The Fed just made the case for being cautious on growth.

Nvidia just made the case for staying invested in it.

Both can be right.

That is the interesting part of this week’s market.

Kevin Warsh’s Jackson Hole comments pushed September rate-hike odds from roughly 25% to 55%. At the same time, Nvidia delivered $96.2B in revenue, $59.7B in net income, and guided to $108B next quarter.

So we now have two powerful forces pulling against each other:

Sticky inflation + higher-for-longer rates
vs.
AI demand that continues to accelerate.

The question isn’t simply “Is the market bullish or bearish?”

The better question is:

Which side is likely to matter more for stocks over the next 6–12 months?

That’s the debate I’m focused on right now.

Before we jump in, make sure to follow the Substack Chat. I share plenty of interesting setups, quick thoughts, and real-time market insights there throughout the week.

Portfolio Performance

Last 2 years: +227.6%

Last 1 year: +132.2%

YTD: +104.2% vs. SPY +12.8%

No options. No leverage.

There’s one high-return opportunity I’m watching closely that could potentially deliver a 3 to 5x return if the setup plays out.

Let me explain why…

Read more

The market had already made up its mind: small-launch company, former SPAC, permanently stuck in SpaceX’s shadow.

I saw something different.

I ran Rocket Lab through the same 6-lens investing framework I had used on Palantir (PLTR) at $7 in 2023.

It passed all six.

I bought.

Rocket Lab eventually traded above $150 before pulling back. Palantir became a 10–12x investment for me before I sold.

The important part isn’t the exact stocks or prices.

It’s the framework.

I built this system to answer two questions:

1. When is a growth stock worth buying?

2. When has the original investment thesis already played out?

Here are the six lenses.

Subscribe now

1. Margin Trajectory

Revenue growth alone isn’t enough.

I want to know:

Are the economics getting better as the company scales?

I look at gross margins, operating leverage, cost per unit, and the profitability of new business lines.

At $9, Rocket Lab wasn’t a highly profitable company.

But the direction was improving.

Electron had already established a real launch business, while Space Systems was becoming an increasingly important and higher-quality revenue engine.

My rule: I want growth with improving economics, not growth that requires endlessly increasing spending.

2. Contracted Revenue & Backlog

Guidance is a forecast.

Backlog is visibility.

I want to know:

• Who is paying?

• How long is the contract?

• How difficult is it to cancel?

• Are customers coming back?

Government, defense and infrastructure contracts can create much stronger visibility than businesses dependent entirely on short-term demand.

Rocket Lab already had meaningful launch and Space Systems contracts when the stock was around $9.

The market was still largely valuing it as a small-launch company.

I was looking at the quality and durability of the revenue underneath the headline story.

3. The Second Engine

This is probably my favorite lens.

What part of the company does the market understand today and what part hasn’t been fully priced in yet?

Palantir at $7 was largely viewed as a government-services business.

But the second engine was the software platform and its opportunity to expand into commercial markets.

Rocket Lab was viewed as a rocket company.

But underneath that was Space Systems, plus the long-term opportunity from Neutron.

The second engine creates the potential for a major re-rating.

Once everyone understands it, the opportunity changes.

That’s exactly what happened with Palantir.

I didn’t sell because Palantir became a bad company.

I sold because the asymmetry I originally bought had changed.

4. Share Count & Dilution

This one gets ignored during bull markets.

I don’t.

Growth funded by issuing endless shares can destroy shareholder returns.

I track:

• Three-year share-count growth

• Stock-based compensation

• ATM programs

• Convertible debt

• Insider selling

• Cash runway

I don’t require zero dilution.

I want the business to grow faster than the share count.

Rocket Lab still carries dilution risk. Space is capital-intensive and Neutron requires enormous investment.

That’s a risk I accept, but I size the position accordingly.

Subscribe now

5. Operational Execution

Great stories are everywhere.

Great execution isn’t.

I want evidence that management can repeatedly deliver difficult things.

For Rocket Lab, that meant:

• Electron launch cadence

• Successful mission execution

• Space Systems deliveries

• Manufacturing progress

• Neutron development

A company doesn’t need a perfect record.

It needs a demonstrated ability to learn, execute and improve.

I trust operating cadence more than management promises.

6. The Forced Re-Rating

The final question:

What will force the market to recognize the value I’m seeing?

A stock can be cheap for years.

I need a reason that eventually changes.

That could be:

• A new product

• A massive contract

• Improving profitability

• Entry into a larger market

• A second business becoming meaningful

• A valuation disconnect becoming impossible to ignore

For Rocket Lab, I saw two potential catalysts:

Space Systems + Neutron.

For Palantir, it was the commercial software opportunity.

The key is knowing whether the re-rating is still ahead of you or already behind you.

Score Playbook

You can use these rules for your investments

6/6 — Buy candidate

The thesis is strong enough to consider a real position.

5/6 — Watchlist

Something important is still missing.

4/6 or lower — Pass

Why I Sold Palantir After 10–12x

This is the part most investors struggle with.

Finding a 10x stock is difficult.

Selling one can be even harder.

Palantir at $7 was an asymmetric opportunity.

After a 10–12x move, the situation was completely different.

The market understood the company.

The second engine was no longer hidden.

The re-rating had largely happened.

That doesn’t mean Palantir was a bad company.

It meant my original investment setup had changed.

I would rather sell a great company after a huge gain than hold forever simply because I was right once.

Why I’m Still Holding Rocket Lab

Rocket Lab hasn’t reached the same point in my framework.

The thesis still depends on execution.

Neutron can be delayed.

Cash burn remains important.

Dilution is a risk.

Space is a difficult industry.

But the original six-lens thesis hasn’t broken.

That’s why I still own it.

And if the facts change, I’ll change my mind.

That’s what the framework is for.

The Real Lesson

The skill isn’t buying Palantir at $7.

It isn’t buying Rocket Lab at $9.

The real skill is being able to look at 50 growth stocks and confidently say:

“No.”

Then, when one clears every lens, having the conviction to say:

“Yes.”

And when a winner has already paid you, having the discipline to say:

“It’s time to move on.”

This is my personal investment framework and is not financial advice. Always do your own research and consider your risk tolerance before investing.

Want the Stocks I’m Watching Now?

This is where I go deeper for paid subscribers.

I share the growth stocks, AI infrastructure companies, space stocks, and emerging opportunities that are currently on my radar, before they become obvious to everyone else.

Paid subscribers also get my detailed investment theses, watchlists, buy zones, portfolio updates, and the reasoning behind my highest-conviction ideas.

If you want to follow the same process I use to identify asymmetric opportunities, upgrade to paid and join me inside the research.

The goal isn’t to find every winner.

It’s to find the few that can make a meaningful difference while avoiding the ones that aren’t worth the risk.

Upgrade Now

I just added a new name to the portfolio tied to the humanoid robotics build-out.

Paid subscribers got the real-time alert last week when I bought it.

Here’s what caught my attention:

• Large multi-year design-win backlog.

• Revenue growth is accelerating.

• Early but meaningful traction supplying vision processors to humanoid robotics makers.

But the story goes much deeper than that.

Below, I’ll break down the full thesis, ownership data, growth numbers, key catalysts, risk framework, and the exact levels I’m watching.

Let’s dive in…

Read more

A lot happened this week.

I added one new position. A beaten-down company operating at the intersection of robotics, automotive sensing and edge AI.

Bitcoin finally started acting like the bottom signals I flagged back in July were.

Portfolio Performance

Last 2 years: +238.2%

Last 1 year: +153.4%

YTD: +116.6% vs. SPY +12.4%

No options. No leverage.

