I ate at Olive Garden this week (we were craving quick pasta fix), and it explained private equity’s strategy in clinical research better than most investment banking presentations ever could.

Olive Garden is not promising the best Italian meal you have ever had, it certainly was far from that although their lasagna chocolate cake is top tier anywhere ( I challenge you find me better). It is ultimately promising that the meals will not surprise you.

You know the menu. You know the price range. You know the service model. You know that the ingredients, kitchen processes, employee training, and customer experience have all been standardized. If something goes wrong at the local restaurant, there is a larger corporation behind it (for Pete’s sake to channel my inner Robert S. Goldman, I am even able to tag them here on LinkedIn).

It might not be the most memorable meal of your life, but the probability of a complete disaster is low.

That is what pharmaceutical companies are buying when they choose a private equity-backed research site network.

They are buying certainty.

Then there are the Montgomery Biscuits Baseball.

The Montgomery Biscuits are a real minor league baseball team. Even the name tells you something important about the minor leagues. They can be playful. They can experiment. They can build a unique local identity. They do not have to protect a multibillion-dollar global brand the way the New York Yankees or Los Angeles Dodgers do.

Most importantly, minor league teams develop players.

That is what small, independent clinical research sites have historically done for the broader drug-development industry. They are the farm system. They hire raw talent, give inexperienced employees broad exposure to clinical trials, and develop many of the coordinators, monitors, project managers, clinical operations leaders, and sponsor executives who eventually move into the industry’s major leagues (just look at Monica Cuitiva B.S, B.A . as one of many examples).

Small sites can also be weird. They can be creative. They can try things that do not fit neatly into a corporate performance dashboard. They can still have some fun.

Private equity’s opportunity is to make clinical research sites more like Olive Garden.

The industry’s risk is that it eliminates the Montgomery Biscuits along the way.

I Have Seen the Private Equity Process Up Close

A few years ago, my site, Yuma Clinical Trials, got fairly far into a process with a private equity-backed organization.

The transaction ultimately did not happen, but my experience was positive. I liked the people involved. I understood what they were trying to build. I came away from the process believing that private equity’s interest in research sites was not a temporary trend.

Since then, the wave has only grown.

The conversation is usually reduced to a simple question: Is private equity good or bad for clinical research? In my opinion it is a much more nuanced discussion than the good vs bad framing I have been seeing on LinkedIn lately.

There is no universal answer.

The answer changes depending on whether you are a pharmaceutical sponsor, contract research organization, site owner, employee, job seeker, investigator, monitor, or study participant. It also depends on the site’s location, therapeutic focus, competitive environment, and existing level of sophistication. While we are all interconnected, we often have conflicting needs and desires and ideals of what positive outcomes may look like.

Private equity may improve quality, efficiency, scalability, and predictability across the site spectrum.

It may also weaken workforce development, job satisfaction, and career opportunities, reduce local accountability, centralize too much power, and slowly optimize some of the humanity out of clinical research. It may end of creating a new generation of specialists precisely at a time when we are going to need more generalists in the face of the AI revolution.

Both can be true.

Private Equity Is Buying the Bottleneck and Selling Certainty

Private equity firms are not buying clinical research sites because they are fascinated by informed consent forms, source documentation, or investigational product accountability.

They see a fragmented market, inconsistent performance, recurring pharmaceutical demand, and an opportunity to create scale.

There are thousands of research sites operating with different technologies, staffing structures, quality systems, recruitment capabilities, financial models, and levels of operational discipline. Matter of fact, there is a new trend spearheaded by Brad Hightower that are switching back to paper source and regulatory for a variety of different reasons.

Private equity sees an opportunity to acquire profitable sites, combine them into networks, centralize certain functions, reduce duplicated overhead, standardize operations, improve business development, and generate better margins.

It is also investing in the part of clinical development that remains stubbornly physical.

Target identification can become more computational. Molecule screening can be accelerated. Portions of protocol design, biostatistics, remote monitoring, data review, and medical writing can increasingly be supported by software and artificial intelligence.

The research site cannot be fully digitized.

A study participant is not merely a digital data point. The participant has to travel to a location, sign consent, have blood drawn, receive an investigational product, complete assessments, undergo imaging, report adverse events, return for follow-up visits, and remain engaged throughout the protocol.

The information eventually becomes data, but its origin is still physical, local, and human.

That makes the site one of the most persistent bottlenecks in clinical development. Bottlenecks attract investment because they are where urgency, scarcity, variability, and pricing power tend to accumulate.

Some independent sites are excellent.

Others overpromise enrollment, enter data late, communicate poorly, commit avoidable protocol deviations, and require constant intervention from sponsors and contract research organizations.

Private equity recognizes that and sees an opportunity for what they do best: standardize and reduce bottlenecks while realizing productivity gains and ultimately, profits.

All of that can be reduced to one word:

Certainty.

Pharmaceutical companies need confidence that a site will activate on time, recruit the promised population, follow the protocol, enter data quickly, communicate with monitors, maintain appropriate documentation, and avoid producing a regulatory disaster.

For a sponsor, choosing an independent site can resemble selecting a small family-owned restaurant in an unfamiliar city.

It might be the best meal you have ever had. The owner might greet you personally. The service might be exceptional. The food might be prepared with more care than anything available from a national chain.

It could also be a complete disaster.

The sponsor is not merely deciding whether the local restaurant might be better than Olive Garden. It is deciding how much variability it can tolerate during an expensive and strategically important clinical trial.

For a pivotal Phase 3 program, predictable may beat exceptional. (Of course both would be preferred but we need to get real about the state of our industry, LinkedIn platitudes aside).

A private equity-backed site network can centralize business development, contracts, budgets, regulatory submissions, recruitment, data entry, quality assurance, training, technology, and financial reporting. Instead of asking one coordinator to perform six unrelated functions, the organization can create specialized departments with clear responsibilities.

That can produce faster startup timelines, more consistent documentation, cleaner data, stronger oversight, and more reliable performance.

It also gives the sponsor someone higher up the corporate structure to call when something goes wrong.

That is the Olive Garden proposition. The food may not be extraordinary, but the process is repeatable, the experience is familiar, and the risk of a catastrophic surprise feels lower.

Pharmaceutical companies have good reasons to be buyers of this offering.

What the Industry Could Lose

The concern is not that private equity-backed networks will necessarily produce worse clinical research.

In many cases, they will produce better clinical research. For what it’s worth, I think they are elevating the bar to some extent and forcing smaller sites to “up their game” when it comes to quality and consistency aspects.

The concern is that the industry may undervalue functions that small sites have historically performed without being explicitly paid for them.

The biggest example is workforce development.

Small research sites have long functioned as the minor leagues of clinical research.

Pharmaceutical companies and contract research organizations are the major leagues. They offer larger organizations, structured career ladders, more specialized positions, greater geographic mobility, and often better compensation.

But many of the people who eventually succeed at sponsors and contract research organizations learn the fundamentals at small sites.

Independent sites frequently hire people with little or no research experience. They teach them how to communicate with participants, prepare source documents, enter data, respond to queries, manage investigational product, work with investigators, support monitoring visits, recruit patients, identify protocol risks, and understand Good Clinical Practice.

They do this partly because experienced employees are expensive and harder to find. A small site may not be able to hire a coordinator with five years of experience.

But founder-led companies can also be unusually good at spotting raw talent.

A founder may see potential in someone whose résumé would never survive the automated screening process at a large corporation. That person gets hired, trained, exposed to multiple parts of the trial, and gradually becomes a broadly capable clinical research professional.

Eventually, that person may leave for a contract research organization or pharmaceutical sponsor.

The site loses an employee, but the industry gains a trained professional.

That is the unseen value of the farm system.

The Montgomery Biscuits Problem

A minor league player does not develop by performing one narrow function in isolation.

The player learns the game.

That is also what happens at many small research sites.

At an independent site, an entry-level employee may help recruit participants, prepare for visits, organize regulatory documents, enter data, sit through monitoring visits, communicate with investigators, and observe how budgets, contracts, and study startup work.

The employee becomes a generalist.

At a larger network, that same person may be hired into a specific department such as recruitment, data entry, regulatory startup, or patient scheduling.

That specialization can improve operational efficiency. It can also produce an employee who understands one function extremely well without understanding how the entire clinical trial works.

The employee becomes a specialist.

There is nothing inherently wrong with specialists. Clinical research needs them.

The concern is what happens if the industry loses the environment that creates generalists.

If independent sites disappear, stop hiring inexperienced people, or are absorbed into highly compartmentalized networks, where will the next generation of broadly trained clinical research professionals come from?

Private equity-backed networks are not obligated to operate as training academies for the greater good of the industry (shoutout to The CRC Academy). Their responsibility is to build valuable organizations and generate returns for their investors. But the industry should recognize the subsidy they have historically received from small sites.

The site industry has been training future sponsor and contract research organization employees for decades, often without receiving compensation or even recognition for doing it.

If consolidation weakens that farm system, the larger industry may eventually have to build a replacement.

Generalists and Specialists Respond to Different Incentives

Private equity improves businesses through specialization, standardization, measurement, and accountability.

One person handles budgets. Another handles regulatory submissions. Another enters data. Another manages recruitment. Another oversees investigational product. Another manages quality. Another performs business development.

This makes responsibilities easier to define and performance easier to measure.

It also changes how employees think about ownership.

At my site, everyone understands that something can be urgent even when it technically belongs to someone else.

If a monitor needs something, someone takes responsibility.

If a participant has a problem, someone handles it.

If a principal investigator needs a scheduling accommodation, someone figures it out.

If a contract or budget document reaches the wrong employee, that employee still understands where it needs to go and why it matters.

Everyone is partly responsible for everything.

That model is less efficient on paper. It depends heavily on judgment, communication, culture, and individual initiative. It also creates a broader sense of ownership.

In a highly specialized organization, employees naturally prioritize the work attached to their roles and performance metrics. A coordinator may receive a message about a budget issue and consider it someone else’s responsibility. A data-entry employee may have little reason to think about participant retention. A recruitment specialist may not understand what happens after the participant enrolls.

That is not necessarily laziness or indifference.

It is a rational response to the organization’s incentive system.

People focus on what their employer measures and rewards.

Founder Mentality Is Difficult to Institutionalize

If I sold my site and became a local site director inside a large network, I would still be the same person.

But my incentives would change.

Today, I can tell my team to treat participants, monitors, principal investigators, business partners, and employees like gold. I can trust that if we consistently do those things, the financial results will eventually take care of themselves.

I do not have outside investors demanding that I maximize every possible point of margin next quarter.

After an acquisition, I would be evaluated according to the network’s priorities, key performance indicators, financial targets, and operational goals.

I might still care just as much about the monitor who needs help or the participant who is considering withdrawal. But I would also have a dashboard telling me which activities affect the quarter and which ones do not.

Founder mentality is difficult to preserve after the founder no longer owns the business.

That does not mean every founder-led site has an excellent culture. Some independent sites are poorly managed, financially undisciplined, inconsistent, and unpleasant to work with.

It also does not mean employees at large networks do not care. Large organizations contain extraordinary people who go far beyond their formal job descriptions.

The issue is not that small automatically means good or that institutional ownership automatically means bad.

The issue is which behaviors the system rewards.

Clinical Research Has Its Own 50-50 Balls

My brother is a college basketball coach, and he once explained the importance of what coaches call 50-50 balls.

A 50-50 ball is loose. Either team can get it. The outcome comes down to effort and desire. Pure hustle.

It may not appear prominently in the official statistics, but coaches track it because teams that consistently win those moments usually win games.

Clinical research has its own 50-50 balls.

A monitor sends an email late in the afternoon and needs an answer before the next morning.

A participant is thinking about withdrawing and needs someone to spend an extra 20 minutes listening.

A principal investigator needs a last-minute scheduling accommodation.

A coordinator notices something that is not technically a protocol deviation yet but could become one.

A local relationship produces an unexpected referral.

None of these moments will transform a company’s quarterly results by themselves. They may not appear on a formal dashboard. But collectively, they determine whether a site becomes trusted.

Founder-led sites often have a strong incentive to win those 50-50 balls.

When a monitor feels supported, that monitor may recommend the site for another study. When a participant feels respected, that participant may be more likely to remain in the trial, refer friends and family, or return for another study. When an investigator feels valued, that investigator may remain engaged and continue working with the site (not to mention they may refer their colleagues for new PI opportunities).

The return is real, but it is difficult to measure immediately.

Private equity is very good at measuring and optimizing what can be seen.

The risk is that some of the most important parts of clinical research remain partially invisible.

Sponsors Should Be Careful What They Wish For

In the near term, consolidation is attractive to pharmaceutical sponsors.

Instead of managing relationships with hundreds of disconnected sites, a sponsor can develop preferred relationships with a smaller number of networks. It gets standardized contracting, centralized escalation pathways, consolidated performance data, and fewer organizations to evaluate.

The risk appears later.

The contract research organization market offers a possible preview.

As major contract research organizations consolidated, they gained scale, global capabilities, therapeutic expertise, technology, and negotiating power. Large pharmaceutical companies benefited from having sophisticated partners capable of running enormous global programs.

Smaller biotechnology companies sometimes experienced the other side of that consolidation.

They worried that the largest customers received the strongest teams while smaller sponsors received less attention. The organization might have every capability imaginable, but the sponsor’s actual experience still depended on which people were assigned to its study.

The same dynamic could eventually emerge among research site networks.

As certain networks become larger, they may gain more negotiating leverage over budgets, contract terms, technology requirements, and study allocation.

A network that controls a meaningful percentage of enrollment in a therapeutic area is no longer merely a subcontractor. It becomes strategic infrastructure.

Sponsors may accept that risk today because their immediate problem is execution. They need studies started, participants enrolled, data entered, and timelines protected.

Concerns about future concentration can feel theoretical when a pivotal program is delayed right now.

Still, consolidation does not eliminate risk.

It changes the form of the risk.

Private Equity Is Raising the Bar

Despite the concerns, I believe private equity is currently more positive than negative for the research site industry.

Private equity-backed networks are putting pressure on independent sites to respond faster, improve quality, enter data sooner, recruit more effectively, strengthen their systems, and operate with greater discipline.

That pressure can be uncomfortable, but clinical research needed higher operating standards at the site level.

Independent sites can no longer rely on personal relationships and good intentions while ignoring documentation, financial controls, key performance indicators, and operational consistency.

The Olive Gardens are forcing the family-owned restaurants to clean their kitchens.

That is a good thing.

The mistake would be assuming that the only two choices are a sophisticated network or an unsophisticated independent site.

The best independent sites can combine institutional-quality operations with founder-level accountability. We at DSCS Sweat Equity & Investments, LLC dba DSCS CRO are attempting to do this with our 80 plus independent site owners in our subscription model.

The best networks can create standardization without eliminating judgment, creativity, local relationships, and employee ownership.

That should be the goal.

Why This Is Bullish for Site Owners

Private equity’s interest also creates something the research site industry historically lacked: liquidity.

A well-run research site is no longer merely a job that generates income for its owner. It can become a transferable asset with strategic value.

That changes how entrepreneurs think about building sites.

Owners have more reason to document processes, improve financial reporting, develop leadership teams, strengthen investigator relationships, track performance, diversify revenue, and reduce dependence on the founder.

Even sites in competitive markets may become attractive acquisition targets.

If one large network establishes a presence in a market, its competitor may want a presence there too. Strategic rivalry can create additional demand for established local operators.

Sites in less competitive markets may become even more valuable because they provide access to investigators, patient populations, and communities that cannot be replicated simply by opening another office across the street.

Private equity is giving site owners potential exits.

It is also creating much stronger competitors.

Large networks have centralized business-development teams, national sponsor relationships, conference coverage, institutional capital, sophisticated analytics, and access to information that many independent sites do not have.

The smart money often knows which studies are coming long before the average site receives a feasibility questionnaire.

Independent sites cannot beat these networks by pretending to be smaller versions of them.

They have to compete using the advantages that remain difficult to centralize: local relationships, speed of judgment, therapeutic specialization, investigator access, community trust, participant experience, and founder-level accountability.

They have to become the best version of the Montgomery Biscuits.

Private Equity Will Convert Some of the Fun Into Margin

There is one final consequence that is difficult to place on a spreadsheet.

Private equity may take some of the fun out of clinical research.

Founder-led sites still have room to experiment. They can make community videos, interview local businesses, test unconventional recruitment ideas, give employees broader responsibilities, and pursue projects without proving an immediate return on investment.

At my site, we create videos featuring local businesses because we believe community visibility matters. A person may see one of our coordinators interviewing the owner of a coffee shop, recognize that coordinator later in a study advertisement, and feel more comfortable contacting us.

Can I prove the exact financial return from that video?

Probably not, but that is also exactly why that type of branding works.

A private equity-backed organization is more likely to ask whether the activity can be standardized, measured, scaled, and repeated across the network.

That is a reasonable question.

It is also how the fun slowly gets converted into margin.

Private equity finds an inefficient but potentially valuable activity, refines it, measures it, and turns it into a repeatable process.

Founder-led sites often operate at a different stage. They are still expanding their capabilities. They are investing in people and infrastructure before fully understanding how much the organization can handle.

They are building a larger bucket so they can collect more rain later. This is the downward slope of the PE favorite “J curve”.

Private equity generally prefers to acquire a bucket that already works, optimize it, and duplicate it.

Neither approach is morally superior.

They represent different stages, different skills, and different incentives.

But they feel very different.

Clinical Research Needs Olive Garden and the Montgomery Biscuits

Private equity is bringing capital, operational discipline, quality systems, technology, and institutional attention to a part of clinical development that has been underdeveloped for too long.

It is giving site owners potential exits.

It is forcing independent sites to improve.

It is giving pharmaceutical companies more certainty.

Those are meaningful benefits.

The risks are also real.

Too much consolidation could produce pricing power and sponsor dependency. Specialization could weaken the industry’s talent pipeline. Centralized management could dilute local accountability. Dashboards could begin replacing judgment. The behaviors that win clinical research’s 50-50 balls could be overlooked because they do not fit neatly into a quarterly report.

The industry does not need to choose entirely between Olive Garden and the Montgomery Biscuits.

It needs the operational consistency of Olive Garden without destroying the farm system, personality, experimentation, and local ownership represented by the Biscuits.

Independent sites need better systems.

Private equity-backed networks need to preserve some form of founder mentality.

Sponsors need predictable execution, but they should also understand where their future employees, local relationships, participant trust, and operational creativity have historically come from.

Private equity is selling certainty, and pharmaceutical companies have good reasons to buy it.

But standardization should not be confused with excellence.

The real question is not whether private equity will change clinical research sites. It already is.

The question is whether the industry can preserve what the Montgomery Biscuits do best before it standardizes every site into another Olive Garden.

I knew something had turned before I read either earnings release.

You can feel it at the site level first. The feasibility surveys start showing up again. Startup calls get scheduled faster. Monitors start fighting each other for calendar slots. Lab kits pile up in the back. None of that is a data point you can put in a deck, but if you have run a site for any length of time you know the difference between a quiet quarter and a busy one, and it’s starting to get busy again.

Medpace and IQVIA Quarter 2 2026 Earnings Calls Came out (IQVIA’s today) and I listened to them both because Joel White who is the real guru of this niche inspired me. (Catch our last interview discussing Medpace’s results here.)

It’s safe to say both companies crushed earings, beat on every line, raised full year guidance, and reported a bunch of net new clinical bookings. Ari Bousbib from IQVIA told analysts he had never seen a cleaner quarter in more than twenty-five years of reporting earnings, at IQVIA or anywhere else.

For three years the industry looked like a contradiction. CROs cut experienced senior-level people, consolidated functions, and pushed more work onto leaner teams. Everybody I know in clinical research read that either as an industry in decline, or a way for CROs to purge from the COVID-era over-hiring and improve profit margins. Investors read the exact same moves as discipline and rewarded them for it. Turns out, both reads on the situation were correct, and now we may even expect another round of hiring booms in the near future.

The recovery is real, but the old CRO model is not coming back.

The next generation of CRO will be more productive, more technically capable, more transparent, and a lot more selective about who it hires. Sponsors will keep outsourcing trials. They will stop accepting layers of labor, manual trackers, and optimistic promises as proof that the work is getting done. Just ask the “Sponsor Monster” aka Robert S. Goldman how he feels about this.

The future CRO is not smaller, nor is it led by a group of junior CRAs empowered by AI. Rather, it will be one that is more transparent, rooted in context, improved outcomes, and accelerated timelines all directed, in my opinion, by a renaissance of senior-level CRA operators.

Medpace Proved That Execution Still Wins

Medpace is the strongest operating story in this industry and it is not close (as I learned from Joel referenced above).

Medpace grew across the board. $707.3 million in the quarter up 17.2 percent, $1.41 billion through six months up 21.7 percent, and the margin held flat at 21.7 percent the whole way.

The signings were the real story. Medpace sold $795.7 million of new work in three months, 28.2 percent more than the same quarter last year and the second biggest quarter in company history.

Here is why that matters if you are a CRA or a coordinator. Work signed today is the study you get assigned to next year (and increasingly becoming sooner). And they signed more than they burned through. For every dollar of work Medpace actually performed in the quarter, they booked $1.13 of new work, up from $1.03 a year ago. Anything above a dollar means the pipeline is filling faster than it is emptying.