High-growth investing comes with volatility, but volatility can also create opportunity.

Let’s dive in…

Read more

Crazy Week for the Portfolio

We just went through one of the busiest earnings periods of the year. Some results were incredible, while a few positions gave me even more conviction in the long-term thesis.

The broader market remains interesting as well. Earnings are keeping investors optimistic, particularly around AI infrastructure, while inflation, Treasury yields and Fed uncertainty continue to create volatility.

For me, the key takeaway is simple: corporate spending on AI infrastructure remains extremely strong despite the macro noise.

AI Infrastructure: Still Accelerating

One of the biggest takeaways from earnings is that AI infrastructure demand is not slowing down.

The opportunity is much bigger than any individual company. The AI data center market is projected to grow from roughly $181B in 2026 to more than $810B by 2033.

Across my holdings, I’m seeing:

• Rapid revenue growth

• Record backlogs and customer commitments

• Accelerating AI demand

• Expanding infrastructure capacity

• Greater visibility into future revenue

Nebius ($NBIS) is the clearest example.

I first bought NBIS at $24. My initial position is now up more than 1,055%.

Its latest earnings only strengthened my conviction. Revenue grew 454% YoY

This is what I look for: a massive market, accelerating fundamentals and a company scaling into the opportunity.

And it wasn’t just AI.

My space, photonics, semiconductor, robotics and defence holdings also had several important developments this week.

I also made some changes:

• Fully sold one position

• Added to two positions

Portfolio Performance

Last 2 years: +284.5%

Last 1 year: +192.6%

YTD: +151.3% vs. SPY +13.7%

No options. No leverage.

Past performance is great, but I’m much more interested in where the next asymmetric opportunities are developing.

Let’s dive in…

Read more

Congratulations.

Seriously.

If you’ve been following me for a while and held onto Nebius through all the noise, fear, and volatility, take a moment to appreciate what just happened.

You didn’t just watch a stock go up.

You watched a thesis play out in real time.

I started buying NBIS last year around $24, when very few people were willing to believe in the story. There were plenty of skeptics, plenty of reasons to stay away, and plenty of people telling investors that the risk was simply too high.

Fast forward to today.

Nebius just delivered an absolutely incredible Q2, with $582.3M in revenue, up 454% YoY, while adjusted EBITDA reached $236M. The stock surged more than 30% following the results.

From roughly $24 to above $250.

That’s more than a 10x.

This is why I invest.

Let’s dive into what actually happened this quarter. The numbers are even more impressive when you look under the hood.

Subscribe now

Why This Print Is a Defining Moment

Nebius is one of the purest plays on the structural shortage of high-performance AI compute. In the neocloud space the key questions have always been the same: Is demand real and sustained? Can these companies price with power? Can they finance massive CapEx without destroying equity holders? And can they move beyond pure infrastructure into higher-value software?

Today Nebius answered all four with conviction.

Group revenue came in at $582.3 million, up 454% YoY.

Annualized run-rate revenue hit $3.0 billion at the end of June, a 56% sequential jump from $1.9 billion at the end of Q1 and nearly 600% YoY.

Adjusted EBITDA swung to $236.2 million (41% group margin) from a $21 million loss a year ago. The AI cloud segment itself posted an adjusted EBITDA margin of approximately 50%, up from 45% in Q1 and just 24% in Q4 2025. That is rapid operating leverage while the company is still investing heavily.

They sold out available capacity again as new clusters came online. They raised the year-end 2026 contracted power target to 5 GW. And management made it clear that demand continues to outstrip supply so aggressively that they could sell their entire planned 2027 capacity today at current terms, yet they are deliberately holding some back for higher-priced, shorter-duration opportunities.

The Landmark Deals That Changed the Narrative

The commercial highlight of the quarter was four landmark AI cloud contracts with an average total contract value of more than $1 billion each. Total contract value won in the quarter grew nearly 4x sequentially. Contract value from new customers jumped more than 9x.

The key deals include:

• Reflection — multi-year agreement to train and run its open-source models on the Nebius platform.

• Cohere — partnership focused on agentic AI solutions for enterprises.

• A U.S.-based AI neolab choosing Nebius for frontier model development.

• A large U.S. quantitative trading firm — one of Nebius’s first major financial services wins, selecting the platform for speed and reliability at scale.

Additional wins included AMI (co-founded by Yann LeCun), Basecamp Research, Prima Mente, and Higgsfield.

Average annual contract value on the core mid-term deals landed in the $20–25 million per MW range. Shorter-term (3–6 month) capacity is clearing at $40–50 million+ per MW. Around 70% of the Q2 deals included customer prepayments that cover 50–60% of the associated capital expenditures. That compressed the expected payback period to roughly one year and ten months.

Exact Words From Management

Arkady Volozh, founder and CEO, set the tone clearly:

“Everything we set out to do this quarter, we did. In most cases, we did more.”

“We closed our largest AI Cloud deals on our strongest terms to date, at prices that represent a step-change in the economics of our business.”

“Demand for AI capacity continues to grow exponentially, and we are converting that demand into contracted, profitable growth.”

“Q2 was our strongest commercial quarter yet.”

Management also stated they could sell the entire planned 2027 capacity under similar terms today, but are retaining some supply for shorter-duration, higher-priced opportunities.

Subscribe now

Capacity, Cash, and the Financing Flywheel

Nebius ended the quarter with approximately $8 billion in cash, supported by $2.3 billion of positive operating cash flow in Q2 alone. CapEx was heavy at roughly $5.7 billion in the quarter, consistent with the full-year $20–25 billion guidance that was fully reaffirmed.

What matters is how that CapEx is being funded. Customer prepayments are expected to exceed $9 billion in 2026. The company also closed its first asset-backed financing facility of approximately $775 million at attractive terms (SOFR + 2.50%). With more than $40 billion in total customer commitments already in place, management indicated they can do more of this type of financing.

The contracted power guidance trajectory has been consistently raised: from >1 GW in August 2025 to the current 5 GW target for year-end 2026.

Platform Differentiation and the Longer-Term Vision

Infrastructure alone is necessary but not sufficient. Nebius is deliberately building a full-stack AI cloud.

Token Factory, the managed inference platform, saw production workloads increase more than 3x in Q2. It is increasingly running open-weight models including Kimi K3, GLM 5.2, and Nemotron Ultra. During the quarter the company brought in the Eigen AI and Clarifai teams to strengthen inference optimization. Eigen at the model level and Clarifai at the system level.

The platform advanced with the Aether 3.6 release, which introduced Nebius Echo, an AI agent built into the console that allows natural-language control of infrastructure. Management’s stated view is that training and inference are converging. Unlike pure inference providers, Nebius aims to serve the full workload lifecycle on one platform, improving total cost of ownership and utilization.

This is the real long-term differentiator. The companies that win in AI infrastructure will not just rent GPUs. They will own the full stack, infrastructure + software + developer experience and turn that into higher utilization, stickier customers, and expanding margins over time.
Other Businesses & Equity Stakes

Beyond the core AI cloud (98% of revenue), Nebius still has a few non-core assets:

Consolidated subsidiaries

• Avride (autonomous vehicles): Fleet nearly 3x YTD, 1M+ autonomous miles, 60,000+ Uber rides. Still early-stage and loss-making (~$40M adjusted EBITDA loss in Q2).

• TripleTen (edtech): Launched new AI programs and expanded B2B. Smaller losses (~$9.4M in Q2).

These two businesses are the main reason group margins (41%) sit below the pure AI cloud margin of ~50%. Management treats them as non-core.

Equity stakes

• ClickHouse: Significant minority stake. January 2026 Series D valued the company at ~$15B. Nebius’s carrying value of the stake rose to $1.52B.

• Toloka: AI data business. Deconsolidated in 2025 after Bezos Expeditions-led investment. Nebius retains a meaningful economic stake but no longer has voting control.

The core AI cloud drives almost all the value.

What Investors Can Reasonably Expect

Looking forward:

• Capacity that came online late in Q2 should begin contributing more meaningfully to revenue in Q3 and beyond.

• The four landmark deals signed in Q2 are expected to ramp later in 2026 and into 2027.

• Pricing power appears intact.

• The combination of customer prepayments, asset-backed financing, and strong operating cash flow gives Nebius more flexibility to fund growth.

• Software and platform attach rates have room to expand.

• Formal 2027 guidance is expected later in the year.

Full-year 2026 guidance was fully reaffirmed: $7–9 billion ARR, $3.0–3.4 billion group revenue, approximately 40% adjusted EBITDA margin, and $20–25 billion CapEx. With first-half performance already strong, these targets look increasingly achievable and potentially conservative.

The broader AI infrastructure build-out is still in its early-to-middle innings. Power constraints, GPU supply, and the race to train and serve ever-larger models continue to favor specialized providers that can deliver density, software differentiation, and reliable long-term capacity. Nebius is positioning itself as one of the clearer winners in that race.

This was a defining quarter. The numbers, the deals, the pricing, the cash generation, and management’s language all point in the same direction: demand is real, the model is working, and the opportunity remains large.

My portfolio is currently up 140% YTD

Paid subscribers get:

• My full current portfolio with exact position sizes

• Real-time buy and sell signals

• Early access to new ideas

• Ongoing updates on how I’m managing risk and taking profits

The subscription price will increase soon. If you join now, you lock in the current lower price forever.

My Positioning, Price Targets & Next Steps

Read more

The market focused on losses and timing. I’m focused on demand, execution and what comes next.

If you only looked at the stock prices after earnings, you might think something went badly wrong.

I don’t think it did.

Both Rocket Lab and AST SpaceMobile reported earnings this week. Neither report was perfect. There were bigger losses, heavy spending and some questions around timing.

And the market reacted quickly.

But when I look past the first reaction, I see something quite different.

I see two companies moving closer to the point where their long-term stories become much more tangible.

The numbers first

Rocket Lab delivered $234 million of revenue, up 62% year over year, and ended the quarter with a record $2.36 billion backlog.

That is a pretty remarkable number for a company that is still in the early stages of what could become its biggest opportunity: Neutron. Rocket Lab also guided for another record quarter in Q3, with revenue expected between $250 million and $265 million.

AST SpaceMobile is a very different business.

Its Q2 revenue came in at $31.5 million, more than double the previous quarter, while backlog reached approximately $1.3 billion. The company ended the quarter with 13 satellites in orbit and continues to work toward commercial service as the constellation expands.

So we have two very different companies, at very different stages.

But there is one common thread:

Demand is showing up before the financial statements look pretty.

That’s important.

Rocket Lab has a growing backlog and a launch business that management continues to describe as extremely constrained.

ASTS is spending enormous amounts of money building a network that doesn’t yet resemble a traditional mature telecom business.

And that creates a strange situation for investors.

The numbers investors can see today are messy.

The numbers investors are really betting on are much further out.

Rocket Lab

The quarter wasn’t just about revenue.

Rocket Lab continues to move further beyond simply being a launch company.

Space Systems is now a major part of the business, while Electron continues to generate demand and Neutron represents the much larger long-term opportunity.

The company also announced the establishment of Rocket Lab Germany, expanding its push into sovereign European space capabilities.

Then there was GHOST.

The company unveiled a containerized launch system designed to provide rapidly deployable launch capability for orbital and suborbital missions.

The first site is planned for Kodiak, Alaska.

I think this is one of those announcements that is easy to overlook when everyone is staring at the earnings-per-share number.

Because it says something about how Rocket Lab is thinking about the market.

It isn’t just trying to launch more rockets.

It wants to become infrastructure.

And then there is Neutron.

Rocket Lab still expects to deliver the vehicle to the pad in Q4 2026, but management acknowledged that the window for an end-of-year launch is becoming tighter.

That was enough to give the market something to focus on.

And it did.

AST SpaceMobile

ASTS was much messier on the income statement.

Revenue was $31.5 million, below expectations, and the company reported a substantial net loss. Capex was also enormous as AST continues spending ahead of the commercial rollout.

But here’s the thing.

I don’t think ASTS investors should be looking at this company primarily through the lens of quarterly EPS.

The question is whether the network is actually getting built.

And it is.

AST currently has 13 satellites in orbit.

The next satellites are moving through the pipeline, with additional BlueBird satellites already in production.

The company continues to target commercial service as the constellation reaches the required scale.

That is the milestone I care about much more than whether ASTS beats quarterly EPS by a few cents.

Because once the constellation reaches sufficient scale, the conversation changes.

It moves from:

“Can they build this?”

to:

“How big can this become?”

That’s a very different investment debate.

So what actually changed?

That’s the question I wanted to answer with this earnings review.

Not whether the stocks went up or down.

I’m interested in the things that could still matter two or three years from now.

For Rocket Lab, that means Neutron, backlog conversion, launch capacity, Space Systems and national-security opportunities.

For ASTS, it means satellite production, launch cadence, commercial service, government contracts and the company’s ability to turn billions of dollars of investment into an operating network.

The market saw the misses.

I think there is considerably more underneath the surface.

Read more

Turning the Dip Into Opportunity

The recent pullback in growth stocks created some of the most attractive entry points we’ve seen in a while. Several positions added during the weakness are already generating strong gains.

• Multiple holdings bought during previous dips have delivered 30–40% gains in just a few days.

• Relative strength returned quickly once forced selling faded.

• The strongest companies recovered rapidly as buyers stepped back in.

This week, I added several new positions across the same high-conviction themes.

The opportunity is still there, but the question now is where I’m seeing the strongest risk/reward.

Below, I share my full portfolio, exact position weights, new buys and updated thesis for every major holding.

Performance Overview

Last 2 years: +235.6%

Last 1 year: +113.1%

2026 YTD: +91.7% vs SPY +13.1%

No options. No leverage.

Let’s dive in...

Read more

Bloom Energy is right in the middle of this AI power mess. Data centers need power now and the grid just can’t keep up. They sell solid oxide fuel cells that make electricity on site. Runs on natural gas mostly, sometimes biogas or hydrogen. No burning anything.

The company started in 2001 in San Jose. They’ve got about 1.8 GW out there now across more than 1,100 sites in 9 countries. The big question is whether this growth can actually keep going or if the risks start showing up in a big way.

Recent Earnings

Q2 numbers were strong.

• Revenue came in at $1.065 billion. Up 165% from last year.

• Product sales did most of the work at $935 million. That part grew over 215%.

• Gross margin hit 33.4% GAAP. Non GAAP was 34.3%.

• Operating income finally looked solid at $182 million on a GAAP basis.

• Operating income around $240 million.

• Cash from ops was $226 million.

• Ended the quarter with $2.67 billion in cash.

Full year 2025 revenue was $ 2.02 billion. They raised the 2026 guide pretty hard.

• Now looking at $3.9 to $4.2 billion.

• Basically doubling at the midpoint.

• Gross margin expected near 34%.

• Operating income $800 to $900 million.

• EPS $2.55 to $2.85.