Total work sold but not yet performed now sits at $3.01 billion, and about $1.96 billion of that is expected to get run over the next twelve months. That is the number that eventually turns into protocols, site activations, monitoring visits, and patients in chairs.

Here is the number I keep coming back to. Backlog conversion hit 24.1 percent of beginning backlog, up from 21.2 percent last year.

Conversion rate is how fast contracted work turns into performed work. When it climbs three full points, studies are moving faster through the system. Sites are starting up faster, patients are being enrolled, visits are happening, labs are running, sites are getting paid. That is the site side of the equation, and it is the hardest metric in a CRO filing to manufacture. You cannot book a patient visit that did not occur.

Then there is headcount. Medpace grew employees 7.3 percent to 6,479 while growing revenue 17.2 percent.

That is my whole thesis in one line. The point was never to fire everybody and let software run the trial. August Troendle said the company expects high-single-digit employee growth through 2026 and probably 2027, concentrated in the US and Asia Pacific including India, helped by turnover staying low.

The fastest growing major CRO is hiring because growth is making it hire.

What Medpace Did Not Get Credit For

I want to slow down here because the easy version of this story is wrong, and Troendle said so himself on the call.

The jump in net bookings from a rough first quarter was driven more by lower cancellations than by higher gross bookings. Troendle called cancellations well-behaved and said they had come down to a very good level, though not an unusually low one. He also said cancellations are a wild card the company cannot predict, and that Medpace deliberately limits what it recognizes in backlog when a program has an interim analysis or a regulatory decision that could kill it.

IQVIA Confirmed It Is Not Just Medpace

You could dismiss Medpace as one exceptional company taking share from weaker competitors. That argument gets a lot harder after IQVIA.

Total revenue of $4.368 billion grew 8.7 percent reported. Adjusted Earnings per Share (EPS) of $3.15 rose 12.1 percent. Free cash flow hit $360 million, up 23 percent. Every metric cleared the high end of guidance on operations rather than accounting.

IQVIA’s clinical research arm, the part that actually runs trials, did $2.575 billion in the quarter, up 8.8 percent. Strip out acquisitions and currency swings and it still grew 7 percent, so that is real growth and not accounting.

They signed $3.15 billion of new trial work in the same three months. That is 19.3 percent more than last year and their biggest quarter since 2022. Same math as Medpace. For every dollar of work they performed, they sold $1.22 of new work.

Zoom out and it is not a one quarter fluke. Over the last twelve months they signed $11.25 billion, up 12.9 percent, and that number has now climbed four quarters in a row. Work sold but not yet run totals $34.2 billion, with a record $9.23 billion of it scheduled to get performed over the next year.

Here is the part I would flag for anyone reading the headlines. A big ratio of new work to performed work is easy to post when your business is shrinking, because the number you divide by keeps getting smaller. IQVIA’s ratio went up while the business was also growing near 9 percent. Both sides moved the right way at the same time. Exactly like Medpace. The pipeline is filling faster than two of the largest companies in the industry can work through. (The other 2 major CROs have not reported as of this article).

IQVIA’s CEO was emphatic that nothing about the quarter was engineered. No single chunky award. Pass-throughs in the normal range. Cancellations in the normal range. A good mix of large pharma, midsize, and emerging biopharma, with full-service outsourcing particularly strong.

CROs are not going away.

The Real Engine Is Emerging Biopharma Funding

Emerging biopharma funding hit $35 billion in the second quarter according to BioWorld, more than double the year-ago period. IQVIA also broke out new customer segmentation showing that emerging biopharma is 35 percent of its R&D Solutions revenue, the highest share among the large CROs.

Emerging biopharma trials are effectively 100 percent outsourced. A company with forty employees and one asset does not have an internal clinical operations department. It hires a CRO, and that CRO hires sites like mine.

Medpace tells you the same thing from the other direction. Small biopharma is 84 percent of its revenue.

So the whole recovery traces back to one variable. Small sponsors got funded again. When capital came back, programs restarted, RFPs went out, and the work landed downstream within a few quarters.

That is worth remembering because it cuts both ways. The 2022 through 2024 downturn was never really about CROs. It was about the biotech financing collapse that followed the 2020 and 2021 boom. Small sponsors cut programs, delayed studies, narrowed pipelines, and sat on cash. Then layoffs happened and the industry overcorrected the other way: consolidation, flattened management, more work per person. We are now simply entering the era of the correction for the over-correction and on top of that, catalyzed by improved funding for emerging biopharma.

The Senior Talent Re-Correction Is Coming

The worst mistake CROs made in the downturn was treating experienced people as expensive people.

Senior clinical research professionals carry context you cannot write into an SOP. They know how a sponsor behaves when a timeline slips. They can look at a feasibility forecast and tell you it is fiction. They know which protocol requirement is going to wreck a coordinator’s day for no clinical benefit, and they know which deviations are administrative noise versus a genuine threat to safety or data integrity. Most importantly they know that two problems that look identical on a dashboard can require completely different responses.

Here is where the calls got interesting. Bousbib (IQVIA CEO) said several large pharma clients are asking IQVIA to add thousands of full-time employees, because they expect AI-discovered molecules to roughly double their study portfolios. He pushed back directly on the idea that AI shrinks the CRO. Pharma has been using AI in discovery for two or three years already, and more molecules entering development means more trials, which means more demand for people who can run them.

That is the largest CRO on earth saying its clients are asking it to staff up, in the same quarter it reports AI-driven productivity gains. On top of this, on the same call, Bousbib said large pharma are working with IQVIA on leveraging their proprietary data as the AIs continuously need better data to train on and the Frontier AL labs have already exhausted everything on the internet itself. In short, this means deep, contextual knowledge matters, something exclusive particularly to senior level individuals.

The jobs will be redesigned. The need for judgment survives. I expect the recovery to hit experienced project managers, clinical operations leaders, data managers, regulatory professionals, quality specialists, and CRAs who can work without a babysitter and who by default can train the “last mile” of the AIs starving for this insider knowledge.

The more of this technology I use in my own companies, the more convinced I am that experienced people are becoming more valuable, not less. The tools are genuinely good now. They read protocols, compare documents, draft reports, flag anomalies, and rank risk faster than any human. What they cannot do is tell you whether the output makes operational sense, whether a specific sponsor will accept the fix you are proposing, or who is accountable when a safety call goes wrong.

Someone has to direct the work, check the result, and sign their name to it.

AI Is Already Winning Bids, and Not the Way People Predicted

The lazy prediction is that pharma deploys autonomous systems, pulls clinical operations back in house, and CROs disappear.

IQVIA booked $3.15 billion in new R&D work in three months. Sponsors are not walking away from outsourcing. They are buying more of it.

What AI is actually doing shows up in three specific places in these results.

It is winning competitive bids. Bousbib described a complex Phase III stroke study where the client explicitly credited IQVIA’s AI-enabled site startup and enrollment capability, alongside therapeutic expertise, as the reason it won. Read that again, because it is a site selection story. The differentiator was not headcount or price. It was the ability to find and activate the right sites faster.

And it is scaling commercially. IQVIA has 294 AI agents deployed across 90 use cases, with 19 of the top 20 pharma companies using its AI solutions. One midsize client extended IQVIA AI across an entire immunology franchise in 95 countries.

None of that replaces the CRO. A system does not have a relationship with a principal investigator. It cannot talk a site into prioritizing a query that has been sitting open for three weeks, and it cannot figure out why a coordinator stopped entering data when the real answer is that she gave notice on Friday and nobody told the CRO.

The tool identifies the signal. A person still has to decide what the signal means. As it turns out, the risk was never that technology kills CROs. The risk is that technically capable CROs make labor-heavy competitors economically obsolete.

Sponsors Will Stop Paying Premium Prices for Administration

The traditional CRO model sold hours. Assign project managers, CRAs, data managers, medical writers, regulatory coordinators, and vendor managers. Build trackers. Schedule meetings. Maintain documents. Bill the sponsor for the people required to keep it all upright.

That model is under real pressure now. Sponsors have stopped believing that a status meeting is a deliverable. Protocol analysis, feasibility review, document QC, TMF inspection, medical writing, query generation, pre-screening, budget and contract workflow: software touches all of it. The functions do not disappear. What disappears is the premium for doing them by hand.

Sponsors will still pay up for results. Faster startup. Enrollment that actually happens. Site relationships. Clean data and a database lock that holds. Regulatory defensibility. Real visibility into what is happening across the study instead of a slide deck describing what happened six weeks ago.

There is a commercial tailwind under this too. New drug launches ran 45 percent higher in the first half of 2026 than the prior year, which is why IQVIA’s Commercial Solutions organic growth hit 5 percent, its best in over a year. Launch spending concentrates in the first two years after approval. More approvals now means more commercial work later, and it means the trials that produced those approvals were running two and three years ago. The cycle is turning at both ends and sponsors that needed the R&D help also need the commercialization help which includes physician relationships fostered via clinical research. IQVIA is the CRO leading this category by far.

The Site Is Still the Hardest Part

No technology removes the site. We identify the patients. We consent them. We run the visits, enter the data, report the safety events, answer the queries, and implement the amendments. I have been banging on this drum for two decades now. The protocol is a document until a coordinator turns it into a patient encounter.

That makes site performance the most valuable asset a CRO can hold, and the industry still selects sites almost entirely on vibes.

Here is how it works. A sponsor sends a feasibility survey asking how many patients we can enroll. Every site owner in the country knows the honest answer and the answer that keeps you in consideration are not the same number. Everybody inflates a little, everybody knows everybody is inflating, and then the study plan gets built on the inflated numbers. Six months later there is a rescue meeting about why enrollment is behind.

That is exactly what IQVIA just monetized. When a sponsor awards a Phase III stroke study and says out loud that AI-enabled site startup and enrollment capability won it, the sponsor is buying its way out of the feasibility survey problem. It is paying for a CRO that knows which sites actually activate and enroll rather than which sites claim they will.

The large CROs have had years of this data sitting in their systems. They have always known which sites miss deadlines and which investigators deliver. What changed is that it is now a line item in a competitive bid.

Why Site Payments and Site-Centricity Matters

Every company tracks how long it waits to get paid. Do the work, send the invoice, count the days until the cash actually lands. Most businesses sit somewhere between 30 and 90 days, and lower is better. The finance term is days sales outstanding.

Medpace’s number is negative 59.6 days.

Negative. That means that the cash shows up about two months before the work gets performed, because sponsors fund them in advance.

That is normal in this industry and it is smart business. It is also the exact opposite of how money reaches a site. We run the visit. We enter the data. We wait for the CRO’s payment cycle to come around. Then we wait another 60 to 90 days after that, sometimes longer.

The same dollar is prepaid to the CRO and paid in late (often very late) to the site. Medpace ended the quarter with $502.7 million in cash and bought back $294.7 million of its own stock. That is a well-run company doing exactly what it should. It also happens to be running on float that originates in work my staff has not been paid for yet.

I am not arguing that is a scandal. I am arguing that site-level working capital is the least examined pressure point in clinical research, and it gets worse in an upcycle, not better. By the way, I am not alone in having this sentiment. Just attend any SOS Save Our Sites event and hear it from others (often more vocal and irate). More visits are happening. Sites are fronting more cash, faster, at exactly the moment they are also being asked to add staff.

As the upstream systems get faster, the site becomes the visible bottleneck. A platform can generate a query in a second. A coordinator still has to answer it. A model can flag underperformance. A CRA or PM still has to drive out there and fix it.

The CROs with real site relationships will own the hardest part of clinical development for a long time. The ones who also fix site payment terms will get first call on capacity when the next crunch hits.

Risk-Based Quality Has to Become Real

We have been talking about risk-based quality management for years. In practice a lot of studies still run on heavy monitoring, manual trackers, retrospective deviation review, and documentation produced mostly to prove documentation was produced.

The modern version focuses oversight on what can actually affect patient safety, participant rights, and the reliability of the result. The technology finally makes it possible to catch delayed data entry, repeat deviations, weird query patterns, missing safety information, and inconsistent performance early enough to matter.

Identifying risk is not managing risk. The quality of the whole system depends on where the thresholds are set, what data feeds the model, and whether the person reading the output knows what they are looking at. A junior employee can follow an escalation procedure. An experienced one knows whether escalating is the right move, how fast it needs to happen, and what it will break somewhere else in the study.

Risk-based quality should cut monitoring waste. It should not become the excuse for removing the people with judgment.

The Opportunity Past Phase 3

The CRO opportunity keeps expanding past traditional Phase 2 and Phase 3 work.

Sponsors need real-world evidence, EHR data, claims, registries, pragmatic studies, long-term follow-up, external controls, and post-market evidence. The strong CROs will generate evidence across the entire product lifecycle rather than closing out one study and moving on.

They are also becoming policy players, which is newer than most people realize. IQVIA was the only CRO invited to the HHS clinical trials roundtable under Operation Trailblazer and the only industry representative to testify before the House Energy and Commerce Subcommittee on accelerating early clinical development. When one vendor is in the room while the rules for faster trials get written, that is a competitive position, not a press release.

The rest of us should be paying attention to what comes out of those rooms, because site-level requirements are downstream of them.

Sponsors also need compliance translators. The regulatory environment gets more complicated every year as trials add eConsent, telehealth, home visits, remote staff, wearables, decentralized procedures, and half a dozen data sources. International studies stack on privacy rules, local submissions, and regional technology standards. Nobody needs a vendor that forwards the new guidance document with FYI in the subject line. Sponsors need a partner who can explain what the requirement does to the protocol, the consent, the vendor plan, the monitoring model, the data flow, investigator training, and the documentation trail.

Converting regulatory language into operational execution is a skill, and it still commands premium pricing.

Who Wins

The large CROs with global infrastructure, real data assets, and deep site relationships stay powerful. IQVIA, ICON, PPD, Parexel, Syneos, and Medpace hold capabilities that are genuinely hard to reproduce.

Size alone guarantees nothing. Medpace proves a focused CRO can outrun much larger competitors on organic growth, disciplined hiring, cash generation, owned technology, and consistent delivery. IQVIA proves scale still wins bids when it is attached to data and working AI rather than headcount.

Specialty CROs get an opening too. Real depth in oncology, rare disease, CNS, metabolic, dermatology, vaccines, women’s health, or real-world evidence beats generic scale for the sponsors who need that depth. Site connected CROs and clinical platforms compete on direct investigator access, actual performance data, and faster study startup.

The losers are just as easy to name. Labor-only staffing shops with no technology and no therapeutic depth get squeezed on price. CROs that cannot prove enrollment lose credibility the first time a sponsor asks for the history. Weak quality systems break under the scrutiny that is coming. And any CRO carrying a backlog full of inactive programs is now carrying a number that analysts have learned to ask about.

The old model stays profitable for a while. The margins get harder to defend every year.

The CRO of 2030

By 2030 there will be a clean split between CROs that modernized and CROs that stayed dependent on generic labor.

The strong ones will be connected directly to sites, running modern technology, deep in specific therapeutic areas, and able to sell a whole program or a single function. They will pair execution with live data, measurable site performance, and risk-based quality that actually works. They will help sponsors generate evidence across the lifecycle instead of finishing one study and starting the sales cycle over. They will cut administrative cost while paying up for the experienced people who make the hard calls.

Two companies just told us the demand is there. Medpace grew revenue 17.2 percent while adding 7.3 percent headcount and raised guidance. IQVIA booked $3.15 billion of new clinical work in ninety days, said its clients want it to hire thousands more people because AI is about to double their pipelines, and called it the cleanest quarter in twenty-five years.

That is not technology eliminating clinical research jobs. That is technology and discipline letting companies grow faster than their payroll while still hiring while the sites front the costs (need a good Brad Hightower rant here).

The industry is booming.

On Thursday, July 16, 2026, a company most people associate with betting on presidential elections and the high temperature at the Phoenix airport quietly opened a new window into an industry that has spent decades trying to earn the public’s trust. Kalshi, one of the largest federally regulated prediction market exchanges in the country, began letting ordinary people wager real money on selected late-stage clinical trials and FDA decisions: whether a Phase 3 trial hits its endpoint, and whether and when a drug gets approved. By reporting from STAT and Bloomberg, the exchange had taken in more than $100,000 in bets shortly after launch.

Crazy right? You can now buy a contract that pays out based on whether a cancer drug hits its primary endpoint. The same tap you use to bet on France in the third-place game of the World Cup now offers you a position on a Phase 3 cardiovascular outcomes study enrolling thousands of real patients.

I run a clinical trial site. Yuma Clinical Trials does exactly the kind of studies Kalshi is now turning into betting lines: the multi-center Phase 3 programs that decide whether a molecule becomes medicine. So when I saw this news, my first reaction was not outrage. It was confusion. I am, by temperament, a libertarian and an optimist. I generally believe companies should be allowed to grow and innovate without a regulator standing on their neck. I bet on prediction markets myself. I use Robinhood for the World Cup, and I am net up for what it’s worth (and for those who may be curious, Go France and Argentina). And yet I sat on my morning livestream and talked myself, out loud and in real time, into the conclusion that this particular product should not exist in its current form. Here is how I got there, and, more importantly, how I think the industry should respond without waiting on Washington.

What Kalshi actually built

Kalshi did not do this alone. It partnered with a company called AppliedXL, a public-intelligence firm that tracks and forecasts clinical trial outcomes, to build the resolution machinery underneath these markets. That word, resolution, matters. Every contract has to settle against something. Each contract establishes in advance the public source, or hierarchy of sources, that will determine settlement. Depending on the contract, that may include ClinicalTrials.gov results, FDA records, advisory-committee materials, or specified company disclosures. In other words, the bet does not resolve on a rumor or a stock chart. It resolves on auditable public evidence.

Kalshi has also tried to build guardrails. The company says it is taking a conservative approach, focusing the pilot on selected late-stage trials and FDA decisions. It requires employment verification for traders. And critically, for the initial pilot, the framework generally favors listing trial-outcome contracts only after enrollment has closed, a curation principle meant to keep site staff and coordinators from trading on what they see while patients are still coming through the door.

That is a thoughtful design. It is also, in my judgment, a design that underestimates the nature of the information it is trying to fence off. Because the problem with betting markets built on top of contests where insiders exist is not theoretical. We have watched it play out, on national television, in a league with far more surveillance resources and positive community sentiment than clinical research has ever had.

The scandal the NBA almost made disappear

Cast your mind back to October 23, 2025. Federal authorities announced related basketball-betting and Mafia-linked poker cases involving more than 30 defendants. Among them were a Hall of Famer and then head coach of the Portland Trail Blazers, Chauncey Billups; a then Miami Heat guard, Terry Rozier; and a former player, Damon Jones. Both Billups and Rozier were placed on administrative leave and both denied the allegations. Their situations have since moved on: the Heat waived Rozier in April 2026, the Trail Blazers hired Micah Nori as head coach on June 23, 2026, and Billups is facing trial, with proceedings set to begin around November 2, 2026. I want to be careful and accurate here, because on my livestream I said from memory that Billups “went to jail,” and that is not right. He was arrested, indicted, and has denied wrongdoing, and the case is still pending.

The mechanics of the Rozier allegations are what should chill anyone in my industry. Prosecutors allege that in a March 23, 2023 game, Rozier let an associate know in advance that he planned to leave early, played about nine minutes, and enabled more than $200,000 in prop bets on his underperformance. A superseding indictment filed on May 28, 2026 further alleged that he agreed to accept a $100,000 bribe, reportedly reduced to about $70,000 after some wagers lost. These remain allegations that Rozier denies. The Billups scheme was different in flavor, allegedly involving Mafia-linked rigged poker games using chip-tray analyzers and doctored shuffling machines, but the theme is the same: people with private knowledge monetizing it through a betting product that assumed everyone was guessing.

Here is the structural point, stated generally rather than as a claim about any specific play. In prop-bet schemes, a participant does not need to change who wins the game. The outcome fans care about can stay untouched while a side wager, a player’s minutes, a point total, a spread, gets quietly steered. The thing the public watches and the thing the bettor cares about are not the same thing, and that gap is where the money hides.

And how do these things surface? Often not because the leagues have flawless surveillance. Go back to Tim Donaghy, the referee at the center of the 2007 NBA scandal. Donaghy admitted to betting on games, including games he officiated, and to using confidential information. The FBI uncovered his conduct while investigating a broader organized-crime gambling operation. The NBA did not detect the scheme first. Donaghy pleaded guilty in August 2007 and received a 15-month federal prison sentence. Betting-integrity monitors do flag statistical anomalies, and unusual wagering patterns are a real detection tool, but the lesson of Donaghy is sobering: in the most watched, most surveilled entertainment product in America, an insider operated across seasons before the catching happened, and it happened through a federal investigation, not the industry’s own systems.

Now transplant that structure onto an industry with a fraction of the surveillance, orders of magnitude more human complexity, and actual lives on the line, as well as an already large public distrust for the majority of the stakeholders, and you have a recipe for permanently damaging the foundation that our industry is built upon.

Clinical research already has this problem. Kalshi just widens the door.

Trading on nonpublic trial data is not a hypothetical we are inventing for a thought experiment. The SEC has been prosecuting it for years, and the cases read like a warning label for exactly what Kalshi is now building.

Consider Daniel Catenacci, a medical school professor who served as lead clinical investigator on Five Prime Therapeutics’ trial of the cancer drug bemarituzumab. According to the SEC, shortly after learning the results were positive, he bought 8,743 shares ahead of the November 2020 announcement, and the stock later rose more than 300 percent. Or Dr. Sai-Hong Ignatius Ou, who the SEC says bought 80,000 shares of Nuvalent on nonpublic knowledge of promising cancer-drug results and benefited by roughly $1.52 million when the stock climbed about 60 percent. Or Dr. Mohammed Bari, a medical investigator on Karuna Therapeutics’ KarXT trial, who prosecutors say began placing share orders within hours of learning the drug worked and earned nearly $120,000. The pattern did not stop: in January 2026, the SEC charged consultant Hong “John” Wang over trading connected to confidential C4 Therapeutics trial information, and it previously charged former Pfizer statistician Amit Dagar in connection with advance knowledge of positive Paxlovid trial results.

Notice what those cases had in common. They involved securities trades routed through brokerage accounts, creating the conventional account, timing, and communications records that the SEC and federal prosecutors routinely investigate. That friction is not a bug. It is most of what deters the marginal insider. The financial disclosure forms we make investigators sign exist precisely because holding stock in the company sponsoring your study is a flashing red flag that regulators know how to read.

A Kalshi contract sits one step to the side of all of that. To be precise about the enforcement picture, these event contracts trade on a CFTC-regulated exchange, not a securities exchange. A suspicious Kalshi trade would more naturally involve Kalshi’s own surveillance, the CFTC, and potentially the Department of Justice, with the SEC becoming relevant only if the person also traded the sponsor’s stock or another security. The CFTC has stated plainly that misuse of material nonpublic information in event-contract markets falls within its enforcement authority under Section 6(c)(1) and Regulation 180.1, and it has already brought cases against Kalshi traders who used nonpublic information. So the tools exist. But it is a newer enforcement environment with less developed precedent than conventional securities markets, and the guardrails are aimed mostly at the obvious insider, the coordinator watching her own eight patients. They do far less against the person the modern site landscape has quietly created: the site-network executive.

The consolidation problem nobody is pricing in

When I started in this business, betting on a trial from inside a single site would have been foolish. A small site like mine might enroll eight or ten patients in a study that runs across 50, 80, 120 locations and a thousand or more subjects. From that vantage point you know almost nothing. You might convince yourself a patient is improving, or guess from the gastrointestinal side effects that someone is on the active drug in an obesity study, but it is anecdote. Two data points out of a thousand is noise, not a signal.

That world is disappearing. Site networks have been consolidating aggressively, backed increasingly by private equity. At an investigator meeting I attended about three years ago for a large international Phase 3 cardiovascular and obesity study with more than 120 sites, roughly half the room belonged to site networks, and not even the same one. Three separate networks were represented. That consolidation has only accelerated since.

Now redo the math from the top of a network. A director overseeing 20 sites in a single study is no longer looking at two patients. That person may have visibility into 15 or 20 percent of the entire dataset, streamed through the exact analytics dashboards these networks are building or buying to run their business. In an obesity trial, where the active drug tends to announce itself through weight loss and GI symptoms, a sophisticated observer with a fifth of the data does not have anecdote anymore. That person has a signal. Not certainty, but a plausible, tradable edge.

But wait, there’s more and it comes in the form of a governance gap. FDA financial-disclosure requirements generally cover investigators and subinvestigators directly involved in treating or evaluating subjects. They do not automatically encompass every coordinator, regional executive, or site-network employee who may gain access to aggregated operational data. A network executive watching dashboards could therefore fall outside that specific framework unless sponsor policies or contractual confidentiality rules separately cover the role. They have more visibility than the coordinator and less paperwork. That is not a loophole Kalshi created or is probably even aware of for that matter. It is one that consolidation created, and that a betting market now makes lucrative to exploit.

The angle almost nobody is talking about: the participants themselves

There is a second failure mode, and it may matter more than the insider one, because it corrupts the science rather than just skimming money off the top. What happens when the people being studied can bet on the study?

Blinding is the load-bearing wall of a rigorous trial. Neither the patient nor the evaluator is supposed to know who got the drug and who got placebo, because the moment either one knows, expectation contaminates the result. Loss of blinding can introduce reporting and performance bias, particularly when outcomes depend on subjective symptom ratings. Participants already compare symptoms and perceived assignments online, in trial forums and social media, out of hope and curiosity. From as far back as I can remember, Sponsors have been legitimately concerned with “empowered patients” breaking their own study blinds by using online forums and social media channels.

Now imagine giving participants a stake in the overall result. That could create an additional incentive to guess their assignment, to coordinate, or to alter their behavior, adherence, symptom reporting, whether they stay in the study at all. A participant who is long on success and believes he is on placebo has a reason to drop out or to shade his diary. A participant who is short has the opposite reason. I want to be clear that this remains a hypothetical risk until evidence emerges from these markets. But it is exactly the risk that bioethicists, physicians, and trial researchers raised to STAT: financial incentives could push people with influence over a study to alter or delay outcomes, shifting the goal from discovering accurate medical evidence toward moving a market. Once that is true even a little, every clean result is suspect too.

The fix should be private and fast, not federal and slow

Here is where my libertarian instincts and my worry finally reconcile, and where I part ways with the people already calling for Congress or the FDA to act. I do not think federal rulemaking is the right first move, and I am not sure it is necessary at all. Formal regulation is slow. By the time a rule works its way through notice, comment, and litigation, the pilot will have grown, mutated, or blown up on its own. Rulemaking is also blunt: written broadly enough to matter, it tends to catch the wrong people and slow down good science.

The faster and, I think, better tool already lives inside our own industry. Sponsors do not need to wait for a regulator to tell them what to do. They can ban this at the point of the contract. We already make investigators sign financial disclosure forms, a self-disclose-and-self-police mechanism that works precisely because the person attaches their name and their liability to a promise. Extend that logic. A sponsor can require that everyone with access to aggregated trial data, not just the investigators and subinvestigators the FDA framework reaches, but coordinators, regional managers, network executives, dashboard administrators, and analytics vendors, sign a plain-language attestation: I will not hold a position, directly or indirectly, in any prediction market or derivative tied to this study’s outcome, and I understand that if I do, I can be held personally liable. Make it a condition of access to the data. Make the penalty real, contractual, and enforceable in the same way a breach of confidentiality already is.

This is not a fantasy. It is the financial disclosure form doing what it was designed to do, pointed at a new derivative that the current form does not name. It is faster than legislation because a sponsor can add it to a clinical trial agreement tomorrow. It is more precisely targeted than a federal rule because the sponsor knows exactly who touches its data. And it keeps the policing inside the relationship where the duty of trust already exists, rather than waiting for the CFTC or the DOJ to reconstruct a paper trail after the damage is done. Sponsors ultimately make the rules even if the FDA may offer guidances. Take for example the requirement by Sponsors (not the FDA) for PIs to be licensed physicians. True, there is no federal mandate that states what license a PI should hold, but no sponsor in their right mind would choose a non-MD/DO individual to be a PI for one of their studies. Problem solved. Now apply that same pragmatism here.

I will be honest about the cost. If you think study startup is slow now, imagine collecting attestations from every employee of a private-equity-backed network before a single patient enrolls. It will add friction. But it is friction the industry controls, and it is a great deal cheaper than the alternative.

Why the alternative is so expensive

I keep asking a question Kalshi has not answered: who is this even for? In my estimation, the market for betting on a single Phase 3 readout is tiny next to sports, elections, and the weather. The mainstream audience that drives prediction markets does not know what a PCSK9 inhibitor is, let alone how to price the probability that one clears its endpoint. This launch does not look to me like a business. It looks like a headline. Which means, in my view, the exchange is taking on real reputational and regulatory risk for a market that cannot move its revenue.

I think a scandal is a likely outcome here if given enough time, not a distant theoretical risk. You do not need an NBA-style conspiracy. You need one coordinator or network executive to be caught by Kalshi, the CFTC, or federal prosecutors trading on confidential trial information. The NBA survived its scandals in part because it sells entertainment and its public relations operation is genuinely world class, and viewership barely moved. Not to mention that gambling has always been associated in some way with sports, again the main product is entertainment. Clinical research does not have that luxury. We went into Covid with a chance to be the heroes and, for a lot of reasons, squandered most of the trust we briefly earned. We enroll patients on the strength of their belief that the people running the study are trying to find the truth, not trying to cash a ticket. A clinical-research version of the NBA betting scandal would not read as sports-page gossip. It would read as proof that the industry games its own results, and it would land at the exact place we can least afford it: the moment we ask a patient to trust us enough to enroll.

I came into this as the last person who wants more government. I still am. That is why my answer is not a new federal statute but a stronger signature line: let every sponsor ban this itself, in the contract, with personal liability attached, and let the people closest to the data put their name on a promise before they ever get near the dashboard. The NBA showed us what happens when you build a betting market on top of a contest that has insiders. The only real question left is whether we fix it ourselves, quickly and privately, or wait for someone to learn the lesson again, this time with all of our (real) futures at stake.

Retail loses money on FDA decisions for one reason. They trade the headline instead of the trial.

They see the word. They do not read the table. And on a biotech catalyst, the gap between the word and the table is the entire game.

I do not read these from a chart. I run clinical research sites. That is my actual job. I read the protocols, the trial results, the FDA review documents, the rejection letters most people never open. So when a biotech has a catalyst coming, I am reading the same data the reviewer is reading, not a press release about it.

Right now the catalyst is Celcuity, ticker CELC, and its drug gedatolisib. The FDA decides on July 17. The internet wants to compress that into approve-up, reject-down. That framing is exactly how people get hurt. Let me walk you through how I actually read it, because the three checks below work on every FDA date, not just this one.

The setup nobody slows down for

Gedatolisib is for HR-positive, HER2-negative advanced breast cancer, the most common flavor of the disease. Instead of hitting one node of the PI3K pathway like the older drugs, it blockades the whole thing. Comprehensive blockade is the pitch, and it works whether or not the patient carries the PIK3CA mutation.

Here is the part almost every recap gets sloppy about. The July 17 decision is specifically for the PIK3CA wild-type population. The data that moved the stock back in June was from a different group, the mutant cohort. So CELC is not trading one event. It is trading the wild-type application in front of the FDA, plus the newer mutant data from ASCO, plus the future label, plus the next filing, all at the same time.

That is what a real catalyst looks like. The stock ran to about a hundred fifty-one in May, got cut down after the June data, and now sits in the low one-teens going into the decision. Three checks. Steal the framework.

Check one. Did the headline match the table?

On June 2, Celcuity presented its mutant-cohort data at ASCO. The stock dropped more than twenty percent in a single day. Every recap said the same thing, that the number came in light versus the whisper. Partly true. But the thing that actually moved the tape was a headline about patient deaths in the trial.

Now read the table instead of reacting to the word.

In that June mutant readout, there was one treatment-related death on the three-drug regimen. The company tied it to palbociclib, the older CDK4/6 drug in the combo. There were two treatment-related deaths on the competitor’s arm, alpelisib plus fulvestrant. And there were zero on the two-drug version of gedatolisib.

Sit with that for a second. The word that crashed the stock points straight at the one ingredient the company can drop. The clean doublet killed no one. So the headline that scared retail out of the position is actually the clue for why the two-drug regimen may be the better commercial story. That is check one. The headline is the reaction. The table is the truth. They are almost never the same thing.

One honesty note, because I would rather you hear it from me than get nitpicked in the comments. That clean death table is from the mutant cohort presented in June. The wild-type cohort the FDA is actually reviewing had two treatment-related Grade 5 events in the triplet, one tied to palbociclib and one a hepatic failure assessed as related to all three drugs. The point survives. The doublet is the cleaner story, and the palbociclib signal is the reason. But do not let anyone tell you the filing under review has a spotless triplet arm. Read the specific cohort, every time.

Check two. Who is positioned against it?

Before any catalyst, find out who is betting against it. Smart money leaves footprints.

On CELC, roughly a quarter of the tradable float is sold short going into a decision the crowd calls a near-certain approval. That is not background noise. That is a heavy, deliberate bet against a supposedly sure thing. Latest read is about 10.67 million shares short, up double digits from the prior report.

Here is the mistake retail makes with that number. They hear high short interest and their brain screams short squeeze. Maybe. But high short interest does not tell you direction. It tells you violence.

If the FDA gives a clean approval with a clean label, those shorts become forced buyers, and that can send the stock through levels people assumed were already priced in. If the label comes in narrow or messy, those same shorts get their confirmation and the downside accelerates. Either way, a short pile that size means the move is going to be sharp. So treat it as a position-size warning, not a buy signal. That is check two.

Check three. What is the label going to say?

This is the one that matters most, and it is the one nobody prices. Approval is not a yes or no. It is a paragraph. The paragraph is where the risk lives.

Look at the wild-type data the FDA is reviewing. In the full trial population, the gedatolisib triplet delayed the cancer about 9.3 months versus 2.0 months for the control. Strong. But in a subgroup, the patients enrolled in the US, Canada, Western Europe, and Asia Pacific, that same triplet posted 16.6 months. That is a 7.3-month spread on the same drug in the same trial, driven by where patients were enrolled. On top of that, the control arm ran only about 2.0 months, which is low for this setting, and a low bar makes any drug look like a superstar.

None of that stops an approval. But those are the exact lines an FDA reviewer circles.

So picture the outcome almost nobody is modeling. The FDA approves it, but writes a narrower label. Maybe it carves the population. Maybe it hedges the combination or the prior-therapy positioning. Technically the company won. The headline says approved. And the stock still sells off, because everyone was trading the simple approve-equals-up wiring, and a narrow label breaks that wiring. Approved, and down.

You only see that coming if you read the label risk instead of the press release. That is check three. And it is not just my theory. The sharper PDUFA accounts on X have already stopped asking whether it gets approved and started asking what the Day 0 label will let investors believe. That is the trade.

The call

If you put a gun to my head on the outcome, here is my split. Around 88 percent full approval with a clean label. About 9 percent approved but narrow. About 3 percent outright rejection.

I do not expect a rejection, and the reason is structural. This went through the FDA’s Real-Time Oncology Review, the collaborative process where the agency and the sponsor are already deep in the data. Add Breakthrough Therapy, Fast Track, and Priority Review. You do not sit in that lane a week out if there is a bomb in the file. I am also not expecting a conditional or accelerated approval, because everything about this points to a standard full approval. Randomized Phase 3, met its primary endpoint on progression-free survival.

So the risk that should keep a shareholder up at night is not the rejection. It is that narrow-label branch. That is the outcome the market is least prepared for, because everyone is wired for the binary.

On where it trades. If I have to commit to one number on a clean approval with an acceptable label, I think it settles somewhere around a hundred thirty-eight in the first day or two. Not the wick. The settle. I would expect a gap open near a hundred thirty, a squeeze toward the old highs around a hundred forty-five to a hundred fifty as the shorts cover, then the anticipation buyers taking profits and easing it back into the high one-thirties. The Street average sits up around a hundred fifty to a hundred sixty, but that is a grind-toward-over-weeks number, not a day-one number, because approval was expected and the commercial launch still has to prove out.

And if the label comes in narrow, flip the whole thing. You can retest ninety to a hundred even on a technical yes. Same event. Opposite chart. One paragraph decides which.

What I would actually do with it

I like the science. Comprehensive pathway blockade is a real idea, and the VIKTORIA-1 data is genuinely strong. But at this price, this is a catalyst trade with a bullish lean, not free money.

The edge was never the approval. The approval is mostly priced. The edge is the death headline everyone misread, the short pile nobody is respecting, and the label language nobody is modeling. The best risk-reward on this name was in the mid-June low, not here in the one-teens with the whole crowd leaning the same way.

So I would size it like the binary it is. If a thirty percent overnight move against you would end your account, your position is too big. That is the entire lesson. Not the ticker. The sizing.

The one line

If you take nothing else from this, take the line I am going to repeat on every catalyst I break down.

Approval is not a yes or no. It is a paragraph.

Read the paragraph. That is the whole job, and it is the one thing the chart guys and the hype accounts cannot do for you. I run this framework on the next FDA date, and the one after that, before they happen. If that is useful, that is what this is for.

Not financial advice. This is educational analysis from a clinical research operator, not a licensed advisor. I may hold positions in the securities discussed. Biotech catalysts are high-risk and binary. Do your own research and manage your own risk.

Sources: Celcuity investor releases, the VIKTORIA-1 wild-type publication in the Journal of Clinical Oncology, the June 2 ASCO mutant-cohort presentation, FDA program designations, and MarketBeat short-interest data. Figures current as of early July 2026 and subject to change into the July 17 decision.

Subscribe now

I posted a short last week on Youtube and Instagram that I knew would be controversial. One stat: 60% of monitoring tasks are predicted to be automated by 2028. That is two years from now. The comments went where you would expect. Will CRAs still exist?

Short answer: yes. Longer answer: we have been here before (remember risk based monitoring?), and the people panicking are misreading what just happened, exactly as they did back in 2016.

The 2016 precedent

Ten years ago I did a video with Darshan Kulkarni (him and I go way back, pre-Baba yaga days) about risk-based monitoring. The fear back then was identical. If we are no longer doing 100% SDV, why would sponsors pay for as many monitors? Then COVID hit, zero interest rates made money free, vaccine research exploded into a broader trial boom, and by 2021 you were getting hired if you had any resemblance to a CRA. The industry could not find enough monitors.

The fear was rational. The prediction was wrong. Because it assumed the number of trials stays fixed while the cost per trial drops due to know standard monitoring tactics such as risk-based monitoring. We can go back to around 2005, when electronic data capture really became the new standard. I personally remember CRAs back then being fearful that EDC would replace their jobs. That is not how this industry works.

Jevons paradox, again

When coal got cheaper and engines got more efficient, England did not use less coal. It used more, because efficiency unlocked new applications. Replace coal with intelligence. Agentic AI doing the grunt work of monitoring is orders of magnitude cheaper than $120 an hour in labor. If the cost of running a trial drops, sponsors do not pocket the savings and go home. They run more trials. Listen to Dave Ricks at Eli Lilly and Company talk about pipeline decisions. Sponsors deprioritize programs because of site and CRO bottlenecks, not because of ambition. Cut the cost of the low-value layer and the math changes on CRO profits (their bottleneck), and they are incentivized to take on a higher volume of trials.

And if the thesis plays out that AI and venture money flow into biotech next, there is a real world where we run out of good CRAs instead of having too many even though it doesn’t seem that way due to the hesitation for CROs to hire senior level CRAs due to the cost-prohibitive nature of their expected salaries. I predict this awkward period will eventually thaw itself out and CROs should start hiring senior level people again IF the pre-clinical AI discovery promises AND the money starts flowing back into the sector in larger volumes.

What actually gets automated

Let me be honest about the task layer, because I have been on record about this for years. Source data verification is not high-value work. Does the date in the database match the physician note? Does the lab value match the EDC and source? A high schooler can do simple SDV if the source is in one place and clear. Oncology is harder because the source is fragmented across progress notes, labs, and imaging, and oncology is our largest segment. Fine. Solve the easy indications first.

AI can already read documents, compare fields, flag mismatches, get trained on a protocol, and surface deviations. That middle layer of the workflow is repetitive, routine, and repeatable. That is exactly what technology attacks first. It is already shrinking.

What does not get automated

AI does not sign the monitoring visit log. It does not sign the follow-up letter. It does not attest that the data is reliable and it does not appear before regulators. Ultimately it does not take responsibility. The FDA has already shown its hand here. Earlier this year, a sponsor that outsourced its entire GMP function to AI got a warning letter (ask Darshan Kulkarni who that was I cannot recall and the specifics are not the prerogative of this article). The precedent is clear: humans carry the accountability, not the vendors and not the model.

And someone still has to determine ground truth. When the PI note says one thing, the hospital record says another, the patient diary says a third thing, and the EDC says a fourth, who decides which version is reality? At the site, that is the PI. At the sponsor, that responsibility is delegated to the monitor. The whole industry runs on it. Not to mention keeping sites motivated, accountable, and on-time. This is human-only level of work.

The best monitors were never valuable because they compared fields well. They are valuable because they spot inconsistencies before they become findings, see workflow breakdowns, catch training gaps and fill them, and occasionally smell misconduct. That is contextual reasoning, not automation.

The rules and the exceptions

My dad is a psychiatrist. He told me once that in medical school you learn the rules, and in practice you learn the exceptions. Monitoring is heading the same way. The rules are the routine tasks. Those go to the machines. The exceptions are the high-risk, high-consequence scenarios that need a human in the loop. Less clerical, more forensic.

So what do you do about it

If you are a CRA, the threat is not AI. The threat is your own skillset. If your job description is “I manually compare information,” you are a task-doer, and task-doers are vulnerable. Have an AI audit how you actually work. Where does your time go. Learn risk-based monitoring properly if nobody taught you. Learn data provenance: where data originated, how it changed, who touched it. Read the peer-reviewed literature in your therapeutic areas until you can hold your own. Become the person who understands how the whole ecosystem fits together.

AI does not eliminate monitors. It exposes where the industry was paying for busy work instead of judgment. The busy work now costs tokens. The judgment still costs $120 an hour, and it will be worth every dollar.