Operating leverage is showing up. Margins moving higher. Cash flow finally positive which is nice to see.

Subscribe now

What They Actually Do

They build modular power systems. Customers start small then add more. Scales up pretty cleanly.

Stuff that matters for AI sites:

• Goes in fast. Weeks or a few months not years.

• Tracks the load in real time so those AI spikes dont kill it.

• Emissions stay low so permitting is easier than turbines.

• Works with high voltage DC which fits the server racks better.

• Fuel is flexible. Natural gas today, hydrogen later if needed.

Big hyperscalers have signed off on it. Oracle has a large deal. AEP and Brookfield helping with financing. Colos and neoclouds are in there too. They claim high share in the commercial fuel cell part of this market.

Service contracts last years after install. Makes the revenue stickier.

Management

K.R. Sridhar started the company and still runs it. Knows the tech deep and talks clearly about the power problem for AI.

Simon Edwards came in as CFO in April from Groq. Background in scaling AI infrastructure finance so that fits.

Rest of the team has solid ops and manufacturing experience. Continuity looks okay. Real test is whether they can keep shipping clean when volumes jump a lot higher.

The Moat

They have some real edges right now.

• The solid oxide stack is hard to copy. Patents and process knowledge matter.

• Time to power is short. Grid and big turbines take way longer.

• Modular so you can add capacity without starting over.

• Lots of field data from years of running these things.

• Once its installed switching costs go up because of the service contracts.

Turbines still win on pure scale sometimes and they have longer track records. Other fuel cell companies are behind on commercial stationary volume for data centers. Current window favours Bloom. But they need cost downs and real long stack life to keep this advantage.

Risks

Performance is one issue. Hunterbrook put out a short report in July using public generation data. Claimed a lot of systems fall below the efficiency and output numbers they market, and earlier than expected. Stack life might be shorter. Bloom says that’s wrong. We’ll see what the field data shows over the next few quarters.

Scandium is the other one. It’s in the electrolyte and helps performance. The short report said they still get material through China linked routes even though management kept saying no China dependence. Supply outside China looks tight if they ramp hard. Bloom put out an 8K and a blog post saying supply is diversified and they have enough for current needs plus 25 GW of capacity. Lawyers already filed class actions. If costs jump or supply gets tight margins will feel it.

Financial side is better than before but not clean.

• Debt around $2.5 billion, mostly convertibles.

• Cash at $2.67 billion covers a decent chunk.

• Dilution from the converts is still there.

• Those big framework deals still have to turn into real shipments at good margins.

Customer concentration is high. Few big hyperscalers drive a lot of it. Financing has to stay available. Gas prices and tax credits can move the numbers. Turbine competition will increase if their lead times improve.

Quality

Volume is helping margins. Cash generation better. Balance sheet equity stronger. Hyperscaler validation is real. Service revenue gives some visibility. Those are positives.

The open questions on how long the stacks actually last and the scandium supply keep it from feeling higher quality.

Valuation and Price Targets

Read more

Why Risk Management Is Crucial In Growth Investing

People obsess with “what stock should I buy?”
Very few obsess over “how big should this stock be?” and “what’s my maximum loss if I’m wrong?”

Growth names are:

• More volatile.

• More sensitive to sentiment and macro.

• More exposed to execution risk.

You can’t escape that volatility, but you can decide how much of your portfolio is exposed and how you react when things go wrong.

In my experience, the investors who survive and compound are the ones who:

• Cap their risk per position.

• Size up and down based on conviction.

• Treat drawdowns as normal

Step 1 – Set A Hard Cap On Risk Per Position

The first rule is simple:

No single idea is allowed to put the entire portfolio at risk.

You’ve probably seen this framed as “never risk more than 1–2% of your capital per trade” in trading literature.

Translating that into growth investing:

• Decide your max loss per position (example: 1–2% of total capital).

• Combine position size and stop‑loss level so that, if your stop is hit, the loss stays under that cap.

Example with round numbers:

• Portfolio: 10,000.

• Max loss per idea: 150 (1.5%).

• If you run a 15% position (1,500), your stop needs to be ~10% below entry.

• If you want a looser stop, you shrink the size.

This is not guessing. You’re doing the math before you buy.

Step 2 – Define Base And Max Position Sizes

• Base position: 5–10% of the portfolio.

• High‑conviction max: up to ~15–20% in very rare cases.

• Absolute cap: nothing goes above ~25% of total.

You can tune the ranges to your risk tolerance, but the logic stays:

• Most ideas live at the base size.

• A small handful earn a bigger slot.

• No single stock is allowed to become your entire future.

This is where your research, expertise, thesis work starts to matter.

Step 3 – Tie Size To Volatility (Not Just Conviction)

Conviction is emotional.
Volatility is measurable.

I like to combine them.

For high‑beta names, I willingly scale down size, even if I love the thesis.
For more stable growth names, I’m comfortable running closer to my base or high‑conviction size.

You can keep this simple:

• High volatility: small position.

• Medium volatility: normal.

• Low volatility: larger, but still within your caps.

This way, each position contributes a similar amount of risk.

Step 4 – Stage‑Based Sizing: Earn The Right To Concentrate

One thing I really like from trader education is the idea of stages.

• Stage 1 (learning/rebuilding):

  • Caps around 5–10% per position.

  • More names, smaller sizes.

  • Focus on execution and discipline.

• Stage 2 (consistent/confident):

  • Base size still ~10%.

  • Move up to ~12–20% when multiple edges line up: fundamentals, technicals, macro, and your process all agree.

• Stage 3 (proven):

  • 20–30% positions are possible but only with strict stop‑loss logic and a deep track record of following your own rules.

You don’t “deserve” a 30% position just because you like the company.
You earn it by showing you can:

• Cut losers without drama.

• Sit through volatility without panic.

• Respect your own plan in real time.

  Subscribe now

Step 5 – Adjust Size To The Market Regime

Your skill is not constant.
The market is not constant.
Your size shouldn’t be constant too.

In strong, trending environments, you gradually scale up.
In choppy or brutal environments, you cut size aggressively. Sometimes stay mostly in base positions.

Some rules that help:

• After a string of losses or a major correction, halve your normal size until you’re back

• Only increase size in steps (e.g., 10% → 12% → 15%). Never jump from tiny to massive.

• Let your equity curve and your stress level guide size, not FOMO.

You’re not fighting the market. You’re listening and adapting.

Step 6 – Sell When The Business Breaks, Not When The Price Dips

Fundamental change is the only sell trigger for high‑conviction growth names.

• Revenue and user growth clearly slow or turn negative.

• Margins and returns on capital deteriorate, not just QoQ noise.

• Management damages the moat with bad capital allocation. Strategy shifting in the wrong direction.

You sell when the story you bought is no longer true.

The Psychological Side

Most people blow up because they:

• Double down in anger after losses.

• Refuse to cut obvious losers.

• Oversize positions out of FOMO or ego.

A few habits help:

• Limit the number of names you hold when you’re still learning. Complexity increases emotional load.

• Write your rules down before entering: entry, stop, adds, trims, max size.

• Treat each trade as one sample in a long series.

You’re building a process that will carry you through many cycles. Do not try to win one trade.

How I Apply All Of This In My Own Portfolio

I use it to manage a high‑conviction portfolio with a couple dozen positions and real swings.

If you want to see how these ideas look in practice:

• How I size my AI, semiconductor, space, and defence names.

• Which positions I allow to get big and which stay small.

• How I deal with drawdowns while still targeting aggressive long‑term growth.

Then you’ll like my portfolio article, where I walk through my full positioning.

If you’ve read this far, you care about real risk management, concentrated growth portfolios, and surviving volatility without losing upside.

That’s exactly what I share with paid subscribers:

• Full current positioning and sizing.

• Deep‑dive theses on my highest‑conviction names.

• Real-time buys and sells

My subscription is still at early‑supporter pricing, and it will increase as the publication grows, but existing subscribers are grandfathered at their original rate.

If you join now, you lock in today’s lower price forever

Subscribe now

It is +82.3% YTD and +202.6% over the last 2 years. S&P 500 is up 9.4% YTD.

The last few weeks have been rough for AI stocks. Retail dumped, leveraged accounts liquidated. Narrative flipped from “AI is going to change everything” to serious doubts about whether infrastructure spending will actually pay off. These kinds of shakeouts are normal in every big tech cycle.

The sentiment has turned bearish but spending isn’t slowing down.

Morgan Stanley’s numbers show hyperscaler capex is expected to more than 3x from $426 billion in 2025 to $1.4 trillion by 2028. This isn’t short-term. It’s a multi-year buildout forced by real physical limits around power, cooling, interconnect and compute.

Our portfolio will benefit massively from this.

The gap between the current fear and spending is exactly where asymmetric opportunities happen.

In July we made a few moves in the portfolio.

• We fully exited 4 positions

• Opened 3 new ones.

Below I’ll walk through the recent portfolio changes and give updates on the major positions.

Performance Overview

- Last 2 years: +202.6%

- Last 1 year: +103.7%

- 2026 YTD: +82.3% vs. SPY +9.4%

No options. No leverage.

Let’s dive in...

Read more

The last few weeks watching AI infrastructure felt like a slow-motion train wreck. First, there was huge selling, then margin calls, followed by forced liquidations. Leopold Aschenbrenner’s Situational Awareness fund got crushed. By June, it was up 439%, hitting almost $45 billion. Then the fund had to dump most of its publicly traded stock onto the market. Citadel ended up buying a big chunk of it.

Also, over 350,000 Korean retail investors got wiped out on their leveraged AI and semiconductor bets. This is insane! The total losses on those leveraged bets ran into the tens of billions. It was a brutal, public mess.

Then, things changed. A lot of people who were hyping the AI supercycle just three months ago are now doubting it. You hear things like “Capex peak,” “ROI problems,” and “the trade is overcrowded”

And yesterday topped it off.

AI infrastructure stocks that got hammered started to climb. The forced selling finally let up. The QQQ closed up over 3%. The big question this morning is pretty simple: Was that just the bottom?

We’ll get into that.

If you’re not already in the subscriber chat, you should be. That’s where I post every new trade as I take it. Thesis, risk, cost basis, the whole thing.

Join the chat →

For the rest of this piece, I’ll cover:

• What actually went down over the last 48 hours.

• If I think the bottom is really in.

• How this dip stacks up against past corrections.

• The specific QQQ levels I want to see to signal an uptrend.

• Which stocks are showing the clearest future signals.

• What I’m focused on after this market pullback.

This is a moment where being ready is key.

Read more

Good evening everyone,

I already updated the paid subscribers in the Substack chat earlier.

I sold three positions recently.
I opened one brand-new speculative name.
And I am still planning more dip buys.

I am executing a clear plan: protect capital, recycle into higher-conviction setups.

Below are my thoughts. ...

Read more

I’ve been staring at this weekly Bitcoin chart for days.

The same RSI condition that showed up near the 2019 bottom and again near the 2022 bottom just printed for the third time. Price is sitting around 64,650. RSI is hovering near 39 after dipping lower. The structure looks familiar in a way that makes me pay attention.

At the same time, we have fresh institutional capital flowing into the crypto complex, regulatory clarity advancing in Washington, and an emerging structural demand driver that most charts completely ignore: AI agents that need programmable money.

This is not a call that the exact bottom is in today. It is a clear-eyed look at why the current zone offers limited downside relative to the upside if the historical pattern and the new catalysts play out.

This is the kind of setup that has worked for me before. The approach that got my portfolio past 240% in the last 2 Years was never about predicting exact bottoms. It was about waiting until a few independent things lined up and then sizing for the asymmetry. That same process has let me outperform the broader markets by a huge margin.

Subscribe now

Reason 1: The Weekly RSI Pattern

Look at the weekly

In late 2018 into early 2019, weekly RSI dropped into the low 30s. Price bottomed and the next major advance began.
In 2022, after the FTX collapse, the same RSI zone was tagged. The low formed and the multi-year rally followed.
Today RSI sits near 39–41 after dipping lower. Price is consolidating near $64,650 inside a range that looks structurally similar to the previous two boxed zones.

Three clear historical instances of the same high-timeframe condition. Markets are never identical, but this is the kind of repeatable technical observation I refuse to ignore.

Reason 2: Citadel Securities Just Wrote a $400 Million Check

On July 16, Citadel Securities invested $400 million in Crypto.com at a $20 billion valuation. It is the first institutional funding round in the company’s long history. The capital is earmarked for expansion into tokenized securities and derivatives.

Citadel is not a retail-facing venture firm. They are one of the most sophisticated market makers in the world. When they put this size of capital into crypto infrastructure, it is a signal that structural demand is expected to grow. Follow the money!

Reason 3: Institutions Are Still Accumulating Coinbase Heavily

Coinbase remains the primary publicly traded proxy for institutional crypto exposure. Recent 13F activity and ownership data show continued buying even through the drawdown. Specific large increases (one notable firm raised its stake by over 600% in the first quarter) and overall institutional ownership remaining high tell the same story the surveys have been showing all year: sophisticated capital is using volatility to add. Not to exit.

Again, follow the big money!

Reason 4: The Clarity Act Is Advancing

Regulatory progress is also moving, slowly but in the right direction. The Clarity Act passed the House last year. Senate Banking advanced it in May, and it hit the Senate calendar in June. It’s still waiting on a full floor vote, but the direction is clearer than it’s been in a long time. That kind of uncertainty has been a real overhang for bigger capital. Any real progress removes friction.

And then there’s the one that feels under-discussed.

Subscribe now

Reason 5: AI Agents Will Need Crypto

This is the catalyst most people are still underweighting.

AI agents cannot open traditional bank accounts. They have no human identity, no KYC profile, and no credit score. Yet they will need to hold value and make transactions.

The emerging solution is crypto-native. Stablecoins (especially USDC) are already becoming the default for high-frequency micropayments and agent-to-agent settlement because they are programmable, settle in seconds, and cost fractions of a cent. Bitcoin is increasingly discussed as the preferred store of value and treasury asset for those same agents.

Multiple 2026 studies of frontier AI models show a strong revealed preference for Bitcoin as a long-term holding and stablecoins for everyday transactions. Infrastructure is already being built (Coinbase, Stripe, Visa and others are racing to support agent payments). This is not science fiction. It is an early but real demand driver that did not exist in previous cycles.

To be honest, if AI agents scale the way the technology suggests, they will need money that works without banks. That money is far more likely to be Bitcoin and stablecoins than traditional fiat.

Reason 6: Terrible Sentiment

You can feel it across timelines, group chats, and comment sections. The tone is exhausted, dismissive, and in many cases outright hostile toward anything crypto. People who were loud during the last run-up are quiet. Funding rates, social volume, and general mood all point to the same place. Most participants have already given up or are waiting for lower prices that may never come. Extreme negative sentiment has shown up near every major bottom I’ve paid attention to. It doesn’t guarantee the low is in, but it is rarely present at the top.