I am bullish on this industry. Get out there and go to work.

Subscribe now

I have spent 21 years running clinical research sites. I have also spent the last few years, out of necessity, watching the same consolidation movie play out in private medical practice that is now playing out in our industry. The medical practice version is a leading indicator and we can try to predict the future of smaller research sites with some degree of certainty.

Here is the question I keep coming back to. Is there hope for an ecosystem with small independent research sites, or are we all going to be absorbed into Headlands, Velocity, CenExel, Flourish, Javara, and Evolution over the next decade?

I see two possible outcomes based on my analysis and in this article I will attempt to walk you through both possible pathways and what I actually think will happen.

What happened to private medical practice

Private medical practice in the United States fell from 60.1% of physicians in 2012 to 42.2% in 2024. Hospitals took most of the share. Private equity took a meaningful chunk in derm, GI, ortho, ophthalmology, and a handful of other specialties. UnitedHealth’s Optum quietly became one of the largest employers of physicians in the country.

The consolidation logic was simple and this was before AI. The back-office burden for a solo or small group practice became too heavy. EHR costs, payor contracting, prior authorization, MIPS reporting, cybersecurity, malpractice insurance. A 2-physician practice could not credibly fund and operate this stack. A 100-physician PE-backed platform could. Roll up the small practices, spread the back-office over a bigger denominator, exit at a higher multiple. That was the trade.

It worked for about a decade. It is now breaking down.

The platforms LBO’d at 3% debt costs cannot refinance at 8% to 10%. Envision filed Chapter 11 in May 2023. Cano Health filed February 2024. American Physician Partners ceased operations July 2023. GenesisCare and Steward followed. The exit market froze. Multiples compressed from 14x to 18x EBITDA at the 2021 peak to 10x to 14x today for the best assets, and 4x to 7x for sub-scale add-ons.

Meanwhile something else happened that nobody was modeling. AI scribes, cloud EHRs, and outsourced RCM collapsed the back-office moat. The cost to start a solo Direct Primary Care clinic in 2026 is $70,000 to $150,000 all-in. In 2010 it was $250,000 to $500,000. Hint Health’s data shows DPC membership grew 837% from 2017 to 2025. The American Academy of Family Physicians reports 9% of family physicians ran a DPC practice in 2023, up from 3% in 2022.

In short, PE rollup activity of private practices slows down meaningfully through 2030, but does not stop. The shape of the activity changes more than the volume.

The middle is dying. Solo independents are bootstrapping back. The platforms are stuck. That is the medical practice story right now.

Why I see the same pattern in research sites

Look at how research sites consolidated. PE platforms formed roughly between 2018 and 2022, capitalized at the same peak multiples with the same cheap debt. Headlands took capital from KKR. Velocity from GHO Capital and Jefferies. CenExel from Webster Equity. Flourish from Webster. Javara from Heritage Group. Evolution from Lindsay Goldberg. The list goes on.

The thesis was identical to medical practice. Small sites cannot afford the modern infrastructure. CTMS, eSource, eConsent, eTMF, sponsor portals, FDA inspection readiness, ICH-GCP compliance, 21 CFR Part 11, sponsor-specific reporting requirements. A 1-PI site with three coordinators cannot credibly fund and operate this stack. A 50-site network can. Roll up the small sites, spread the back-office, become a preferred provider on master service agreements with the big sponsors, exit at a higher multiple.

Same logic. Same buyers. Same debt structures. Same window.

Now look at where we are. The refinancing wall is coming for our networks the same way it came for the medical practice platforms. The 2018 to 2022 LBOs need to refinance into a different interest rate environment. Several of the networks have had to lay off staff, close underperforming sites, or restructure debt.

I am not predicting any specific network bankruptcy. I am saying the structural conditions that broke the medical practice rollup model are present in our model too.

The case that independent sites survive and grow

Here is the optimistic scenario. AI tooling and modern infrastructure collapse the operational cost of running a small site the same way they collapsed the cost of running a small medical practice. eSource automation, AI-assisted source documentation, automated regulatory binder management, AI query response, AI-driven recruitment screening, decentralized trial infrastructure. The stack that used to require a 10-person operations team starts to require 3 people and good software. As a shameless plug example, my team and I are building coordinare.co to give free tools such as CTMS, budget generators, document managers, and project management task boards to small sites for free.

Sponsors get burned by network consolidation. They notice that when 60% of their study volume sits at three or four networks, those networks have negotiating leverage that the sponsors do not love. They start deliberately seeding independent sites again to maintain optionality and to access diverse patient populations that the networks cannot reach efficiently. Anecdotally, I have been told personally by a half dozen CRAs as well as some sponsor representatives that they are glad there are independent sites like Yuma Clinical Trials still operating.

FDA Diversity Action Plans, decentralized and hybrid trial guidance, and continued sponsor focus on patient diversity push study volume toward community-embedded sites. Independent sites in underserved geographies, owned by PIs with deep community ties, become structurally advantaged for diversity-mandated trials.

Operators leaving networks in frustration start their own sites. PIs who sold to a network and stayed for the earnout watch the operational autonomy disappear and the patient relationships degrade, and when their earnout completes they leave and start fresh. New PI-founders enter the space because the cost to start is lower than it has ever been.

A new generation of site enablement infrastructure emerges. Besides DSCS Sweat Equity & Investments, LLC dba DSCS CRO‘s coordinare.co which offers free tools and a monthly low cost subscription fee for additional assistance and handholding, many independent sites plug into other low cost enterprise-grade CTMS, regulatory QC, budget negotiation, and sponsor contracting infrastructure without selling. They retain ownership, clinical control, and operational independence while accessing the scale benefits that used to require selling to a network.

In this scenario, independent sites are not extinct in 2035. They are the dominant model again, just in a different shape. Smaller, leaner, AI-enabled, often single-therapeutic-area focused, often with deep sponsor-specific relationships, often plugged into enablement platforms rather than owned by networks.

The case that independent sites mostly disappear

Here is the pessimistic scenario, and I have to be honest, it is also plausible.

The sponsor preference for scale is real and it is not going away. When a sponsor is launching a Phase 3 with 80 sites and a 36-month enrollment window, the easiest path is the master service agreement with three or four networks that can deliver those sites with a single contract, a single budget template, a single payment process, and predictable enrollment. The procurement and clinical operations teams at large sponsors are structurally biased toward this model. Independent sites have to win each study individually. The activation energy is higher.

Patient acquisition for clinical trials is harder for independent sites than patient acquisition for medical practice. A Direct Primary Care (DPC) physician markets to a defined geographic patient base. A research site has to find patients who match very narrow inclusion criteria for whatever indications it is currently studying. Networks have shared patient databases, centralized recruitment teams, and proven recruitment marketing playbooks that an independent site has to either replicate or do without.

The capital required to start an independent site is not actually that low when you include the working capital to survive 6 to 12 months of slow startup activations, the regulatory infrastructure investment, the quality system buildout, and the staff you need to keep on payroll before you have study revenue. The bootstrap DPC equivalent for sites is harder than people make it sound.

The supply of new independent site founders is constrained. A burned-out physician can start a DPC practice with their medical license. A burned-out site operator does not necessarily have the experience or the network to start a site from scratch. The talent pipeline for new independent PIs is thinner than the talent pipeline for new independent physicians, however good entrepreneurs can transform the raw ingredients of a busy private practice based physician and turn them into a decent investigator. The question remains if sponsors are more agreeable to giving these newer PIs a chance heading into 2030. The answer may sadly be no.

Sponsors continue to consolidate their own preferred-provider lists, not expand them. The CRO industry continues to consolidate, and the large CROs all prefer to work with networks rather than independent sites for operational reasons. The economic gravity favors scale.

The networks that survive the refinancing wall come out leaner and more disciplined. They learn from the medical practice platform failures, they shed unprofitable sites, they invest in technology that lowers their unit operating costs, and they end up structurally more competitive than they were at peak. The 2030 version of Headlands or Velocity may be a better operator than the 2022 version was.

In this scenario, independent sites are not extinct but they are a small niche. Most clinical research volume runs through 5 to 8 large networks and the hospital-affiliated academic centers. Independent sites serve specific therapeutic areas, specific geographic communities, and specific sponsor relationships, but they are not the dominant model. The general-purpose PI-owned site with 5 to 15 active studies across multiple indications is mostly gone.

What I actually think

Both outcomes are real possibilities. Anyone telling you with certainty which one wins is selling something.

What I believe is that the truth sits closer to the middle but leans toward the optimistic version with a critical caveat.

The middle dies. Legacy independent sites with no enablement infrastructure, no modern quality systems, and no sponsor-specific specialization will not survive the next 5 years. That is not a prediction. That is happening right now. The general-purpose site, owned by a PI who started it in 2008, running on paper source documents, using a legacy CTMS, with three coordinators who do everything manually, and occasionally-involved specialist providers is the buggy whip of clinical research. It dies regardless of which scenario plays out.

The top consolidates further. The networks that survive the refinancing wall absorb the weaker networks, the FDA-finding-prone sites, and the operationally undisciplined platforms. Expect a couple of high-profile network restructurings or sales in the next 24 to 36 months. Expect at least one network to consolidate into another at a discount.

The bottom regenerates, but only the tech-enabled, community grassroots-affiliated, sponsor-specialized version of it. The independent site that survives 2030 looks different from the independent site that started in 2010. It has 2 to 7 PIs, modern eSource, AI-augmented operations, deep specialization in 1 to 6 therapeutic areas, direct sponsor relationships in those areas, and it plugs into enablement infrastructure rather than building everything itself. It is faster, cheaper to operate per study, and higher quality than the legacy independent it replaced. The founders are also more involved and readily available to troubleshoot those pesky sponsor demands.

The critical caveat is this. The enablement infrastructure that the new independent site needs does not yet exist at the scale it needs to exist. Privia, Aledade, and Pearl Health serve physicians. The equivalent for sites is a patchwork right now. CRIO, Florence, RealTime-CTMS, Veeva, Inato, Gleam and a handful of others like my own Coordinare and PatientAce provide pieces. There is no integrated stack that gives an independent site the back-office, the regulatory QC, the budget negotiation, the sponsor contracting, and the AI tooling in one place. That is the gap.

Whoever builds that gap honestly, and I mean honestly in the sense of building it for the operator and not for the immediate profits, will determine whether the optimistic scenario or the pessimistic scenario wins. If the enablement infrastructure gets built, independent sites survive at scale. If it does not, the pessimistic scenario plays out and the networks win by default. This is why I started building tools this year (coordinare.co, sorry to continuously plug it but hey, it’s a noisy world out there).

I am not the only one building. Other operators see this too. There is a small but growing community of people who came up through site operations, lived the consolidation wave from the inside, and are now building the tools and platforms that let the next generation of independent sites survive and compete.

So the answer to the question that started this piece is yes, there is hope for a clinical research ecosystem with small independent research sites at all times. But the independent site of 2035 is not going to look like the independent site of 2015. It is going to be smaller, leaner, more specialized, more AI-augmented, and plugged into an enablement infrastructure that we are still in the process of building.

The middle dies. The top consolidates. The bottom regenerates in a new shape.

That is the truthful version. It is less inspirational than independent sites rise again and more interesting than consolidation wins forever. It is also what I see happening every week from where I sit in Yuma, Arizona, running my site, watching the data, and talking to other operators at SOS Save Our Sites who are trying to figure out what comes next.

The path forward is not nostalgic. The legacy independent site is not coming back. The path forward is to build the new shape of the independent site and the infrastructure that lets it survive. That work is happening. Whether it happens fast enough to matter is the open question.

I think it will. But I would not bet my career on it if I did not also believe that the operators making that bet have to actually do the building, not just talk about it. Get out your picks and shovels.

For two years my business partner Chris Sauber has been telling me the same thing about clinical trial budget

Subscribe now

negotiations. When the first version of a study budget arrives as a Word document instead of a spreadsheet, the entity sending it is testing you. The asymmetry is intentional. A Word doc lists visits and a total dollar amount per visit. It hides everything underneath. No schedule of events breakdown, no per procedure line items, no per assessment pricing. With only a per visit total, you cannot tell whether the EKG, the labs, the PK draw, the cognitive assessment, the diary review, the AE collection, the urine pregnancy, and the IP dispense were priced at fair value or rolled into a number that quietly excludes half of them.

I always believed him. But until yesterday I had never quantified it. I didn’t have a way to test it, it was just an interesting thought. AI has now made this a 5 minute assignment. Admittedly this is not a real study that would be comparable with one done by an industry think tank, but we may be at the age where we do not need those anymore. Anecdotes can be tested. So we did.

We pulled every Initial Sponsor Offer Chris has saved across his DSCS Sweat Equity & Investments, LLC dba DSCS CRO budget negotiation work. Two hundred and one unique studies pulled directly from his email and our google drive. Every subsequent round, every counter, every redline aside, just the first version each entity sent the site. Then we classified each by file format and by who sent it. Direct from a sponsor, or routed through a CRO acting on the sponsor’s behalf.

The headline number is uncomfortable yet all-too-familiar for anyone who works in a budget capacity at the site level.

Direct from sponsor: 4.8 percent of Initial Offers arrived as a Word doc or a PDF. The other 95 percent came as a spreadsheet broken down by schedule of events.

CRO routed: 53.3 percent of Initial Offers arrived as a Word doc or a PDF. Roughly the same number as the Excel files.

That is an eleven-fold obfuscation rate. Not double. Not three times. Eleven times higher when a CRO is sitting between you and the sponsor. Even if a larger n-number study were ever to be done (doubtful) and this number were cut in half, it is still a meaningful signal.

Why does the asymmetry exist? Sponsor incentives and CRO incentives are not aligned the way most sites assume. The sponsor wants the trial completed on time, with clean data, at fair cost. A clearly priced budget helps that goal, because the sponsor pays sites to actually run the work, not to negotiate it for three months. The CRO has a softer incentive on the same axis. The longer a study runs, the more change orders the CRO bills. The more PM hours stack. A budget that lands at fair market values on round one is not the CRO’s most profitable outcome. A budget that takes three rounds, opens five clarification threads, and arrives at a number ten percent below site fair value is.

So what does a site do?

You ask. Every time. Without exception. The line is, “could you send the budget as a spreadsheet broken down by the schedule of events.” Five years ago the answer was almost always no. Today, in Chris’s actual data, the answer is yes more often than not. The Word doc itself is the first test. Asking for the spreadsheet is the first move. Sites that ask get treated as savvy operators and negotiated against at closer to fair value. Sites that do not get squeezed.

That is the entire trick. One sentence in one email. Eleven times more leverage on day one.

(Data: 201 Initial Sponsor Offer files, Chris Sauber DSCS CRO drive, 2024 through May 2026. Sponsor and CRO names omitted under standing CDA obligations. Pattern only, not deal-level pricing.)

Subscribe now

Sponsors and CROs are racing into agentic AI. Half of sites still run paper source. PE-backed networks are vibe-coding their own tech while mom-and-pops are getting left behind. GPT wrappers are grifting and selling themselves as infrastructure. And underneath all of it, sites remain the last physical-world play in a digital-only industry. Here is the actual map.

Last month Sarah Wang and team at TECH WEEK by a16z published an analysis on what happens to enterprise software when agents replace user interfaces (UI). The short version: when an agent can read and write directly to the database, the UI stops being the moat. What is left as defensible is the proprietary data the software uniquely generates (for the customer), the action layer it owns, the network effects it sits inside, and any real-world execution it coordinates. Pure Software as a Service (SaaS) gets commoditized. Anything tied to physical execution or a walled garden of proprietary data gets more valuable.

Sites generate 100% of study data yet most don’t know the first thing about owning their own data exhaust. Every screen failure, every randomization, every query closed, every protocol deviation, every dose held, every adverse event reported, every patient retained or lost. All the demographic data. All of it originates at the site. The sponsor does not generate that data, although they pay for it. The CRO does not generate that data they just move it around. The vendor does not generate that data, they just give sites the “privilege” of storing and organizing it. The site generates it, and then hands it off, and then forgets it exists. That is the single most expensive mistake in the industry right now. Because in a verify-then-trust world, the site that owns its own data exhaust is the site that wins the next study. The site that does not is the site that gets discounted on every feasibility, every budget negotiation, every selection decision, forever.

Owning your data means three things. Tracking it (medians, denominators, time windows, study-by-study breakdowns, indication-by-indication breakdowns, not just gut feel). Storing it (in a system of record that survives staff turnover, not in the head of the coordinator who left last quarter). Translating it (into sponsor-readable risk language on every surface you show them, the way Part 2 of this series will lay out in detail). Most sites do none of the three. They generate the data, they live it every day, and they let it evaporate. Sponsors do not let it evaporate. Sponsors capture it on their side, through CROs, EDC’s, eCOAs, and yes even through the Shared Investigator Platform, much to the chagrin of most study coordinators. The sponsor and CRO knows your site’s real numbers better than you do. Even the sponsor appointed vendors know your site better than you do. They have been keeping receipts for years. This is the central operational failure of the site layer in 2026, and it is also the largest opportunity. The site that flips this script (that captures, stores, and translates its own data on its own terms) becomes the site that can negotiate, that can pitch, that can refuse to overcommit, that can demand a fair budget, that can pick which studies to take. The site that does not reverse course becomes extinct very soon.

The PE-backed networks figured this out. That is half of why they exist. They are not just operational consolidators, they are data consolidators. Every site in the network feeds operational metrics into a shared pool, which gets translated into sponsor pitches, which wins more studies, which generates more data, which sharpens the next pitch. Which strengthens their moat and reduces stakeholder buffer between them and the source of all the revenues the sponsors. The mom-and-pop site does not have that flywheel, fortunately for them, it is becoming democratized very quickly.

The agentic shift, and where it breaks down in clinical research

In the SaaS era, the system of record was defensible because humans lived in the interface. Sales reps logged into Salesforce every day. CRCs logged into Medidata Rave, SIP, or some IRT every day (amongst a dozen other portals if I may add). The UI created habit, muscle memory, and the shared vocabulary an organization operated from. That was the moat. Agents kill that moat in enterprise SaaS. They are starting to kill it at the sponsor and CRO layer of clinical research, too. Lilly just announced a $1 billion co-innovation lab with NVIDIA spanning drug discovery, clinical development, and manufacturing, building on their existing OpenAI antimicrobials partnership and a $2.75 billion AI drug discovery deal with Insilico. Medable and Medidata are both building agentic AI that do source data verification and I wouldn’t be surprised if source data review into their platforms. The top of the industry is moving fast and the large frontier AI models such as Anthropic and OpenAI are getting paid off of all of that.

Here is where it breaks down. Roughly half of US clinical research sites still use paper source. They were never on a UI to begin with. The agentic transition is hitting an industry that never finished the previous transition, and the gap between the leading edge and the trailing edge is widening every quarter. The a16z framework needs a translation layer to apply to clinical research. The bottleneck is not migrating from “humans logging into a UI” to “agents calling an API.” It is leapfrogging from “no system at all” to “agents calling an API” in some parts of the industry, and from “paper logs” to “anything” in other parts. The transition has more dimensions than the enterprise SaaS version.

The bottleneck migrates, and shifts somewhere else

There is an essay by Zohar Atkins applying the Jevons paradox to learning. The Jevons paradox is the nineteenth-century observation that when an input gets cheaper, demand for it goes up, not down. Watt’s steam engine made coal more efficient. Britain ended up burning ten times more coal, not less, because cheap mechanical work created entire industries that did not exist before. If you are paying attention as a site owner you are probably noticing that tech “solutions” are beginning to drop their costs. Atkins writes: “When consultation gets cheap, the bottleneck migrates somewhere else.” That is the single sharpest sentence I have read on what is happening to enterprise software in 2026. The bottleneck does not go away when tools get cheap. It moves.

The question for every site, every sponsor, every vendor is where it moves to. This shows up at the site level in ways that surprise people. We had a conversation on a recent PatientACE livestream about this. The ads start working. Lead automation kicks in. Pre-screening scheduling automates. The funnel is humming. And suddenly the bottleneck is not patient acquisition anymore, it is dry ice, it is lab personnel, it is experienced CRCs. There is not enough dry ice/personnel in Yuma to handle all the work the streamlined funnel is generating. You shifted the constraint, you did not eliminate it. The bottleneck has to live somewhere. The question is whether you are the one deciding where.

For sponsors and CROs, the bottleneck is moving from “do we have the AI capability” (they do, increasingly) to “can the sites we need actually feed our agents real data.” For sites, the bottleneck is moving from “can we afford the software” (suddenly, increasingly, yes) to “do we have the operational discipline to run anything well, regardless of what tools we use.” The vendor landscape is reshaping around these moving bottlenecks. There are now three tiers of software vendors selling into sites, plus a fourth category that pretends to be Tier 2 but is not.

BYOT vs. mandated tools, the split that runs through every tier

Before the tiers, one critical distinction. Every piece of software at a site falls into one of two buckets. Bring-your-own-tech (BYOT) tools are what you choose to buy and run at your site. CTMS. Patient recruitment platforms. CRMs. E source. E reg. Coordinator dashboards. Source document management. Invoice reconciliation services. Internal analytics. You decide what to use, you pay for it, you train your staff, you live with it. Mandated tools are what the sponsor or CRO forces on you, study by study. eCOA platforms. eDiaries. EDC systems (Medidata Rave, Veeva CDMS, and the all-too-infamous SIP).