Putting the Pieces Together

Technical pattern that marked the last two major lows.
Real capital from one of the world’s top market makers.
Continued institutional buying of the leading regulated proxy.
Regulatory clarity moving through Congress.
A new structural buyer class (AI agents) that prefers crypto-native money.

I talk about these kinds of setups actively with the people who are in the chat. If you want to be part of those conversations instead of just reading the finished notes, join the chat. That’s where the real-time thinking happens.

Next Steps

• Keep the weekly Bitcoin chart with RSI (14) open and mark the three historical lows yourself.

• Watch Coinbase institutional ownership updates and any further large 13F increases as a flow proxy.

• Track the Clarity Act Senate calendar. A floor vote would be a meaningful sentiment and capital catalyst.

• Monitor early metrics on AI agent transaction volume and wallet infrastructure. The preference data is already there; volume will tell us the speed.

• Size any exposure according to your own risk rules so that a failed pattern does not damage the broader portfolio.

If you want the full picture, join the paid side.

Here’s what you get:

• The complete portfolio

• Live position sizing on setups like this one

• The exact risk rules behind the 240%, large outperformance versus the market

• Ongoing updates as things develop

Right now a paid subscription still costs about the same as a burger.
That price will not last.

Anyone who joins today, locks in the current rate forever.
Once it goes up, the old price is gone.

The free posts will keep giving you the big picture.
The paid side is where you see the actual book.

Subscribe now

Share

Everyone on X was calling it dead. “Government software company, no growth, AI is overhyped.” The stock had been crushed. I just kept buying. Same shares I’d already owned for a while. That position ended up doing about 14x from those levels.

Same thing happened with Ethereum during the COVID crash. Bought a bunch when it was sitting around $100 while people were screaming the world was ending. A year and a half later, it was over $4,000.

Conviction is key.

I keep coming back to these charts. The numbers don’t really change. Bull markets have historically lasted much longer and made much more money than the short, nasty bear markets.

Average bull market: 5.3 years and +254%.

Average bear: about 1 year and –31%.

The up periods simply last longer and go further. It is not some optimistic theory. That’s the actual history.

Most people still treat every 10–15% drop like the end of the world. They sell good companies, sit in cash, and then watch the recovery without them.

What the people who actually made money did

I used to think successful investors had some secret timing skill. Turns out most of them just didn’t sell.

There’s a Vanguard study that looked at hundreds of thousands of accounts with at least $500K. The ones that did well weren’t the clever market timers. They held a big chunk in U.S. stocks, barely traded, and almost none of them sold during the COVID crash. 99% just did nothing!

It matches what I’ve seen in my own account. Look, the biggest gains didn’t come from perfect entries. They came from not interrupting the compounding when things got ugly. For example just to name a few.

This chart shows the same story differently. The long orange bars are the bull markets. The short blue dips are the bears. You can stare at it for a minute, and the pattern becomes obvious. Bull runs simply last longer and deliver a lot more.

Why quality companies get cheaper

The thing is, when the market falls, it doesn’t care about the good businesses from the bad ones. Everything gets marked down. That’s your opportunity.

In 2023, when Palantir was $7, the same thing was happening across a lot of growth names. The long-term story on the ones I already liked hadn’t changed. The price just got better. Same with Ethereum in March 2020. The network was still growing. The price had just collapsed with everything else.

That’s the real edge of staying long. You’re not trying to catch a falling knife. You’re buying more of something you already believe has a long runway, only at a better price.

Charts that Explain Market Volatility, Corrections and Bears.

This one puts the averages side by side. Bull markets have historically delivered roughly +279% over about six years. Bear markets took about 14 months and cost around –33%. The math is not complicated. The hard part is remembering all these.

A simple way I handle dry powder now

I keep a small amount of cash or short-term treasuries on the side. Not a huge pile. Just enough so that when a quality name I already own drops 20% or 30%, I can buy more.

When the market is calm, I just keep buying on my normal schedule. When things get ugly, I lean in a bit harder. When it recovers, I go back to normal size. No leverage. No options. Just a slight tilt toward the same businesses I already like.

I used some version of this in every meaningful pullback since 2020.

How the S&P 500 Performed During Major Market Crashes

This last chart is a good reminder of how the biggest crashes eventually get left behind. The people who stayed invested ended up fine. The ones who sold and waited for the all-clear usually didn’t.

Subscribe now

The other chart that always hits me is this one:

The Cost of Missing the 10 Best Days in the Stock Market

Miss just the ten best days over a couple of decades and you cut your returns by 50%. A lot of those best days happen right after the worst ones, when everyone is still hiding. That’s why trying to time your way out of pullbacks usually ends badly.

Pullbacks are normal. A 5% drop happens a few times a year. A proper 10% correction is basically annual. Bigger 20%+ ones show up every few years. The market has always recovered from them.

This drawdown chart is messy on purpose. It shows how often the market falls and how it always eventually climbs back. The people who lose money long-term are usually the ones who sell during the red parts and never fully get back in.

I’m not saying you should buy every falling knife. I’m saying that if you already own high-quality businesses with real competitive advantages and long runways, a pullback is often the best time to buy more. That’s exactly what I did with Palantir in 2023 and with ETH in 2020. The price got better. The long-term story didn’t change.

My portfolio is up around 210% in the last 2 Years. That number didn’t come from calling bottoms or being a genius. It came from staying invested through the noise and adding to the same names when they went on sale.

That’s it.

Want to know the exact names I’m adding during this pullback?
They’re in the article below. ⬇️

Pullbacks will keep coming. They always have. The people who treat them like opportunities instead of reasons to panic are usually the ones who end up with the results everyone else is chasing.

Stay long. Buy the good ones when they get cheaper. Then go do something else with your day.

Cheers!

Subscribe now

Share

Core Thesis

AI growth is no longer limited by chips or models. It is limited by power, specifically reliable, high density, 24/7 power that can be delivered at speed and at scale.

Most of the market is still focused on the obvious winners. The real opportunity sits one layer deeper: the companies that can actually supply the electricity AI needs when and where it needs it. Some of these will be high quality compounders that keep growing earnings for years. Others will be more asymmetric, smaller, less followed names that can re-rate hard once the market properly prices in the power bottleneck.

In my view, the next leg of the AI cycle will be defined less by who builds the best model and more by who can keep the lights on. That shift is already underway, and it creates a clearer, more investable set of opportunities than the broader AI trade.

Constellation Energy (CEG)

Price ~$250 | Market Cap ~$90B

Business
Constellation Energy is the largest producer of carbon-free energy in the United States. It operates the country’s biggest commercial nuclear fleet and, following the Calpine acquisition, also controls a large flexible natural-gas generation portfolio. The combination delivers firm, 24/7, largely carbon-free power, exactly the profile hyperscalers require for AI data centers.

Why critical today
AI data-center electricity demand is projected by the IEA, Goldman Sachs, and grid operators to more than double this decade. Intermittent renewables cannot meet the firm, around-the-clock requirement at scale. Nuclear PPAs with multi-decade duration have become the preferred solution for Microsoft, Meta, and other large load customers. CEG owns the assets and the commercial platform best positioned to capture that demand.

Partnerships

• Microsoft: 20-year PPA for the full ~835 MW of the restarted Crane Clean Energy Center.

• Meta: 20-year agreement for 1,121 MW from the Clinton plant.

• CyrusOne: 380 MW (plus option for another 380 MW) at Freestone in Texas.

• Additional clean-energy contracts with Walmart and a growing pipeline of data-center co-location discussions.

• Calpine integration adds flexible gas capacity that can firm the nuclear baseload.

Contracted Backlog
Long-duration nuclear PPAs already signed convert previously merchant or subsidized generation into premium, contracted cash flow. Management continues to highlight a robust pipeline of additional hyperscaler and large-load opportunities.

Management & Institutions
CEO Joe Dominguez has consistently executed on nuclear restarts, large M&A (Calpine), and commercial structuring. The team has delivered four consecutive years of beating adjusted operating EPS guidance since the company became public. Institutional ownership is broad and high-quality.

Revenue Trajectory & Profitability

• 2023: ~$24.9B

• 2024: $23.6B

• 2025: $25.5B (+8.3% YoY)

• 2026: Strong growth expected with full Calpine contribution (Q1 2026 revenue already $11.1B); adjusted operating EPS guidance $11.00–$12.00

Profitability is robust and expanding. Adjusted operating EPS reached $9.39 in 2025. Management targets base EPS growth from roughly $6.65–$6.75 in 2026 to $11.40–$11.90 by 2029, a compound annual growth rate of more than 20%. Free cash flow before growth is guided at $8.4 billion across 2026–2027 and $11.5–$13.0 billion across 2028–2029.

Balance Sheet
Post-Calpine, total assets stand at approximately $97B. The company maintains investment-grade credit ratings, strong free-cash-flow generation, and an increased $5 billion share-repurchase authorization. Debt rose with the acquisition but is supported by the enlarged cash-flow profile. Market capitalization is ~$90B.

Customer Concentration
The commercial platform serves a broad C&I base (including a large portion of the Fortune 100). Hyperscaler PPAs are growing rapidly but still represent a manageable portion of the overall portfolio, giving CEG better diversification than pure-play smaller AI-energy names.

Challenges, Risks & Execution Realities
Regulatory and permitting timelines for nuclear restarts, successful integration of Calpine, potential changes in nuclear production tax credit economics, and competition for new large-load contracts remain the primary risks. Execution has been strong to date.

CEG is a strong company.

But the four names I’m most focused on for this theme offer what I believe is meaningfully higher asymmetric potential:

Let’s dive in…

For me, this is when I get excited.

Pullbacks like this don’t scare me. They reset prices on ideas I already like and create the cleanest entry points of the cycle. The best opportunities almost always show up when everyone else is scared.

My portfolio, built on the same high-beta asymmetric approach that delivered over 195% growth, has felt the impact. Some positions have given back gains as the AI theme corrected.

Portfolio sizing is key. It’s what lets you survive the drawdowns and then aggressively add when the fear is at its peak.

I was able to secure a clean 10x on SIVE before it collapsed to 33SEK. I bought early and exited into strength, shared with paid subs. How I’m applying those same risk-management lessons right now, and exactly where I’m seeing the strongest risk/reward in the current AI stock pullback.

Performance Overview

- Last 2 years: +195.2%

- Last 1 year: +101.7%

- 2026 YTD: +81.3% vs. SPY +9.6%

No options. No leverage.

Let’s dive in...

Read more

They focus on hype, short-term momentum, or the latest product announcement. What actually creates multibagger outcomes, stocks that can return 10x or more, is far more consistent and identifiable than most people realize.

After studying what worked in previous cycles and applying it to the current AI environment, I’ve identified seven core characteristics that show up repeatedly in the biggest winners. This is the same framework that helped me identify and hold through volatility in three different opportunities — PLTR, NBIS, and SIVE — that each went on to deliver 10x returns.

These traits are even more relevant in 2026 as the AI buildout moves from hype to real infrastructure and enterprise adoption.

Here are the seven characteristics of true multibagger growth stocks in the AI era:

1. Massive and Expanding Total Addressable Market (TAM) Powered by a Secular Trend

The biggest winners almost always operate in markets that are growing rapidly for structural reasons, not cyclical ones.

In the AI era, this means exposure to exploding demand for compute, data infrastructure, enterprise AI platforms, specialized semiconductors, or power solutions for data centers. A company can have great execution, but without a large and expanding TAM, the upside is mathematically capped.

How to evaluate it: Look for markets where spending is projected to grow at 30%+ CAGR for the next 5–7 years, driven by AI adoption. The best opportunities sit at the intersection of multiple tailwinds (for example, AI + data center power demand).

2. Ability to Reinvest Capital at High Returns for a Long Period

This is one of the most important and underappreciated traits.

Multibaggers don’t just grow; they grow while generating high returns on the new capital they deploy. This is what allows compounding to accelerate over many years.

In the AI space, look for companies that can take the cash flow from early wins and reinvest it into new data centers, new models, new verticals, or geographic expansion at similarly high returns.

Red flag: Companies that need to raise large amounts of dilutive capital just to maintain growth usually fail this test.

3. Durable Competitive Moat That Strengthens Over Time

In AI, moats are evolving. Traditional advantages like brand or manufacturing scale still matter, but new ones are becoming more powerful:

• Proprietary data advantages

• Network effects in platforms

• Switching costs in enterprise software

• Technological leadership in specialized chips or infrastructure

• Relationships and integration depth with hyperscalers

The best multibaggers don’t just have a moat — their moat widens as they scale because more data, more customers, and more integrations make them harder to displace.

4. Exceptional Management Execution and Capital Allocation

Great businesses with mediocre management rarely become multibaggers.

In the AI era, execution matters more than ever because the pace of change is extremely fast. You want teams that have shown they can deliver on ambitious roadmaps, adapt quickly, and allocate capital intelligently between growth and returns.

Look for consistent delivery on guidance, transparent communication, and a track record of making smart long-term bets even when the market is impatient.

5. Improving Unit Economics and Operating Leverage as They Scale

Early-stage growth companies often lose money. The multibaggers are the ones whose margins expand significantly as revenue grows.

This is especially important in AI infrastructure and software. Look for gross margin expansion, improving contribution margins, and clear paths to operating leverage. Companies that can grow revenue 50%+ while expanding margins have much higher odds of delivering outsized returns.

6. Reasonable (or Asymmetric) Valuation Relative to the Growth Opportunity

This doesn’t mean buying cheap stocks. It means the valuation is reasonable relative to the quality and duration of growth.

Some of the best multibaggers in AI have traded at high multiples — but the growth and returns on capital justified it. The key is whether the market is underestimating the duration or magnitude of the opportunity.

In 2026, this often shows up in companies that are still early in monetizing new AI capabilities or expanding into new verticals while the market prices them as if growth is slowing.

7. Resilience Through Volatility and the Ability to Compound Over Many Years

The biggest winners almost always go through brutal drawdowns. What separates them is that the business keeps improving even when the stock price doesn’t.

Multibagger candidates should have strong enough balance sheets, customer stickiness, and management conviction to survive temporary setbacks without permanent damage. This trait has become even more important in the AI sector, where sentiment can swing violently.

How I Apply This Framework

When I’m evaluating a new idea, I score it across these seven characteristics. The highest-conviction opportunities tend to score very well on at least five or six of them.

This is not about finding perfect companies. It’s about finding businesses where the combination of these traits creates asymmetric upside, limited downside relative to the potential reward if they execute.

This same lens helped me identify and stay invested through volatility in PLTR, NBIS, and SIVE, three names that each delivered 10x returns. The framework forces you to focus on the business fundamentals that actually drive long-term compounding instead of getting distracted by short-term noise.

Actionable Takeaway

Next time you’re looking at an AI growth stock, run it through these seven questions:

1. Is the TAM large and structurally expanding?

2. Can they reinvest at high returns for years?

3. Is the moat durable and strengthening?

4. Does management have a strong execution track record?

5. Are unit economics improving with scale?

6. Is the valuation reasonable relative to the growth duration?

7. Can the business survive volatility without permanent damage?

The more “yes” answers you get, the higher the probability of a multibagger outcome.