Sometimes the tools may be BYOT and mandated at the same time. eConsent systems. eSource. eReg systems sometimes. You did not choose these. You have to learn a new one for every sponsor. Your staff lives between five and fifteen mandated platforms on any given week. This split matters because the agentic transition hits each bucket differently, though not in the way most sites assume. The common mistake is thinking mandated tools are off-limits to your own agentic workflows. They are not. You can point Claude Cowork or any agentic tool at any mandated platform you have access to (provided you are not opting in to training that agent’s frontier model on sponsor proprietary data). You can have an agent read your EDC, summarize open queries, draft response language, navigate the Shared Investigator Platform, prep your reg binder. The sponsor does not control what you do with your access. They control the architecture underneath. That is the real distinction.

On BYOT tools, you own both the architecture and the interaction layer. You choose the system of record, you choose the integrations, you choose the workflow, and you can stack agents on top however you want. On mandated tools, the sponsor owns the architecture (when the EDC ships native agentic features is their call, not yours) but you still own the interaction layer (what you do with your own agents pointed at their interface is yours to decide). The leverage on mandated tools is asymmetric. You cannot rewire how Medidata Rave validates data entry. But you can have Gleam fetch and fill, or Cowork pre-fill the form, flag inconsistencies before submission, batch-process queries, and surface deviation patterns the monitor will eventually ask about anyway.

Here are the Tiers of the tools as I see them:

Tier 1: Commodity tools

Generic CRMs, basic task boards, feasibility trackers, recruitment vendors. I used to pay for Monday.com until I just built a task board myself. Now anyone can have it free on Coordinare.co, along with the feasibility copilot, budget generator, GCP cert, cultural competency cert, bookmark manager, and a chatbot trained on my entire content library. All free.

These tools were sticky in the enterprise SaaS world because someone logged into them every day. The UI was the product. Agents do not care about UIs. Any site with a frontier AI can rebuild a basic working version in an afternoon. I said this on stage at SOS Save Our Sites during the AI panel. Why would you buy a CRM in 2026 when you can build your own? The clinical research wrinkle: most sites never had this layer to begin with. The 50% still using paper source skipped the SaaS era entirely. They are not migrating from CRM A to CRM B, they are deciding whether to use any CRM at all. Tier 1 has no moat anymore. We are giving these tools away free because charging for them in 2026 would be charging for something a site could build themselves. If a vendor is trying to charge you a real subscription for what is essentially a fancy template wrapper, ask hard questions. Or better yet, run.

Tier 2: Connective infrastructure (and why agents need it)

This is where it gets interesting. CRIO . Inato . PatientACE . Gleam . Veeva Systems . The EHR layer, eClinicalWorks and Epic. The Shared Investigator Platform. eConsent providers. These vendors are not selling you a UI or a GPT wrapped product. They are selling you something that survives the agentic transition and gets more valuable in it, for three specific reasons.

• They have proprietary data exhaust. CRIO has your actual historical screen-fail rates, study by study, indication by indication. That is not data you can rebuild. It only exists because you ran trials on their platform. Inato has the network of thousands of sites and their de-identified cohort data, which no individual site could ever replicate. Gleam’s fetch-and-fill is not just storing data, it is generating new data exhaust about your operational performance every time it runs. PatientAce has the entire patient journey from ad click to study completion.

• They own the action layer, they sit inside network effects, and they play along nicely together. Here is the concrete proof point. We almost lost a study recently because our EMRs would not connect to Inato. Inato’s verification stack requires API-level integration, not Excel exports, because they de-identify data after ingestion. Our four PIs each run a different EMR, and getting permission and access for each one is slow. We were minutes from being filtered out. Patient Ace saved us. They had the API. They had the network compatibility. They could feed Inato what Inato needed. Without that Tier 2 integration, we would have been a paper-source site as far as the sponsor was concerned, and we would have lost the slot. That is what Tier 2 is for. It is not a CRM you bought because you needed task management.

• It is the infrastructure that lets your site exist in a verify-then-trust world. The reason agents make Tier 2 more valuable, not less, is that agents need somewhere reliable to read from and write to. The system of record becomes the connection point for every agent that talks to your site. If you are picking software to commit to, this is where you should be willing to spend, integrate deeply, and build process around. Full disclosure on a few of these. I am an advisor and investor in Gleam. CRIO and Inato, sponsor my podcast and I am a co-owner of PatientAce. I am telling you about all of them because the framework holds regardless. Use the ones that fit your site. The point is the category is defensible.

The Tier 2 imposters: GPT wrappers in disguise

There is a fourth category of vendor showing up in clinical research right now that needs its own filter. GPT wrappers pretending to be Tier 2 infrastructure. These vendors claim proprietary data exhaust they do not have, claim action layer ownership they do not have, claim network effects that do not exist, and price themselves like Tier 2 infrastructure. Underneath, they are calling OpenAI or Anthropic’s API with a system prompt and putting a UI on top. They have no defensible moat. The frontier model is doing all the work, and the wrapper is collecting the margin. Your site can do this yourself if you really wanted to. How to spot them. Ask three questions.

• What data do you generate that I could not get anywhere else? If the answer is hand-wavy, they are a wrapper.

• What action does your software take in the world, not just summarize or recommend? If they only describe, not do, they are a wrapper.

• What is your underlying model and what happens when the next frontier model releases?

If they do not have a clear answer, they are a wrapper and they know it. I am not saying GPT wrappers are useless. Some of them are well-designed products. But they are not Tier 2. They are Tier 1 with better marketing. Price them accordingly. They should cost what a workflow wrapper costs, not what infrastructure costs. The GPT wrapper problem is going to get worse before it gets better, because every vendor pitching sites in 2026 is going to claim AI capability. Most of them are wrappers. The framework above is your filter. There is also a Steve Jobs / Mark Twain test worth running. If a vendor is pitching you 14 features they can do and 22 integrations they support, ask which one you actually need. Mark Twain said if I had more time, I would write fewer words. Steve Jobs built his entire design philosophy on what to remove. A lot of these GPT-wrapper vendors are throwing every feature they can prompt-engineer at the wall and using you as their market research. The worst Tier 2 imposters give you a feature list a mile long because they have no idea what they are actually selling.

Tier 3: The execution layer, and the restaurant analogy

This is the part most software people do not understand, and it is the part that matters most. The site itself is not software. The CRC who actually closes the query is not software. The PI who actually consents the patient is not software. The phlebotomist who actually draws the blood is not software. This is the real-world execution layer. Humans, in physical space, doing operationally complex work that involves judgment, relationships, and presence.

Sites are like restaurants. The sponsors are the customers. The CROs and platforms like Inato are the delivery and aggregator layer. The sites are where the food actually gets made. The analogy starts to wobble if you push it (are the patients the ingredients? the chefs? this is where it gets weird), so I will not push it. The part that holds is the spectrum. Like the restaurant industry, the site landscape ranges from franchise networks with PE backing, professional kitchens, and standardized operations, all the way down to mom-and-pop shops running on instinct and the owner’s relationships.

This analogy explains the bifurcation happening right now. The PE-backed site networks (AES, Velocity, Headlands, Flourish, M3) are operating like restaurant chains. They have capital. They have engineering teams. They are hiring developers and vibe-coding internal tools that they keep proprietary. They are not buying Tier 1 commodity software, they are building it in-house with frontier AI assistance. They are absorbing the agentic transition by absorbing the engineering capability that produces it. The mom-and-pop sites are operating like independent restaurants. Skilled operators, sometimes great work, often surviving on the PI’s reputation and the staff’s hustle. No engineering depth. No capital for software. Still using paper source, sometimes. Getting squeezed between the PE networks above them and the sponsor expectations below them.

The middle layer (single-location sites that are professionally run but not networked) is the most interesting to watch. They are the equivalent of the well-run independent restaurant or small local chain. They cannot vibe-code their way to a Tier 2 integration like a PE network can. They cannot survive on the PI’s reputation alone like a mom-and-pop can. They have to be operationally excellent with the tools available to them, which is exactly where the consulting layer becomes essential.

Middle to middle, not end to end

There is a useful frame for thinking about how agents fit into the site workflow. Agents do middle-to-middle work. Humans own the ends. On the front end, the coordinator decides which leads to pursue, which days to allow pre-screening visits, which patients to prioritize. On the back end, the coordinator validates which prescreened patients become actual screening visits, which screen failures to follow up with, which protocol deviations need escalation. The agent does the part in the middle: the automated text the second a lead comes in, the scheduling link to the site’s calendar, the reminder the day before the visit, the lead categorization across stages, the synthesis of EMR cohort searches. The site can dial those ends in or out. If a study is hard to recruit for, the site owner gets on the phone personally because randomizations matter more than streamlining. If a study has plenty of demand, you push the middle further toward both ends and let the agents handle more. The framework is flexible. The principle is not. The ends are human. The middle is agentic. The bread is baked between the human ends.

Where the consulting layer fits, and why it gets more valuable

There is a fourth dimension the a16z analysis hints at but does not name explicitly. The technical capability of the buyer. Most sites do not have the engineering depth or time to integrate Tier 2 infrastructure properly, build their own Tier 1 tools, train their staff on agentic workflows, and translate their operational metrics into the language sponsors actually score on. The PE-backed networks have that capability in-house. The mom-and-pops do not. The middle is where the consulting layer earns its keep.

Atkins makes the deeper version of this point. He writes: “The supply of accessible material generates a demand for synthesis that did not exist when the material was inaccessible.” That sentence describes the consulting layer of clinical research in one line. When the data was inaccessible (the old world), nobody needed a translator. Sites guessed, sponsors discounted, monitors verified later. There was nothing to synthesize because there was nothing to access. Now the data is accessible. Every site can pull CRIO historicals, Inato demographics, Gleam-generated eSource performance traces, Patient Ace recruitment pipelines. The supply is suddenly enormous. And the demand for synthesis (turning that supply into sponsor-readable risk language or more opportunities for the site) is exploding to match.

My own consulting company, DSCS Sweat Equity & Investments, LLC dba DSCS CRO and other advisory firms, budget specialists, regulatory consultants. The whole services layer of clinical research. We sit at the translation point between the infrastructure (Tier 2) and the real-world execution (Tier 3). We help sites pick the right software, integrate it properly, run it well, and turn the resulting operational data into sponsor-readable risk language. That layer is getting more valuable, not less, as the tools commoditize. Because the tools amplify judgment. They do not replace it. A site with great tools and no judgment will still lose awards to a site with mediocre tools and great judgment. A site with great tools and great judgment will dominate.

The bread and the baking

There is a parable Atkins quotes from the kabbalistic tradition that captures the whole industry shift better than anything I could write myself. A man from the mountains comes down to the city, eats bread for the first time, and asks what it is. He is told it is made from wheat. He returns to his mountain, gathers raw wheat, eats it dry by the handful, and goes home convinced he has tasted what the city tastes. He has eaten the input. He has missed the thing made when the input is ground and kneaded and salted and left to rise and exposed to fire. Bread lives only in the baking. The tools are the wheat. Cheap, abundant, suddenly available to everyone. The whole industry is reaching for the wheat and eating it dry by the handful, then wondering why clinical trials are chronically under-enrolled and over-budget. The wheat is not the bread. A GPT wrapper selling itself as Tier 2 infrastructure has sold you raw wheat with frosting on top and called it cake.

The bread lives in the baking. The judgment. The synthesis. The PI or sub-I who knows which study to put which patient in, or the Site Director that knows exactly which CRC to put on which trial or what tool to call up on at the right time. The deviation log that gets filled before any CRA lists it as an action item. The tools are commoditizing. The baking is not. Sponsors are paying for bread.

The physical molecule play

Here is the final framing that ties everything together. Sites are the last physical-molecule play in an otherwise digital-only industry. At the end of the day, the entire clinical research stack organizes information. Sponsors will soon run AI drug discovery on NVIDIA supercomputers. CROs will soon run agentic CRAs on Medidata and Medable. Vendors run network-effect data infrastructure across thousands of trials. All of it is information moving through software. Sites are different. Sites organize information with the added caveat of taking a real patient on a physical journey. A real human who has never heard of a clinical trial or thinks it’s “sketch”, shows up at a real site. Real blood gets drawn. A real coordinator schedules the next visit and follows up when the patient does not show. The trial does not exist until that journey happens.

This is why every other layer of the industry exists in service of what happens at the site, and why no agent is going to displace it. You cannot draw blood through an API. You cannot consent a patient with a prompt. You cannot run a phase three trial without a human investigator looking at a human patient and making a clinical judgment in physical space. The whole industry is going digital except for the part that cannot. Sponsors are racing to AI. CROs are racing to AI. Vendors are racing to AI. The PE-backed site networks are racing to AI. The mom-and-pop sites are getting left behind. The middle layer is hiring consultants to keep up.

All of it is happening at the same time, in the same industry, in the same year. And underneath all of it, the trial still requires a patient to walk through a door. What AI can rebuild is getting cheaper. What AI runs on is getting more valuable. What AI cannot replace is the physical molecule play and the synthesis of creating. That part is not going anywhere. That part is the whole reason the rest of the industry exists.

Next week, Part 2: what this map means in practice. Specifically, how to rewrite your website, your feasibility responses, your cold outreach emails, and your conference handouts for the verify-then-trust era. The framework above tells you why the bar moved. Part 2 tells you exactly what sponsors are looking at on every surface you put in front of them.

References: Sarah Wang and team at a16z on headless software in the agentic era, and Zohar Atkins’s essay “When Knowledge Is Cheap, Insight Is Everything”, which applies the Jevons paradox to learning in a way that maps cleanly onto what is happening to clinical research right now.

Subscribe now

HIMS reported first quarter 2026 earnings on 5/11/26. Revenue came in light. EPS missed by a margin so wide that some headlines are calling it a 1,100 percent miss. The stock dropped 16.5 percent in after-hours trading, settling around 24 dollars.

If you only read the headlines, you would think this is a broken company.

I sat through the entire earnings call. I’m not a Wall Street analyst. I run a clinical research site in Arizona. I’ve spent 21 years inside the clinical trials industry, negotiating budgets with biotech sponsors, watching drugs move from Phase 1 to commercial launch, and tracking how the FDA actually behaves when new categories emerge.

The market is reading this print on a quarterly timeline. The actual story unfolds over the next 36 months. But there is also a credible structural argument that this company’s best margins are permanently behind it. Both views deserve honest engagement.

Here’s what I think most analysts missed, and what the bulls need to be honest with themselves about.

First, what HIMS actually is

For readers who only know HIMS from the Super Bowl ads, here is the plain-English version.

HIMS is a telehealth platform. You fill out an online questionnaire, a licensed provider reviews your case, and prescriptions get shipped to your door. They serve men and women across hair loss, sexual health, mental health, skincare, dermatology, hormone optimization, and most famously, weight loss with GLP-1 drugs like Wegovy and Ozempic.

For most of 2024 and into early 2025, a large portion of their weight loss revenue came from compounded GLP-1. Compounding is when a pharmacy legally produces a custom version of a drug that is in shortage. HIMS sold compounded semaglutide, the active ingredient in Wegovy, at high margins because branded supply could not meet demand.

In March 2026, that ended. The FDA declared the shortage over. Novo Nordisk, the maker of Wegovy, sued. HIMS settled, killed the compounded business, and pivoted to distributing the actual branded drug through a direct relationship with Novo, with Lilly’s Zepbound added by referral in April.

That pivot is the entire story of tonight’s earnings print.

The numbers, in context

Revenue came in at 608 million dollars, up 4 percent year over year. Subscribers grew 9 percent to 2.6 million. Adjusted EBITDA was 44 million dollars.

The headline-ugly number is a 33 million dollar restructuring charge. That charge represents the writedown of compounded GLP-1 inventory and infrastructure that HIMS can no longer legally use. That is the EPS miss in one line. It is not the business breaking. It is strategic cleanup.

Now here is the part most coverage is glossing over.

HIMS raised full-year 2026 revenue guidance to a range of 2.8 to 3 billion dollars. Q2 is guided to 25 to 28 percent growth year over year. International revenue grew nearly tenfold to 78 million dollars. And within six weeks of switching on branded Wegovy distribution, the company fulfilled over 125,000 prescriptions.

CFO Yemi Okupe said it directly on the call. HIMS is adding more than 100,000 new weight loss subscribers per month. That is a record. Higher than the New Year’s resolution spike. Higher than the Super Bowl ad spike. Branded distribution is bringing in more customers than compounded ever did.

For anyone who runs an operations-heavy business, this move is recognizable. You give up unit margin to win territory. Once you own the territory, you bring efficiency back in. It is the same playbook Amazon ran on retail and Uber ran on transportation. Right now it is the HIMS playbook on consumer health.

The bear case bulls need to take seriously

Before going further into the bull case, the structural bear argument deserves honest engagement, because parts of it are correct.

The Novo settlement was struck March 9, 2026. Branded Wegovy distribution didn’t go live until March 26. That means Q1 still captured roughly two-and-a-half months of the high-margin compounded business. Yet the print still missed and margins still compressed.

The implication matters. If Q1 looks like this with most of the compounded business still flowing, Q2 may look significantly worse when the full quarter reflects branded-only economics. Yemi essentially acknowledged this on the call when he warned of near-term margin headwinds as one-month branded shipments fully replace two-month compounded cadences.

The deeper structural argument is harder to dismiss. HIMS’s exceptional margins in 2024 and 2025 were partly the result of aggressive positioning in the regulatory gray zone created by the GLP-1 shortage. That window has closed. What replaces it is a branded distribution business operating at referral economics, where LifeMD’s own CFO has publicly admitted, quote, we don’t make as much. Amazon One Medical launched a competing GLP-1 program in April at 25 dollars a month with insurance. The branded distribution rail is real, but it is not a high-margin business in isolation.

If you are bullish on HIMS, you have to be willing to make a platform argument, not a distribution argument. The distribution business alone does not justify a premium multiple. The question is whether the data infrastructure, the international expansion, the non-GLP-1 specialties, and the peptide opportunity combine into something the market is currently not pricing.

Here is why I think they do.

What Wall Street missed, part one: the peptide opportunity

For readers unfamiliar with the term, peptides are short chains of amino acids that act as signaling molecules in the body. Several are being studied for longevity, recovery, hormone optimization, and metabolic health. They sit in a category adjacent to GLP-1 but represent a much broader frontier of biology.

The FDA is currently reviewing how to regulate consumer-facing peptide therapies. A meeting at the end of July is expected to provide clarity on which peptides may move into a regulated, physician-supervised distribution pathway.

More than half of the retail investor questions submitted to HIMS ahead of tonight’s call were about peptides. And CEO Andrew Dudum said something during the call that, if you understand how regulated medicine actually works, should make you stop and pay attention.

He said the company will not be first to market with peptides. He said HIMS will only enter the category once they can deliver it with physician oversight, validated supply chains, and clinical protocols. He said, quote, we won’t launch access to peptides until we meet these very high standards that we believe everyone should be meeting.

There is a fair critique that this could become a repeat of the compounded GLP-1 playbook, where HIMS exploits a regulatory edge until that edge closes. I think the opposite is more likely, and the reason matters. The compounded GLP-1 business triggered a DOJ referral, a Novo lawsuit, and forced regulatory cleanup. Dudum is not stupid. The peptide framing he is using now, regulated, physician-supervised, supply-chain-validated, is the language of a CEO who learned an expensive lesson about operating in the regulatory gray zone and is positioning HIMS as the legal compliant option specifically because the gray market is going to get cleaned up.

If that read is right, here is the asymmetry. HIMS already has 2.6 million American subscribers, more than a million square feet of US-based compounding infrastructure, a domestic peptide manufacturing facility they acquired in 2025, lab testing infrastructure through their YourBio and Trybe Labs acquisitions, and clinical workflows built around the FDA’s actual regulatory framework.

When the FDA opens the peptide door, HIMS is the only company in the United States with every piece of infrastructure required to legally and safely deliver peptides at consumer scale on day one. Not the gray market suppliers. Not the boutique longevity clinics charging 2,000 dollars a month in Beverly Hills. Not the international platforms running trials offshore.

The market is pricing HIMS like a GLP-1 distributor. In six months, the market may be pricing it like a longevity platform.

That said, this thesis depends on the FDA actually opening the pathway. If the July meeting is restrictive or pushed into 2027, this catalyst weakens significantly. Honest bulls need to hold that risk in view.

What Wall Street missed, part two: the data architecture

The other underweighted section of tonight’s call came from Mo Elshenawy, the HIMS chief technology officer. He spent roughly ten minutes discussing the company’s AI infrastructure. Most viewers will tune out at the phrase “AI copilot.” For readers in clinical research and life sciences, this is the part to lean into.

Mo described an AI copilot that drafts responses for HIMS care coaches, which licensed providers then review and approve before they reach the patient. He described a labs AI agent that interprets biomarker results in the full context of an individual patient’s history. He described a weight loss AI companion launching soon that will proactively engage patients between provider touch points.

The substance underneath those features is what matters. Every patient interaction on the HIMS platform is being labeled by a licensed clinician. The intake. The diagnosis. The treatment selection. The dose adjustment. The reported side effects. The downstream outcome. All of it lives in a single integrated stack. All of it is reviewed and approved by a real provider.

That is the highest quality training signal available in healthcare AI. You cannot buy it. You cannot scrape it. You can only build it by running an end-to-end vertically integrated platform for a decade, which is what HIMS has been doing since 2017.

For readers in clinical research, the implication is significant. Real-world evidence is one of the most valuable data assets in modern drug development. Biotech sponsors pay clinical research sites enormous amounts of money to generate it on a study-by-study basis. HIMS is generating it continuously, across 2.6 million subscribers, across multiple specialties, across multiple countries, all owned in their proprietary stack.

That is not a telehealth company. That is a real-world evidence engine that happens to sell prescriptions to fund its own data infrastructure.

The bear pushback here is fair. Data is only valuable if it gets monetized. Right now, HIMS is not selling RWE contracts to biotech sponsors. The platform argument requires you to believe management will eventually convert this data asset into a real revenue line, either directly or through AI-enabled care products that justify higher pricing (wearables?). That’s a thesis on execution, and execution is not guaranteed.

What Wall Street missed, part three: international expansion

International revenue grew nearly tenfold in Q1, hitting 78 million dollars. The acquisition of Eucalyptus, expected to close in the second half of 2026, will add an estimated 450 million dollars in Australian annual recurring revenue and extends HIMS’s footprint across the UK, Germany, Japan, and Canada.

The Eucalyptus close is not in the current revenue guidance. That means the 2.8 to 3 billion dollar 2026 guide already reflects strong domestic growth, with international upside layered on top once the deal closes.

This is the most underappreciated catalyst on the calendar. International scale matters not just for revenue, but for data. The disease profiles, treatment responses, and prescribing patterns from a global subscriber base feed back into the same AI infrastructure described above. A platform trained on multi-geography data is structurally more valuable than one trained on a single market.

It is also the catalyst that does not depend on the FDA, the peptide pathway, or the US regulatory cycle. That makes it the cleanest part of the thesis.

Three things to watch over the next 18 months

For readers who want to follow this story rather than trade it, here are the three catalysts that will determine whether tonight’s drop becomes a buying opportunity or the start of a longer reset.

One. Q2 gross margins and EBITDA dollars. This is the swing factor. Q2 will be the first full quarter of branded-only GLP-1 economics. If gross margins stabilize above 65 percent and EBITDA dollars step up sequentially as Yemi guided, the strategic pivot is validated. If margins compress meaningfully below that and EBITDA flatlines, the structural distributor thesis gains weight and the platform thesis needs to be revisited honestly.

Two. The July 2026 FDA meeting on peptide classification. If the agency provides a clear regulated pathway for any major peptide categories, HIMS is positioned to be the first compliant US distributor at scale. If the ruling is restrictive or delayed indefinitely, the longevity platform thesis weakens.

Three. The Eucalyptus acquisition close. Expected in the second half of 2026 with a 240 million dollar payment at close. A clean close on time with the projected ARR intact validates the international expansion thesis. A delay or downward revision changes the math.

The takeaway, depending on who you are

If you are a general reader trying to understand what happened tonight, HIMS took a short-term financial hit to execute a long-term strategic move. The compounding business that drove last year’s margins is permanently gone. The branded distribution business that replaces it is bigger in reach but lower in margin. The real question is whether the platform being built underneath both businesses is worth more than either business alone.

If you work in clinical research, HIMS is worth watching as a structural competitor for real-world evidence contracts within the next five years, or even as a study sponsor in the wearables sector. The infrastructure they are building will eventually intersect with traditional sites, CROs, and sponsor data partnerships. That should change how all of us think about positioning.

If you are an investor, the market just handed you a chance to disagree with the headline reaction. But you also need to be honest about which side of the trade you’re on. The bear case is structural and defensible. The bull case requires you to weight the platform, data, international, and peptide arguments highly enough to override the structural margin reset. Reasonable people can land on either side. The question is whether you are evaluating the next quarter or the next decade.

Most retail headlines tonight will focus on the miss. Most CNBC commentary will focus on margin compression. Most short-term traders will react to the after-hours drop.

The operator’s read is different. HIMS just told the market it is willing to sacrifice short-term margin to win long-term territory. It quietly described the only US-based regulated infrastructure ready for the peptide era. It walked through a data architecture that, in any other industry, would be valued as a moat in its own right. And it raised full-year revenue guidance while doing it.

The structural bear case is the cleanest argument that has been made against this company in 18 months. It deserves to be answered, not dismissed. The Q2 print, the July FDA decision, and the Eucalyptus close are the three places where the bull thesis either earns its premium or has to fold.

This is not a buy or sell call. This is a frame.

The companies that own the distribution rails for emerging healthcare categories over the next decade will be among the most important businesses in the consumer health economy. HIMS just made the case, in a quarter that the market read as a stumble, that it intends to be one of them. Whether it earns that position depends on the next two quarters more than on tonight’s print.

Whether the market figures that out in three months or three years is the only question that actually matters.

This is The Clinical Trials Guru. I write from the operator’s seat, not from Wall Street, not from a Bloomberg terminal, but from the actual room where the drugs you read about get tested. If this analysis was useful, subscribe and share. New essays every week on the intersection of clinical research, biotech investing, and the future of consumer health.

Nothing in this newsletter constitutes investment advice. Do your own research.

The AI drug discovery argument is basically settled.

Isomorphic has compounds in human trials. AlphaFold 3 models protein, DNA, and RNA together.

Dave Ricks

said the innovation in the lab is no longer a bottleneck and that aspect will only improve further. The chemistry side is real. You can argue about how much AI improves preclinical success rates, but you can’t argue the molecules aren’t real anymore.

So the smart money has already moved. The next fight isn’t about whether AI can design a drug. It’s about who owns the layer between the molecule and the patient. And the innovations on the preclinical side mean more shots on goal coming up at research clinics all over the world.

Jensen Huang

said something recently that captures this perfectly. AI isn’t end to end. It’s middle to middle. The ends of any real-world workflow still need humans. What AI does is compress everything in between.

That’s the entire clinical research industry in one sentence. Humans still need to strategize and prompt and refine. Then they need to validate and conduct quality assurance on the outputs. The way I see it, if we get more opportunities in trials by volume, and AI is automating the middle grunt-work tasks, we may actually need more humans on both ends of the bottleneck but the skillset will be different than what it has been to get us here.

The front end still needs humans. A scientist still has to decide what disease to chase, what mechanism to target, what trial to design. Then the team slightly downstream needs to discover which Investigative sites are appropriate for the trials. The back end still needs humans. A patient in a community still has to be recruited, consented, dosed, retained and followed by a coordinator and a PI who actually know them. Also, and this is quite important, these study participants cannot be generated by AI. Someone they trust has to educate them on the process.

But the middle? Protocol to first patient enrolled. Patient identification to first contact. Data capture to query resolution. Site selection to activation. That entire middle is getting compressed, and that’s where the next 5 years of winners and losers get decided.

Here’s the analogy that makes this click. Think about how Amazon treats the last mile.

Amazon automated almost everything in the middle, and most of that was before AI. Warehouses are robotic. Inventory is algorithmic. Routing is now AI. The middle of their operation is one of the most optimized systems in human history. But the last mile, the part where a human actually walks a package to a door in a specific neighborhood, is still the hardest, most expensive, and most defensible part of the whole business. It’s why Amazon spent billions building out their own delivery network instead of just leaving it to UPS and FedEx. The middle is leverage. The last mile is the moat.

Clinical research is the same shape. The middle gets compressed. The last mile, which in our world is the community site enrolling and following the actual patient, is where the moat is. Protocol compliance at (greater) scale. No amount of repeated use of the word “real time” can change this fact. The real time data that the FDA has recently pushed for is not even particularly valuable as it is typically referred to as “not clean” until database locks (another human dependent bottleneck). I do not see drugs getting to market significantly faster but I do see more drugs possibly getting approved over time, but this again depends on humans running both ends of the process effectively.

Do community sites get absorbed into data platforms? Does the line between tech company and site continue to blur? What about between site network and CRO? Do they get bypassed entirely as agentic systems route around them? Or do they finally get repriced as the last mile of the entire industry?

I have opinions on all, and they’re not the same opinion for every context.

Sites that are pure throughput and especially in competitive markets, get squeezed. The agentic stack eats them. There’s no defensible moat in runnings standard ads to get leads for simple trials almost anyone can do, or being a body in a chair when an AI can do study startup, screen patients against EHR data, and route enrollment automatically. That’s middle work, and the middle is exactly what gets compressed.

Sites with real community trust, real PI relationships, and real access to populations the data platforms can’t reach from a Salesforce dashboard are the opposite. These individuals have influence within their communities. They can do studies that are challenging and require many moving parts. They become more valuable, not less. That’s last mile work, and they will be paid handsomely. A data platform can identify a candidate in Yuma or Killeen or San Bernardino. It can’t consent them, dose them, draw their labs, or follow them for two years. It doesn’t have the privilege to have ever earned their trust. Just like Amazon can route a package from Seattle to Phoenix in 18 hours but still needs a human to walk it up your driveway (drones are coming by the way).

The middle is where you get squeezed. The middle is where most of the industry currently lives and fights over because it is easy to build the workflows.

The CRO side is going to feel this even harder. If agentic study startup or site selection actually works, a meaningful chunk of what mid-size CROs charge for becomes software. Don’t get me started on source data verification. The big CROs adapt or acquire. The small ones either specialize into something the agents can’t do, or they get rolled up.

This is the part of the AI conversation almost nobody in our industry is having, because they’re still arguing about whether AlphaFold counts as drug discovery or whether the FDA hollering “real time” is a monumental shift towards some fantasy. Meanwhile the actual industry is restructuring around them.

I’m very bullish on AI in clinical research. I’m probably the biggest advocate for it that I know. But the next 5 years of winners and losers in this space don’t get decided by who has the best molecule. They get decided by who owns the compressed middle, and who owns the first and last miles.

If you’re a sponsor, a site operator, a CRO, or an investor in any of the above, this is the conversation worth having right now.

Where do you think the value accrues? The data platform layer that owns the new middle, the agentic operations layer that runs it, or the community site layer that owns the last mile?

There is a specific kind of setup I look for in clinical stage biotech. It has four ingredients: peer-reviewed validation of the science, a clearly scheduled pivotal trial start, analyst conviction measured in raised price targets not just maintained ones, and a regulatory tailwind that the market has not fully priced. AtaiBeckley (NASDAQ: ) has all four, right now.

I’ve been watching atai Life Sciences since before it merged with Beckley Psytech. The merger closed in November 2025, and the combined company is the cleanest public pure-play on psychedelic therapeutics anywhere in the world. Most mental health biotech falls into two buckets: derivative SSRIs and proven ketamine analogs. AtaiBeckley has broken out of both.

The BPL-003 data is real

Let me start with the science because that is what actually matters. BPL-003 is an intranasal formulation of mebufotenin benzoate, a short-acting psychedelic. It is designed to be administered in a clinic with a ninety-minute window, which is roughly half the operational footprint of psilocybin therapy models and competitive with esketamine on clinic throughput.

The Phase 2a data, which was just published in peer-reviewed form in CNS Drugs, showed roughly 66.7% of treatment-resistant depression (TRD) patients responded by Day 2. Not Week 6. Day 2. And the majority of responders held that response through Week 12, all while staying on their SSRIs. No serious safety issues reported.

I want to be careful here. Phase 2a is not Phase 3. A 66.7% response rate in a small open-label readout is not the same as a 66.7% response rate in a randomized controlled Phase 3. But there are two things about this readout that matter for where the stock goes from here.

First, the speed of onset is differentiated. TRD patients have already failed multiple SSRIs and likely esketamine. What moves the needle clinically and commercially is rapid durable response, and that is what the data shows.

Second, peer review in CNS Drugs is not a trivial signal. That journal’s editorial process scrutinizes mechanistic plausibility, safety profile, and trial design. Publication is not endorsement, but it reduces the risk that the Phase 2a data falls apart on closer inspection. Combine that with FDA Breakthrough Therapy Designation and a successful End-of-Phase 2 meeting, and you have a science package that is built for Phase 3.

The Phase 3 clock starts this quarter

AtaiBeckley reiterated this month that BPL-003 Phase 3 initiation is on track for Q2 2026. Two parallel Phase 3 studies in TRD. The End-of-Phase 2 FDA meeting was successful. The company is in the design window, lining up sites, locking down protocol.

This is the single most important calendar item for the next six months. First patient dosed will be a stock-moving event. The protocol design, once disclosed, will also tell us how the company is thinking about blinding, which in psychedelic trials is the hardest problem to solve.

As someone who runs clinical trials for a living, I will tell you the blinding problem is not solved here. Psychedelics produce unmistakable subjective effects. If patients know they are getting drug versus placebo, expectancy contaminates the readout. The way FDA and sponsors have handled this is with active placebos, dose-ranging, and careful statistical handling. I want to see which approach AtaiBeckley takes in Phase 3.

Analyst upgrades are stacking in one direction

Guggenheim raised its ATAI price target from $11 to $16 while maintaining a Buy rating. Deutsche Bank initiated coverage with a Buy and a $12 target. The mean Street target is roughly $13.33 against a current price near $4.71.

That is not a maintained rating in a sleepy note. That is fresh conviction. Guggenheim specifically cited rising conviction in BPL-003 as it enters Phase 3. Deutsche Bank initiating is especially interesting because it means a large sell-side desk decided this name was worth formal coverage right now, not six months from now.

The passive flow bid

Effective March 23, 2026, AtaiBeckley was added to the S&P Total Market Index, the S&P Completion Index, and the CRSP U.S. indices. CRSP inclusion, in particular, triggers mandatory position-building by passive index funds tracking more than $3 trillion in assets. Vanguard’s Total Stock Market funds are the most visible buyer.

This is a structural bid that the short thesis does not account for. When a small cap biotech gets added to these indices, you get forced buying at every dip, regardless of the clinical setup. It does not change the long-term value, but it does change the tape.

Regulatory environment: rare tailwind, not headwind

For most of my career, FDA’s posture on psychedelics was skeptical at best. That has changed. The current administration has signaled, via executive order, support for psychedelic therapy review pathways. FDA has shown willingness to engage on REMS-style programs, building on the Spravato (esketamine) playbook.

What does that mean operationally? Spravato sites already exist. Clinics that can administer esketamine under REMS can, with training, administer BPL-003 under a similar program. The infrastructure is half-built. That is a real commercialization advantage versus the classical psilocybin model, which requires six-to-eight-hour monitoring windows.

Spravato did roughly $1 billion in annual revenue by 2024. That is the proof-of-concept commercial data point for a clinically administered rescue therapy in TRD. BPL-003 is not Spravato, but the go-to-market pattern rhymes.

Competitive landscape, honestly

I want to be honest about competition because reflexively talking my book is not useful.

Compass Pathways (COMP) has psilocybin in Phase 3 TRD. They were first. They also have the longer monitoring window, which is a real operational headwind.

Johnson & Johnson’s Spravato is the incumbent. It has the brand, the REMS network, the reimbursement playbook. But it has also hit the ceiling of what adjunct esketamine can do for patients.

MindMed (MNMD) has LSD in generalized anxiety, a different indication and a very different patient population.

AtaiBeckley’s wedge is rapid-acting plus shorter in-clinic time plus a second asset (VLS-01 buccal DMT film) running behind BPL-003 in the same indication. Two shots on goal in TRD from the same company is unusual.

What could go wrong

I am not here to be a cheerleader. Here is what could break this setup:

1. Blinding failure in Phase 3. If expectancy confounds the data, regulators may push back on the pivotal results.

2. Placebo response rate creeps up. TRD trials are notoriously susceptible to high placebo response. A high placebo arm response rate can eat the effect size.

3. REMS burden. If FDA asks for a more onerous REMS than Spravato, commercialization economics get harder.

4. Competitor acceleration. Compass Pathways is not standing still. A surprise positive Phase 3 readout from them would change the competitive math.

5. Dilution. AtaiBeckley raised about $150 million in October 2025. They will need to raise again to fund two parallel Phase 3 programs. Dilution is coming.

None of these risks are disqualifying, but they are real and should be sized in.

How I’m thinking about position sizing

At $4.71, the Street mean target is $13.33. Guggenheim’s raise to $16 is the bull case. Deutsche Bank’s $12 is the measured case. If you agree with the setup, the question is not whether to own it, but how to own it.

My personal approach in catalyst stacks like this one is to build a base position and then add into data events rather than ahead of them. The first patient dosed announcement will move the stock. Phase 3 protocol publication will move the stock. Any competitor data from Compass will move the stock. Manage the exposure, do not chase.

Bottom line

The setup is: peer-reviewed validation, Phase 3 starting this quarter, two analyst upgrades, $3 trillion in passive AUM forced to build a position, and a policy tailwind from Washington. That is as dense a catalyst stack as I have seen in small cap biotech this year.

If BPL-003 works in Phase 3, the stock is a multi-bagger from here. If it does not, the loss is real. Position sizing is the whole game.

As always: this is my read as someone who runs clinical trials and watches this space. I am not your financial advisor. Do your own work. I do own ATAI at time of publication.

One uncomfortable truth keeps resurfacing in clinical research, no matter which angle you approach it from.

The sites that figure this out now will thrive. The ones that don’t will be left wondering what happened.

The Bottleneck Is Us

The clinical research site is the real bottleneck. Not the sponsor. Not the CRO. Us.

Sponsors do not want to run trials through traditional sites. They have to. The regulatory framework, patient trust, investigator relationships, and physical infrastructure all run through us. That is real leverage.

But leverage only works if you use it.

Too many sites today cling to familiar, inefficient workflows and call it stability. It is not stability. It is slow obsolescence.

Understand Your Workflow Before You Automate It

Before you hire another recruitment vendor, build custom tools, or spend heavily on software, do one thing first: truly understand your own workflow.

Most sites struggle with new systems not because the software is bad, but because they have never clearly mapped what they actually do, step by step and exception by exception.

A strong CTMS or AI platform forces that discipline. It is not just an organizational tool. It is a mirror.

The smart approach is phased. Start with your patient database and recruitment. Then layer in billing, metrics, and compliance. One disciplined piece at a time.

You cannot automate chaos. You must understand it, streamline it, and then automate it.

The Question That Changes Everything

What if you could hire an employee who works 24 hours a day, 365 days a year, and costs about $200 a month?

That is no longer hypothetical. That is agentic AI today.

Sites already using these tools are filing documents, managing task boards, triaging emails, and flagging protocol amendments before they become compliance issues. Not perfectly, and never without oversight, but fast, consistent, and at a fraction of the old cost.

The shift underneath all of this is straightforward. We are moving from measuring tasks to measuring results.

The Real Problem Is Not Technology

The technology exists. The barrier is that our industry still rewards activity instead of outcomes.

Think about two coordinators on the same study with the same patient volume. One randomizes 70 percent. The other randomizes one. On a task-based scorecard they can look similar. Both showed up. Both made calls. Both logged activity. The output gap is what matters now, and it is becoming impossible to hide.

Large site networks are already raising the bar through faster data entry, tighter monitoring windows, and better performance metrics. The coordinator who once had a week to catch up now has days. The one supported by an AI agent finishes in minutes.

This is not a threat. It is a fact pattern.

Sponsor-Centric Reality and the New Rules of Outreach

Clinical research is fundamentally sponsor-centric. Greg Lipset said it plainly: we need to stop pretending otherwise. Sponsors pay for everything, and they have been burned too many times by inflated claims, missed timelines, and underperformance.

That changes how sites must operate.

Generic outreach no longer works. Sponsors want facts: proven experience with the specific indication, enrollment metrics, query response times, and clear evidence of reliable performance. Feasibility responses, websites, and outreach emails all need to reflect this standard. Transparency is non-negotiable. Daylight is the best disinfectant.

Sponsors are doing more thorough feasibility assessments than ever. Even strong sites get dropped if they respond slowly, fail to act as true partners, or under-deliver on what they promised.

What Jensen Huang and Dave Ricks Are Both Telling Us

Jensen Huang’s view is that the AI moment does not produce a world where everyone builds their own software. It produces an explosion of specialized software companies building better tools. More competition. More options for sites. The sites that adopt early get to pressure-test those tools and pull ahead. The sites that wait inherit whatever the network above them mandates.

Dave Ricks at Lilly has been equally direct. The bottleneck in drug development is not IP. It is everything downstream. The trials. The sites. The documentation. The enrollment. He is thinking about how to solve it from the sponsor side. If sites do not solve it from the inside first, someone will solve it from outside on terms sites will not like.

Both of them are pointing at the same window. It is open right now. It will not stay open forever.

What This Means Practically

You do not need to build your own CTMS. That is not your core competency. Randomizing patients and delivering clean, high-quality data is.

What you do need is an honest look at three things. First, whether you actually understand your workflows precisely enough to use the tools that exist. Second, whether you are measuring activity or outcomes. Third, whether your communication with sponsors reflects performance or just promises.

The sites already working on all three are pulling ahead. Not because they are bigger or better funded. Because they decided to start.

The water is warm. Jump in.

Pick one study. One workflow you actually understand. One tool. Shift your focus from activity to output.

You might be surprised how fast good becomes excellent.

Aquestive Therapeutics received positive feedback from its Type A meeting with the FDA on March 30, 2026. The meeting addressed issues raised in the Complete Response Letter issued in January 2026 for Anaphylm, the company’s sublingual epinephrine film. Investors welcomed the update as a meaningful de-risking event for the program.

During the meeting, Aquestive and the FDA reached general alignment on several key items. The FDA agreed that potential accidental chewing of the film can be managed through clear labeling language rather than requiring new clinical trials. This regulatory shortcut is expected to save significant time and development costs.

The agency also provided constructive comments on the design of the Human Factors validation study. Aquestive is incorporating the suggested changes to user groups and plans to submit an updated protocol for review. In addition, the FDA offered guidance on the supportive Pharmacokinetic study to ensure consistency with data from previous successful trials. No new efficacy or manufacturing issues were identified.

Aquestive also presented an improved pouch opening mechanism designed to make the package easier to open in an emergency while reducing the risk of tearing the film. The FDA acknowledged the design enhancements, which will be evaluated in the updated Human Factors study.

Anaphylm aims to become the first and only non-invasive, orally delivered epinephrine product for the emergency treatment of severe allergic reactions, including anaphylaxis. The thin film, roughly the size of a postage stamp, dissolves under the tongue and is absorbed through the mucosal membrane. Its compact size and durable, weather-resistant packaging offer clear advantages over traditional needle-based auto-injectors such as the EpiPen.

The company reaffirmed its plan to resubmit the New Drug Application in the third quarter of 2026. Official meeting minutes are expected in early May 2026, providing written confirmation of the verbal agreements. Aquestive is also advancing regulatory filings in Canada and the European Union, with submissions targeted for the second half of 2026.

While the regulatory path has improved, risks remain. Aquestive carries 13.5 percent Senior Secured Notes with principal repayments beginning in June 2026. The company’s primary revenue source, Suboxone, continues to face declining market share as a sunsetting product. Cash burn will persist until a potential commercial launch, currently targeted for 2027.

Competition in the needle-free epinephrine space is growing. Neffy, an epinephrine nasal spray, gained approval in 2024 and is already on the market. Aquestive must successfully differentiate Anaphylm on portability, ease of use, and durability to capture meaningful market share.

Overall, the Type A meeting delivered the clarity investors sought. With a defined timeline for NDA resubmission and no major new hurdles identified, Aquestive has strengthened its position. The coming months will focus on completing the required studies and preparing the resubmission package. Success here could mark a major milestone for the small-cap biotech and open the door to the first oral epinephrine rescue therapy.

Veeva Systems is worth over $40 billion. Peter Gassner built it because he saw that life sciences companies were being forced into software that wasn’t designed for their world. He came from Salesforce, understood the gap, then built something better. Early hyper-verticalization of software as a service (SaaS) paradigm that made many billionaires.

Here’s what I keep thinking about: Gassner was not a life sciences operator, he worked at Salesforce and IBM. He had deep software expertise and a co-founder with pharma commercial experience, and that combination was enough to build one of the most successful vertical SaaS companies in history (and a current SOS Save Our Sites sponsor).

So what happens when the operator is also the builder?

We’re starting to see it. Raymond Nomizu founded CRIO after running sites. Rick Greenfield, who I’ve recently interviewed on my show, founded RealTime-eSOURCE from the same place, as someone who had lived the problem and decided to solve it. Andrew Mitchell worked inside Veeva, Intralinks, and IQVIA on the SUSAR side of things, understood the pain from the operator’s seat, and went out and built YEZA.ai. These are not billion-dollar companies yet, and Yeza is still a startup, but they’re proof of concept for something important.

The person who understands the problem most deeply is the person who should be building the solution.

I’ve spent 21 years at the site level. And what I know, for that matter, what any site director, study coordinator, or CRA with enough time in this industry knows, is that the majority of the bottlenecks in clinical research still live at the site. The site is where the rubber meets the road. It’s where every idealistic white paper and well-funded theory eventually meets real-world execution. It’s where digital bits and 1’s and 0’s meet blood, sweat and tears from human beings. And it’s where the friction compounds quietly, year after year, normalized into workarounds that everyone accepts as just part of the job.

Those normalized workarounds are where I think the next biggest companies get built.

When everyone has accepted the pain as the cost of doing business, the person who removes it looks like a miracle worker. The person most likely to see which pain to remove is the operator. Not the VC who attended a conference and heard the panel on site challenges. Not the sponsor who reads the quarterly enrollment reports. The person who has been living it.

The timing for this has never been better. Jensen Huang put it plainly: AI is an intelligence factory. You buy the tokens, you turn them into high value solutions, you sell the solutions at a higher price. The new development language is English. The scarcity is no longer technical ability. The scarcity is operator knowledge, knowing which problem is worth solving and understanding it at a level deep enough to build something that actually fits. The skills needed are the ones that you have honed in the trenches combined with a natural curiosity as well as a good sense of truth and aesthetics.

Patient recruitment. Study startup. Financial transparency. The sponsor-site communication layer. These are not new problems. They are old problems that have been waiting for someone who understood them from the inside to finally have the tools to act.

That person exists. They are probably in this industry right now. They might be reading this. They might be you, or me.

I’m building. I’m investing. And I’m paying attention to who else is doing the same.

Dan Sfera is the founder of Yuma Clinical Trials, Coordinare, and DSCS CRO, and co-founder of Save Our Sites. He invests in early-stage clinical research infrastructure companies. Investor thesis and portfolio at dansfera.com.

Rocket Pharmaceuticals ($RCKT) has a date with the FDA on March 28th. In three days, the agency will either approve KRESLADI, a lentiviral gene therapy for severe LAD-I (a pediatric immune disorder that kills most children before age two), or issue a second Complete Response Letter that would put the company’s survival in serious question. There is no middle ground here. This is as binary as biotech gets.

The clinical case is nearly airtight. Nine pediatric patients treated. One hundred percent survival through up to 45 months of follow-up. Zero treatment-related serious adverse events. The data was published in the New England Journal of Medicine. The FDA waived the Advisory Committee requirement, a strong signal the agency has no unresolved clinical concerns (remember $AQST?). The prior Complete Response Letter in June 2024 was purely a manufacturing documentation issue, not safety nor efficacy. CMC paperwork. That distinction matters enormously when assessing resubmission probability.

The question has always been manufacturing. The FDA flagged CMC issues twice, once with a review extension in February 2024 and again with the CRL in June 2024. Two bites at the same apple suggests this wasn’t a simple paperwork fix. Rocket spent 13 months addressing what they described as “limited” additional data requirements before resubmitting in October 2025. The FDA accepted the resubmission. No new information requests have been publicly disclosed since. The three models converge on roughly 78 to 85% approval probability on public information alone.

Then you add the non-public signals. Insiders were buying in February, approximately $1 million in open market purchases, the largest insider buy in nearly a year, executed 4 to 6 weeks before the PDUFA. These are people who know exactly what was submitted to the FDA and what the agency’s response has looked like throughout the review. They are buying with information the market doesn’t have. That pushes the probability estimate meaningfully higher.

But the shorts are building. Short interest just spiked to approximately 18% of float, rising sharply in the final weeks before March 28. Someone with real capital is betting against approval, or betting the commercial story doesn’t justify the stock even if approval comes through. LAD-I affects roughly 50 to 100 diagnosable US patients annually. Even at $2 million per treatment, the revenue ceiling is modest. Bluebird bio got three gene therapies approved and went bankrupt anyway.

The PRV changes the math. Upon approval, Rocket qualifies for a Rare Pediatric Disease Priority Review Voucher, a transferable FDA fast-track ticket that large pharma companies buy to accelerate their own blockbuster launches. Recent transactions have cleared $150 to 200 million. Rocket’s entire cash position is $189 million. A PRV sale essentially doubles the balance sheet overnight with zero dilution, bridging the company to Danon Disease Phase 2 data in 2027 to 2028, where the real commercial scale lives.

The setup in plain terms. Insiders are buying. Hedge funds exited in Q4 2025, not because they think it fails, but because they don’t carry binary risk. Shorts are pressing the trade. If approval comes, 16 million short shares need to cover into a thin market. That’s a mechanical squeeze on top of fundamental approval upside.

The science is compelling. The signal quality from insiders is high. The short build is the only credible counterargument, and even that may be positioning for sell-the-news rather than genuine CRL conviction. I know I have been wrong on these binary plays but I think I will take a bite.

Three days. Watch the tape on March 28th.

This is not financial advice. Biotech binary events carry extreme risk in both directions. Position size accordingly.

There is a moment in every biotech investment thesis where the external variables start aligning with the internal ones. For Iovance Biotherapeutics, that moment may have quietly arrived on March 6, 2026, when FDA Commissioner Marty Makary announced that Vinay Prasad would be leaving the agency at the end of April.

To understand why this matters, you have to understand what Prasad represented, not just as a regulator, but as an ideological force inside the FDA’s Center for Biologics Evaluation and Research (CBER), the exact division that oversees cellular therapies like Iovance’s Amtagvi. I have done numerous discussions on his book, Malignant, where he blasts single arm trials and surrogate endpoints, the exact strategy Iovance is following. Let’s break it down.

The Prasad Problem

Prasad came to CBER with a long and well-documented academic record arguing that the FDA had been too permissive with accelerated approvals based on surrogate endpoints. He believed single-arm trials showing objective response rates were insufficient evidence to put drugs in front of patients. He wrote about this for years before he had any regulatory power. Then the Trump administration handed him the keys to CBER and he tried to turn the philosophy into policy.

The result was one of the most disruptive regulatory tenures in recent FDA history and I have since changed my mind to some extent on the value of anti-establishment folks occupying positions of power no matter how good it sounds on paper. In less than a year, the agency denied or discouraged approval applications for at least eight drugs. Companies received rejection letters or demands for additional randomized studies, adding years and tens of millions of dollars to development timelines. A gene therapy developer for Huntington’s disease became the flashpoint that finally ended his run, but the damage to the broader biologics space had been accumulating for months before that.

Prasad will resign in April, and his departure matters for Iovance for a reason that goes beyond general industry sentiment. Iovance’s entire pipeline architecture is a precise blueprint of everything Prasad spent his career opposing.

Why the IOVA Thesis and the Prasad Philosophy Were Fundamentally Incompatible

The Barclays conference presentation by Dr. Brian Gastman in March 2026 laid out a pipeline strategy built almost entirely on the accelerated approval framework Prasad disdained. Single-arm registrational trials. ORR as the primary endpoint. Surrogate data in small, well-defined rare disease populations. The sarcoma program heading into a registrational trial in Q2 2026 has a 50% ORR signal from a pilot study at Memorial Sloan Kettering, with no randomized comparator arm planned. The NSCLC program in the IOV-LUN-202 trial is targeting the post-immunotherapy second-line setting with a durability argument rather than a head-to-head survival readout.

These are not unconventional designs. They follow a well-established regulatory playbook that the FDA’s Oncology Center of Excellence has validated repeatedly. Between mid-2025 and early 2026, the agency granted accelerated approvals to zongertinib, sevabertinib, datopotamab deruxtecan, sunvozertinib, and telisotuzumab vedotin, all in NSCLC, all on single-arm data with ORR endpoints. In the sarcoma space, afamitresgene autoleucel got through on a 43% ORR in 44 patients for synovial sarcoma. The precedents are stacked in Iovance’s favor.

But precedents mean less when the person reviewing your file has a philosophical objection to the evidentiary standard those precedents were built on. That was the Prasad risk. It was not primarily a scientific risk. It was an ideological one, and ideological risks are the hardest to price because they do not respond to better data.

The Replacement Bar Is Low, and That Is the Point

The bull case on Prasad’s exit does not require an optimistic successor. It requires a normal one.

Any career FDA reviewer who follows established precedent, respects the accelerated approval framework that Congress has codified, and does not carry a personal crusade against surrogate endpoints is already a materially better regulatory environment for Iovance’s pipeline. The bar is not high. The bar is “not Prasad.”

When the Q2 2026 sarcoma registrational trial opens and data begins to accumulate, whoever sits in the CBER director chair will look at a 50% ORR in a disease where the current standard of care produces less than 5% response rates and a median overall survival of nine to ten months. They will look at the afami-cel approval in synovial sarcoma sitting one shelf over as a direct precedent. Under normal regulatory conditions, the path from that data package to accelerated approval is logical and well-supported.

Under Prasad, it might have triggered a request for a randomized controlled trial that would have taken four additional years and been ethically questionable to run in a patient population with no meaningful alternatives. That scenario is now off the table.

The Underlying Thesis Remains Intact

None of this changes the biology. The real-world evidence for Amtagvi is outperforming the pivotal trial, with a 44% ORR across four authorized treatment centers compared to the 31.4% seen in C-144-01. Patients receiving the therapy after two or fewer prior lines are hitting 52% response rates, reinforcing the push toward earlier treatment sequencing. The gross margin improved from 43% in Q3 2025 to 50% in Q4 2025 as the internalized manufacturing model at the Iovance Cell Therapy Center continues to drive efficiency gains.

The infrastructure overlap thesis is equally undisturbed. The 80-plus authorized treatment centers built around the melanoma indication already utilize thoracic surgeons who manage lung metastasis resections for melanoma patients. When the NSCLC commercial launch arrives, those surgical teams are already trained, already onboarded, and already integrated into the TIL logistics network. The incremental cost of adding a new indication to existing infrastructure is minimal. No competitor starting from scratch can replicate that without years of investment.

The polyclonal mechanism still offers a durability profile that no ADC or TKI can match on biological grounds. The five-year melanoma data, with a median duration of response not yet reached beyond 36 months and 20% of patients achieving five-year survival, remains the most compelling long-term efficacy story in solid tumor immunotherapy.

The Cleaner Runway

What Prasad’s exit does is remove a layer of regulatory noise that was sitting on top of an otherwise strong thesis. It does not guarantee approvals. It does not eliminate the execution risk of expanding into community oncology settings, managing the logistics of a living therapy at scale, or hitting enrollment targets in rare sarcoma subtypes. Those risks remain and deserve honest accounting.

But the sarcoma registrational trial is opening into a post-Prasad regulatory environment. The NSCLC durability data will mature under a CBER director who presumably respects the evidentiary standards that produced every single-arm approval in the agency’s recent oncology track record. The next-generation programs in IOV-4001 and IOV-5001 will be evaluated by someone who did not spend a decade writing papers arguing that the entire accelerated approval framework is a failure of scientific rigor.

For a thesis built on the intersection of clinical innovation and regulatory pragmatism, that is not a small thing. The Iovance story was already compelling on the science. Now the regulatory environment is simply returning to the baseline that the science always deserved.

Most biotech analysts covering Aquestive Therapeutics have focused on the obvious story: a needle-free epinephrine film, a Complete Response Letter, a resubmission timeline, and a stock trading at a steep discount to analyst price targets. Those are all real and relevant. But after running the AQST thesis through multiple layers of analysis and years of being a shareholder, the most important question turns out to be one that almost nobody is asking out loud.

Can sublingual epinephrine actually be absorbed reliably during a severe anaphylactic episode?

It sounds technical. It is technical. And it matters enormously.

Anaphylaxis triggers systemic physiological responses that could theoretically compromise sublingual drug delivery. Angioedema causes oral swelling. Vasoconstriction reduces mucosal blood flow. Dry mouth and altered saliva pH change the absorption environment. The concern is not hypothetical. It is a legitimate pharmacokinetic question that the bull case needs to answer directly.

Aquestive has been answering it. Their NDA package included eleven studies covering approximately 967 administrations, including Oral Allergy Syndrome challenge studies designed to simulate the oral changes accompanying anaphylaxis. Results showed allergen-induced oral changes did not meaningfully alter Anaphylm’s pharmacokinetic profile. Median Tmax came in at 12 to 15 minutes, with Cmax bracketing autoinjector data at key intervals. The prodrug design, using dibutepinephrine that converts to active epinephrine post-absorption, structurally addresses the vasoconstriction problem. Chief Medical Officer Matt Greenhawt presented data at the AAAI conference showing Anaphylm produces no initial dip in diastolic blood pressure unlike injectable epinephrine, a meaningful pharmacodynamic advantage in anaphylactic shock where maintaining coronary perfusion pressure matters. The OAS gap between challenge studies and true severe anaphylaxis remains, but it is as narrow as ethical trial design permits, and the CRL flagged packaging and instructions rather than efficacy, which is the most important signal available.

The competitive landscape deserves honest treatment. ARS Pharmaceuticals’ Neffy is approved, needle-free, and on market. But its first full year of commercial performance tells a more complicated story than the competitive threat framing suggests. ARS generated 72 million dollars in U.S. net product revenue in 2025 while burning through 230 million dollars in sales and marketing expenses and posting a net loss exceeding 170 million dollars. The company is expanding its sales force from 106 to 150 reps in Q2 2026, a signal of continued market development effort rather than established dominance. In a roughly three billion dollar epinephrine market, Neffy’s year-one tally represents meaningful but far from conclusive penetration. More importantly, Neffy is still a nasal spray requiring assembly and priming. Anaphylm is a stamp-sized film with no device, no priming, and no needle, addressing the stigma and inconvenience barriers the December 2025 Duke-Margolis workshop identified as primary drivers of low carriage rates among adolescents and caregivers. Neffy validates that the noninjectable market exists. It does not close it.

The regulatory picture has continued to clarify favorably. ARS’ citizen petition challenging Aquestive’s clinical data package was denied by FDA in the weeks following the CRL, providing independent validation of the underlying science. The company has already modified packaging, contracted CROs for both the human factors validation and PK studies, and submitted a Type A meeting request. Cash at year-end 2025 stood at 121 million dollars. The RTW revenue interest financing of 75 million dollars has been extended through mid-2027. Launch Salesforce has been expanded from 50 to 75 reps. The Q3 2026 resubmission target is credible and management’s execution in the 31 days following the CRL supports that credibility.

The policy environment adds further tailwind. The Duke-Margolis workshop convened by FDA and the Duke-Margolis Institute in December 2025 produced explicit momentum toward expanded epinephrine access in community settings, nonprescription pathways, and alternative delivery forms. Every barrier that workshop identified maps onto what Anaphylm was designed to solve.

The thesis is not merely intact. It is better supported today than it was before the CRL, because the CRL itself clarified that the science is not the problem. The packaging was.

Let that sink in for a moment. A company spends years developing a novel sublingual epinephrine film, runs eleven pharmacokinetic studies across nearly a thousand administrations, navigates one of the most complex regulatory environments in medicine, and the thing standing between them and approval is essentially the same fold-and-tear pouch design used for Suboxone. The new packaging for the human factors validation study drops the fold entirely. Just tear and go. Problem, presumably, solved.

The instructions issue is similarly humbling. A meaningful portion of participants in the prior human factors study placed the film on top of the tongue rather than underneath it. Sublingual means under the tongue. It is in the name. To their credit, Aquestive is now printing administration instructions directly on the packaging so that the correct placement is impossible to miss. Whether this rises to the level of a formal regulatory deficiency or represents the most expensive reminder in pharmaceutical history is a matter of perspective. Either way, the fix exists and it is not complicated.

Then there is the chewing question, which deserves its own moment of reflection. Management is prepared to include an arm in the PK study addressing what happens if a patient chews the film rather than letting it dissolve. Their position, backed by existing data showing therapeutic absorption even when swallowed with eight ounces of water, is that chewing likely does not matter because absorption happens regardless. They plan to raise this with FDA at the Type A meeting and let the agency decide whether the arm is actually necessary. It is a reasonable scientific argument. It is also a sentence that would have seemed implausible as a pivotal regulatory discussion point before this process began.

To be clear, none of this diminishes the underlying science. The diastolic blood pressure data is genuinely interesting. The prodrug mechanism is elegant. The pharmacokinetic profile is competitive. The Duke-Margolis workshop handed the entire thesis a policy tailwind. Neffy burned through 230 million dollars in sales and marketing to generate 72 million in revenue, which means the noninjectable market is real, developing, and very much still available.

The most important remaining catalyst is not the packaging or the chewing arm. It is the self-administration arm of the PK study, which will directly compare outcomes when a healthcare provider administers Anaphylm versus when a patient does it themselves under realistic conditions. That is where the real-world thesis gets validated or complicated.

Not financial advice. Just a longtime follower, fan, and stockholder making lemonade from what turned out to be a surprisingly administrative bowl of lemons. Watch the Type A meeting. Watch the self-administration PK data. Those are the catalysts that matter.

Back in 2020, I was bored out of my mind working from home during COVID. I picked up Walter Isaacson’s audiobook “The Code Breaker”, yes the same Isaacson who wrote Steve Jobs and Leonardo da Vinci. He chose Jennifer Doudna as his next subject because he genuinely believes CRISPR is the next great revolution after the digital age. After listening to that book, I got so excited I bought CRISPR stocks like crazy in a frothy market. CRISPR Therapeutics, Intellia, all of it. I lost around 70% of my money. Sold everything and said we are just a little early for this.

That was six years ago. And this week I bought Caribou Biosciences.

Here is the situation. Caribou has $159 million in cash sitting in the bank right now. The entire market cap is about $180 million. I bought some at $1.90 this week. That means the market is valuing their CRISPR platform, their clinical data, their patent portfolio, and everything else at roughly $20 million. Doudna and CEO Rachel Haurwitz both live in the Bay Area. There are houses out there worth more than $20 million. That is the paradox you are looking at with this stock.

Caribou was co-founded in 2011 in a Berkeley lab basement by Doudna and Haurwitz, who was Doudna’s grad student at the time. Nine years later, in 2020, Doudna won the Nobel Prize in Chemistry for co-developing CRISPR. Haurwitz became CEO from day one and has never left that seat. That kind of founder continuity is rare in biotech and it matters.

The company has had a rough run. Peak was $32 in October 2021. It has been a series of pivots and failures since then, which happens when your stock goes from $32 to $1.50. But they have something new now. They call it chRDNA, which stands for CRISPR hybrid RNA-DNA guides, and it gives their molecular scissors far more precision than first-generation CRISPR tools. They pivoted to off-the-shelf CAR-T cell therapy, which I have covered before on this channel. The problem with today’s CAR-T is that it is autologous, meaning it needs your own cells, costs $400,000, requires weeks of manufacturing, and only works at specialized centers. Patients who live more than a couple of hours from one of those centers often just never get it. Caribou is building the off-the-shelf version using healthy donor cells made in advance. Ship it anywhere. Any community oncology center.

The early data is genuinely interesting. Their lead program showed an 86% overall response rate and 53% of patients progression-free at 12 months in the best-performing cohort. Those numbers are comparable to Yescarta from Gilead, which is the current gold standard and costs a fortune, made from each patient’s own cells. Their multiple myeloma program showed a 92% response rate in early data presented at the Tandem Meetings earlier this month.

Now for the honest part, because I have been burned before and I am not going to sugarcoat this.

Phase 1 data in 35 patients is not Phase 3 data in 250. That best-performing cohort is cherry-picked, with younger donors, better matching, and carefully selected patients. When you open it up to a broader population in a randomized trial, efficacy almost always softens. There were also two Grade 5 events across the two programs. Grade 5 means fatal. In small Phase 1 populations that is a serious signal, and if similar events occur in Phase 3 the FDA could issue a clinical hold. Dilution is also likely coming. They burn about $27 million a quarter and a full Phase 3 will cost more than what is currently in the bank.

So why am I in it? The Phase 3 trial for the lymphoma program is expected to start in the first half of this year. You do not need to wait for data. You just need the study to start. In my world, study startup is a real event that moves markets on its own. Q4 earnings come out around March, which will tell us a lot about the funding situation. Any partnership news, particularly from Pfizer who already put $25 million in at $5.33 per share back in 2023, would be a significant catalyst.

I will be honest with you. Part of the reason I am in this one, beyond the undervaluation, is because of “The Code Breaker.” I have been a fan of Doudna’s work for years. I have done videos on this book. I am still trying to get her on the show, though Nobel Prize winners are tough to book on a small podcast. But at $1.90, with Nobel-level science behind it and a founder-CEO who has been in the seat since the Berkeley basement days, the valuation disconnect relative to peers is real enough that I think it is worth a position.

I am not planning to hold this through Phase 3 data, which is years away. I am watching the near-term (2026) catalysts. And I could absolutely be wrong. I have been before with this exact sector. This may not be a company I would mind holding onto long term if need be.

As always, this is educational content only. I am not a financial advisor. I am an amateur investor who works in clinical research. Nothing here is investment advice. Let me know what you think in the comments. I will be doing updates on this one as the catalysts develop.

The biotech sector is often a landscape of binary bets and speculative hype, but every so often, pattern recognition in the pharmaceutical industry allows you to see a perfect storm forming before the rest of the street. On February 24, 2026, Iovance Biotherapeutics (IOVA) hit a turning point that felt less like a standard earnings report and more like a PDUFA-level event. Between the 30% quarterly revenue growth to $87 million in Q4, full-year product revenue of $264 million that achieved guidance and demonstrated gross margin expansion to 50%, a breakthrough 50% response rate in sarcomas, and a new FDA Fast Track designation for lung cancer, the message is clear: the launch pad is ready.

The Pattern Recognition: Moats and Management

To understand why IOVA is currently undervalued, you have to look past the noise and engage seriously with both the bull and bear cases. Success in biotech requires a specific hierarchy of research: product quality, competitive landscape, management execution, and pipeline data. Iovance checks all of these, but the bear case deserves acknowledgment before dismissing it.

The bear case is real. Iovance lost approximately $391 million in 2025 on $264 million in revenue, and their $303 million cash position is likely insufficient to reach profitability without a capital raise. Dilution is a near-certainty in 2026 or 2027. Goldman Sachs maintains a sell rating for exactly these structural reasons. Any honest analysis has to put those facts on the table.

Those structural concerns are legitimate, and anyone sizing a position in IOVA should hold them clearly in mind. What tips the scale toward the bull case is not that the risks disappear, but that the underlying science and commercial momentum are outpacing the timeline those risks assume. Unlike many biotechs experimenting with unproven mechanisms, Iovance’s Tumor Infiltrating Lymphocyte technology has been researched for 40 years by pioneers like Dr. Steven Rosenberg at the National Cancer Institute. They are not taking a scientific gamble; they are industrializing a proven biological mechanism. By building a network of Authorized Treatment Centers and partnering with McKesson for logistics, management has constructed a commercial infrastructure that is genuinely difficult to replicate quickly. The manufacturing complexity itself, a 32-day turnaround from tumor resection to infusion, functions as a barrier to entry rather than just an operational challenge. Proleukin, which contributed $44 million to 2025 revenue, further strengthens this moat as a bundled component in the TIL regimen and serves as a leading indicator for Amtagvi uptake, reflecting coordinated use in treatment protocols.

Moving Up the Stages: From Last Resort to Second Line

The most important development in the IOVA story right now is not the sarcoma data or the lung cancer designation, as significant as both are. It is that Iovance is moving up the treatment line. Historically, TIL therapy was seen as a salvage treatment for the sickest patients who had exhausted all other options. The five-year data from the C-144-01 trial changed that perception among Key Opinion Leaders. With a median duration of response exceeding 36 months and roughly 20% five-year overall survival, oncologists are increasingly describing Amtagvi not as a last resort but as a treatment to deploy earlier, when patients are healthier and their T-cells are more functional.

This shift is commercially significant for a straightforward biological reason: earlier treatment means better T-cell quality, which means higher response rates, which means more centers treating more patients sooner. The real-world ORR of 44% reported at the 2026 Tandem Meetings, rising to 52% in patients with two or fewer prior lines of therapy, is the data behind that shift.

The Sarcoma Data: Platform Validation, Not a One-Off

The 50% Objective Response Rate in undifferentiated pleomorphic sarcoma and dedifferentiated liposarcoma patients was striking precisely because sarcomas have historically been among the coldest tumor types for immunotherapy. These were heavily pretreated patients with a mean tumor burden of 117 millimeters and more than two prior lines of therapy. Current second-line standard of care in this setting achieves less than 5% ORR with median progression-free survival of two to three months. The gap between those numbers is not incremental; it is categorical.

This data did not emerge randomly. It came from an investigator-sponsored pilot trial at Memorial Sloan Kettering that Iovance supported, building on the hypothesis that TIL therapy’s mechanism of recognizing diverse tumor antigens might overcome the immunotherapy resistance typical of sarcomas. What exceeded expectations was the magnitude of the response, not the attempt itself. That distinction matters because it speaks to platform biology rather than tumor-specific luck. Iovance now plans a registrational trial in Q2 2026 with FDA engagement on an accelerated approval path, which is consistent language with a Breakthrough Therapy Designation application in the background.

The Valuation Gap: What the Comparables Actually Suggest

The most useful valuation exercise for IOVA is not a discounted cash flow model given the current loss profile. It is a scenario analysis based on addressable market penetration.

Start with melanoma alone. Iovance generated approximately $220 million in Amtagvi revenue in 2025, the first full year of launch, in a market of roughly 8,000 addressable advanced (metastatic/unresectable post-PD-1) melanoma patients annually in the U.S. That penetration is still early. Add the NSCLC opportunity, which management describes as roughly seven times the size of the melanoma market, with interim data showing 26% ORR and a median duration of response not yet reached at 25-plus months against a standard-of-care benchmark of 12.8% ORR and 5.6 months duration. Add a sarcoma indication covering more than 3,000 U.S. patients annually with no approved immunotherapy option. Add endometrial cancer currently in Phase 2. Add a next-generation pipeline including PD-1 inactivated TILs (IOV-4001) and an IL-12 engineered TIL (IOV-5001) heading toward IND submission.

For context on how the market values high-conviction oncology assets, consider that Summit Therapeutics currently holds a market cap exceeding $12 billion with zero product revenue, on the strength of conviction around a single oncology lead asset ivonescimab (a bispecific antibody combining PD-1 and VEGF inhibition, essentially a next-generation checkpoint inhibitor). Iovance, with a proven FDA-approved product, $264 million in annual revenue, 50% gross margins trending upward, and a multi-indication pipeline, is valued below $1.5 billion at current prices. The gap between those two numbers is not explained by fundamentals. It is explained by market skepticism about whether a company burning cash at this rate can reach escape velocity before diluting shareholders significantly.

The path to a dramatically higher valuation runs through two milestones: the NSCLC supplemental BLA expected to support accelerated approval in the second half of 2027, and gross margin expansion toward 60 to 65 percent as manufacturing internalization completes. If those two things happen on schedule, the revenue and margin profile begins to resemble a viable path to profitability without continuous dilution. That is when institutional capital typically re-rates a biotech aggressively.

What to Watch

The near-term catalysts worth monitoring are a potential Breakthrough Therapy Designation for sarcoma, updated NSCLC data at a major medical conference in 2026, UK and Australia regulatory approvals expected in the first half of 2026, the EMA resubmission for European access, and the IOV-5001 IND filing in the first half of 2026.

We are watching the early stages of a new standard of care being built across multiple solid tumor types. The science is not speculative. The commercial infrastructure is real. The remaining question is execution and timing, and the data released today suggests both are moving in the right direction.

This is not investment advice. Do your own diligence, and remember that clinical trial results in small pilot studies do not always replicate in registrational trials.

We have entered into what economist Ray Dalio calls a “Stage 6” period: high debt, geopolitical rivalry, and fundamental restructuring of global orders. Between now and 2036, we’re facing a volatile decade where economic strain meets technological disruption. The “growth at all costs” mentality that dominated the 2010s is dead. What matters now is building systems that can outlast external shocks while integrating AI as core infrastructure, not just another pilot project.

Understanding Stage 6: Why This Decade Is Different

Ray Dalio studied 500 years of economic history and found that empires and economies move through predictable stages in roughly 100-year cycles. Understanding where we are in this cycle is critical for anyone working in clinical research.

Stage 1 emerges after major conflict when new institutions form. For clinical research, this was post-World War II when the Declaration of Helsinki and IRBs were established in response to concentration camp atrocities.

Stage 2 brings peace and prosperity. The system works, credit flows, innovation happens. This was the 1950s through early 1970s when modern clinical research frameworks matured and academic medical centers were the hubs of innovation at the time.

Stage 3 sees debt accumulation. The U.S. left the gold standard in 1971, ushering in decades of artificially suppressed interest rates and growing financial bubbles through the 1980s and 90s.

Stage 4 is the debt bubble bursting. The 2008 financial crisis forced central banks to print massive amounts of money rather than let the system collapse. Many millennial clinical research professionals like myself started their careers and/or sites during this chaos.

Stage 5 is where printed money creates wealth inequality and social tension. Asset owners got richer while the middle class got squeezed. Political polarization intensified from 2010 through the early 2020s. CROs hit record high valuations more or less around the end of this period (slightly after post COVID boom).

Stage 6 is where we are now. Internal conflicts intensify, debt reaches unsustainable levels, and the existing order starts fracturing. Geopolitical rivalry heats up. “Might makes right” replaces rules-based international systems. Governments make desperate moves because they’re out of good options (DOGE). Sites now emulate CROs when it comes to becoming the dominant power centers of the industry.

Stage 6 either leads to Stage 7 (war and destruction of the old order) or somehow muddles through to a new Stage 1. We can’t control this part but we can control our chances of success regardless. Stage 6 is volatile and dangerous for the unprepared. Yet within this chaos lies what I believe will be the golden decade for clinical research sites and professionals who understand how to navigate it.

The Capital Discipline Mandate

U.S. federal debt is projected to hit 120% of GDP by 2036. When governments carry this much debt during Stage 6, they cut things that seemed untouchable. Late 2025 saw over 380 clinical trials terminated overnight following NIH budget cuts. While these were federally funded trials, the ripple effects hit private industry through reduced basic science feeding the pipeline and constrained biotech funding.

For sponsors, this means every Phase III trial must justify itself before a single patient enrolls. The days of broad endpoints and “let’s see what happens” experimentation are over. Capital is scarce and expensive, especially with newly nominated Federal Reserve Chairman Marsh taking a deflationary stance on monetary policy. This means sponsors are moving toward “predict-first” operations, using AI and alternative preclinical models to de-risk assets upstream.

For site owners, you can no longer rely solely on “we can recruit diverse patients” as your value proposition. You must demonstrate efficiency, data quality, and compliance from day one. Sponsors cannot afford sites with learning curves. The margin for error has evaporated.

Solo sites can likely survive for a few more years, but joining a site network provides advantages in technology adoption, operational efficiency, and sponsor relationships. Most people are predicting massive consolidation at the site level during this next decade. The golden decade ahead will favor sites that are lean, fiscally disciplined, and technology-enabled. Budget negotiations are improving, but awards go to sites that demonstrate operational excellence, not just patient access.

AI Moves from Pilots to Production Infrastructure

By 2026, successful organizations have stopped running AI pilot projects. They’ve built it into their operational DNA. As one technology founder put it, “We don’t even call it AI anymore. It’s just the next iteration of tech.”

We’re talking about Agentic AI: semi-autonomous systems handling predictive enrollment modeling, real-time data quality checks, and automated regulatory submissions. These aren’t future concepts. They’re operational realities creating two types of workers across all stakeholders.

1. Type one learns to work alongside AI, using it to eliminate grunt work and focus on tasks requiring human judgment.

2. Type two resists it, ends up doing the grunt work AI hasn’t automated yet, for lower pay, until that too gets automated.

For site staff, the transformation centers on unified technology platforms that integrate your CTMS, eSource, and participant engagement into a single command center. No more logging into six different systems. Companies like CRIOare already delivering integrated platforms that accept additional tools like patient recruitment portals PatientACE as one example, creating one hub where coordinators can actually work instead of constantly switching between disconnected systems.

Data entry transcription from eSource to EDC is being eliminated by tools such as Gleam that work with both electronic and paper sources. This represents inning one of a nine-inning game. The technology exists now to cut administrative time by 40-60%, freeing coordinators to focus on patient care rather than paperwork.

What used to be entry-level training tasks (data entry, regulatory binder filing) are disappearing. Sites no longer need to “overhire” and hope one out of three new hires becomes a coordinator. You can hire appropriately, let new coordinators shadow experienced staff, and focus on actual study coordination work. The regulatory burden in clinical research means this shift will take a decade rather than happening overnight, but it’s coming.

The sustainability paradox is real: AI data centers may require 200 gigawatts of power by 2030. Leading biopharma like Lilly and AstraZeneca are responding with “Green AI” initiatives, optimizing cloud infrastructure and implementing smart energy IoT to align digital transformation with ESG commitments.

Geopolitical Fractures and the Global Diversification Imperative

Stage 6 geopolitics is reshaping the global research network. Data nationalism is real. The U.S. CLOUD Act creates sovereignty vulnerabilities for cross-border trials. China and Europe are building regulatory moats. The seamless global system is cracking.

CROs have been outsourcing certain functions to Europe and Latin America for years, particularly budget management and administrative tasks. Now entire trial operations are shifting geography for two reasons: cost pressure when capital is scarce, and geopolitical fragmentation when data sovereignty matters.

Latin America offers 40% faster enrollment at 30% lower cost than the U.S. Border sites are becoming strategic hubs and launching pads for trials expanding into Latin American markets. Sites positioned along the U.S.-Mexico border have geographic and cultural advantages in this shift.

Sub-Saharan Africa bears 25% of global disease burden but conducts a tiny fraction of trials. The market is projected to reach $1.68 billion by 2032, driven by diversity mandates and infectious disease research.

Asia-Pacific hubs like Singapore and Australia have become go-to locations for cell and gene therapy. Australia has long been favored for Phase I trials and now leads in decentralized trial frameworks.

For early-career professionals, opportunities over the next decade will concentrate in emerging markets. Learning Spanish or Portuguese isn’t just nice to have anymore. Understanding regulatory frameworks in Brazil, Mexico, or Southeast Asia could make you invaluable.

For experienced site staff, this creates consulting and training opportunities. Emerging markets need people who understand Good Clinical Practice, can train local teams, and help them meet FDA and EMA standards.

The Three Critical Skill Sets

Digital Literacy means understanding how data flows through systems, troubleshooting when eSource doesn’t sync, reading audit trails, and spotting anomalies. The jobs being created are “digital trial coordinators” and “informatics nurses” who manage AI-driven workflows while maintaining patient relationships. If you’re a CRC with five years of experience, this is your upgrade path. Learn the technology platforms sponsors actually use.

Regulatory Intelligence matters more than ever. The FDA’s “Plausible Mechanism” pathway allows ultra-rare disease drugs to gain approval based on evidence from a handful of patients, provided there’s clear mechanistic rationale and robust Real-World Evidence plans. However, what the FDA says versus what it does can diverge, as evidenced by recent rejections despite meeting stated pathway criteria. Understanding trial design implications, patient recruitment strategies, and post-approval data collection requirements will differentiate valuable site staff from interchangeable labor.

Risk Management is now foundational. The 2026 FDA-EMA collaboration established ten “Guiding Principles of Good AI Practice” mandating data traceability, algorithm transparency, and patient consent for AI use. These aren’t suggestions. They’re requirements. Sites that demonstrate understanding of risk rather than just compliance checkboxes will win contracts.

The Outlast Philosophy: Why Trust Still Matters

Despite automation acceleration and Stage 6 chaos, a counter-trend is emerging. Eric Markowitz’s “Outlast” philosophy captures this: longevity comes from centering trust, not just efficiency metrics.

We’re seeing rejection of the toxic “growth at all costs” culture that dominated the 2010s and early 2020s. The burn-and-churn approach where sites were treated as interchangeable vendors and staff turnover hit 40-50% annually is dying. Why? Stage 6 economics make it more expensive to constantly onboard new sites and train new staff than to invest in relationships (and agentic technology) that work.

The industry is rediscovering that you can’t automate trust. Study participants don’t trust machines; they trust people. Patient retention requires human beings, hopefully using technology to reduce burnout rather than increase it. Site staff training requires human mentorship, not just AI modules.

For individual professionals, this means reputation matters more than ever. Build real relationships with medical monitors. Develop therapeutic area expertise. Become known for quality data delivery, not just fast enrollment. In a world where everything else is being automated, trust is your leverage. As I’ve been saying since 2015: EQ over IQ.

Strategic Imperatives for Survival and Success

For Site Owners:

• Implement unified technology platforms and demonstrate ROI on eliminating duplicate work

• Train staff on eSource platforms your top sponsors use

• Track and publicize data quality metrics as competitive advantage

• Join or build site networks to share technology and operational costs

• Diversify therapeutic areas rather than depending on single specialties

• Maintain fiscal discipline with tight overhead and pivot capability

• Consider strategic positioning near borders for international trial access

For Site Staff (CRCs, Coordinators):

• Learn technology stacks deeply, don’t just use them superficially

• Develop therapeutic area expertise where you become genuinely valuable

• Network with professionals running international trials

• Understand Agentic AI concepts and experiment with personal applications

• Focus on skills machines can’t replicate: patient relationships, cultural competency, clinical judgment

For Early Career Professionals:

• Prioritize digital skills and regulatory intelligence over generic experience

• Learn Spanish, Portuguese, or Mandarin for emerging market opportunities

• Understand AI tools and how they change workflow rather than fearing them

• Seek positions at sites investing in technology infrastructure

• Build competency in decentralized trial models and remote patient engagement

Conclusion: Building for Endurance

The next decade won’t be won by having the fanciest tools or fastest timelines. It will be won by building resilient systems that balance AI efficiency with the Outlast philosophy of trust and long-term thinking.

We’re transitioning from a world of scientific promise to one of operational control, where success goes not to the fastest movers but to those built to withstand historical volatility. Clinical research has survived regulatory overhauls, financial crises, and pandemics. We’ll navigate Dalio’s Stage 6 too.

The consolidation wave is already here. Mergers and acquisitions at the site level are accelerating. Biopharma faces a patent cliff creating its own consolidation pressures. These are tailwinds for prepared sites and professionals, not headwinds.

This is the golden decade for clinical research, but only for those who build for endurance rather than speed. Quality matters more than velocity. Build the right skills. Build real relationships. Build systems that can deliver what sponsors need: efficiency without sacrificing trust, technology without eliminating humanity, progress without burning out your people.

Stage 6 periods in history are rough. But they’re also when new orders emerge and opportunities crystallize for those positioned to seize them. The winners over the next decade will be those who understand that survival isn’t about moving fast. It’s about outlasting the cycles.

Watch my latest podcast on this topic here (please subscribe) I am almost at 100k!