Final Thoughts

The AI era is still early, but the companies that will deliver the biggest returns over the next 5–10 years will almost certainly share most of these seven characteristics. Focusing on them dramatically improves your odds compared to chasing narratives or momentum.

If you want deeper research on companies that score highly on this framework, ongoing analysis of new opportunities, and access to my full portfolio thinking, consider joining the paid tier.

The current low subscription price will not stay this cheap. It is increasing soon. Anyone who subscribes now will be permanently locked in at today’s rate for life. Future price increases will not apply to early subscribers.

You can see the latest details on how this approach is playing out in my most recent portfolio update here:

Thank you for reading. Drop your thoughts on these characteristics in the comments. I read everyone.

Stay focused on the fundamentals that actually matter.

Subscribe now

Here’s why it stands out:

• Revenue expected to surpass $500 million in 2026

• Delivering explosive year-over-year growth

• Holding over $1.4 billion in cash

• Building a complete autonomous defense stack from the ground up

• Goldman Sachs and Morgan Stanley see massive tailwinds in autonomous systems and defense robotics

The name, full valuation breakdown, and price target are below.

Let’s dive in…

Read more

In June 2026, SharonAI closed an oversubscribed $1.6 billion strategic financing. Combined with $2.2 billion in take-or-pay contracts already signed (revenue ramping from Q3/Q4 2026), this marks a genuine inflection point. The company is transitioning from a small pre-revenue operator into a scaled provider of sovereign, high-performance AI compute for enterprises, startups, researchers, and governments across the Asia-Pacific.

Let’s break it down properly: the business, the technology moat via NVIDIA, the demand drivers, execution roadmap, financials, risks, and an institutional-grade valuation framework.

Subscribe now

Table of Contents

• Start of the Business

• What is SharonAI’s business about? (The Neocloud Explained)

• Why specialized GPU cloud / Neoclouds are critical today

• TAM growth and future demand in AI compute

• The NVIDIA Partnership, 40,000 GB300 GPUs and DSX AI Factory

• $2.2 Billion in Contracted Backlog — Real Visibility

• Recent $1.6B Strategic Financing & Capital Position

• Management, Location Advantages & Institutions Buying In

• Challenges, Risks & Execution Realities

• Current Financial Picture

• Valuation model

• My positioning

Start of the Business

SharonAI Holdings Inc. (NASDAQ: SHAZ) emerged from a business combination between Roth CH Acquisition Co. and SharonAI Inc. (closed December 2025), followed by a Nasdaq uplisting in February 2026 that raised approximately $125 million. The company operates as a leading Australian Neocloud, a specialized high-performance computing (HPC) and AI cloud provider focused on accelerated GPU workloads.

It is not trying to be AWS or Azure. It is purpose-built for the most demanding AI training, inference (especially agentic and RAG-style), and sovereign workloads that require dedicated, secure, high-bandwidth clusters with strict data residency and compliance needs.

What is SharonAI’s business about?

Neocloud = Specialized GPU Cloud, Not General-Purpose Hyperscaler

Think of it this way:

Hyperscalers(AWS, Azure, Google Cloud): Offer everything, general compute, storage, networking,, plus GPUs. Great for broad use cases, but GPU capacity is often shared, expensive for sustained large clusters, and subject to allocation queues during AI booms. Data sovereignty can also be an issue for governments and regulated industries in APAC.

Neoclouds (CoreWeave, NBIS, IREN, SharonAI): Pure-play accelerated compute platforms. They focus almost exclusively on NVIDIA GPU clusters optimized for AI/ML/HPC. They offer virtual private clusters with dedicated hardware, high-speed networking (InfiniBand or equivalent), full isolation, and often better price/performance or availability for large, long-running jobs.

SharonAI’s edge is geographic and strategic: Australia.

Australia offers:

- Large land availability for massive AI campuses.

- Growing renewable energy capacity and relatively competitive power in certain regions.

- Strong sovereign AI push from governments and enterprises in Australia, India, Southeast Asia, and parts of East Asia that prefer (or require) local data residency and compute for security, latency, regulatory, or national interest reasons.

- Political stability and rule of law attractive to international customers.

SharonAI delivers Virtual Private Clusters with complete network isolation from the public internet, a big deal for enterprises, research institutions, and regulated industries that cannot or will not put sensitive workloads on shared public cloud GPUs.

Why Neoclouds and high-performance GPU cloud are critical today

The AI buildout has moved far beyond “just more GPUs.” Training frontier models and running large-scale inference (especially agentic AI with long context, tool use, RAG over massive knowledge bases, and continuous reasoning loops) creates enormous demand for dedicated, high-bandwidth, low-latency GPU clusters that can run for weeks or months without interruption.

Hyperscalers are capacity-constrained on the highest-end GPUs during peak demand. Many AI labs, startups, and even large enterprises are turning to specialized providers for:

- Guaranteed access and priority allocation.

- Better economics on sustained workloads.

- Sovereign/private infrastructure.

- Faster deployment timelines in some cases.

Australia/ APAC has been underserved relative to the U.S. and Europe. Sovereign AI initiatives (national or regional control over AI capabilities and data) are accelerating. SharonAI is positioning itself as the go-to partner for this wave in the region.

TAM growth and future demand

The global AI compute / GPU cloud market is expanding at a blistering pace. Demand for high-end accelerated compute continues to outstrip supply in many segments, especially for large contiguous clusters.

Key tailwinds for SharonAI specifically:

- Sovereign AI buildouts across APAC.

- Enterprise and research demand for dedicated, compliant capacity.

- Inference workloads (often more latency- and bandwidth-sensitive than training) favoring specialized providers.

- NVIDIA’s own push to expand its ecosystem of trusted cloud partners (SharonAI is a certified NVIDIA Cloud Partner and was recently highlighted by NVIDIA as an early participant in new AI infrastructure initiatives).

Management and industry commentary point to multi-year visibility of strong demand. The company’s own contracted backlog already reflects this.

The NVIDIA Partnership — A Defining Validation

In mid-2026, SharonAI announced a six-year strategic compute collaboration with NVIDIA

Key elements:

- Deployment of NVIDIA’s DSX AI factory design.

- Scaling up to 40,000 Grace Blackwell GB300 GPUs.

- Targeting 72 MW of new AI data center capacity in Australia.

- NVIDIA is highlighting SharonAI as an early partner in its expanded AI infrastructure initiatives (including revenue-sharing/credit-support models for cloud providers).

NVIDIA GB300 Superchip, the powerhouse at the heart of SharonAI’s planned AI factory expansion.

This is not a vague “partnership.” It is a concrete plan to build one of Australia’s largest AI computing platforms using the latest NVIDIA silicon. The GB300 (Grace Blackwell architecture) delivers massive leaps in performance and efficiency for both training and inference workloads.

For SharonAI, this collaboration provides:

- Access to the most advanced GPUs and reference architectures.

- Credibility and go-to-market support.

- A clear technology roadmap aligned with NVIDIA’s ecosystem.

Subscribe now

$2.2 Billion in Contracted Backlog — Real Visibility (Not Just Hype)

This is one of the most compelling parts of the story.

SharonAI has announced ~$2.2 billion in total contract value (TCV) from two major take-or-pay agreements, both expected to commence revenue contribution in Q3/Q4 2026:

- $1.25 billion over five years with ESDS Software Solutions (Indian counterparty) — for deployment and operation of a large dedicated GPU cluster (~8,200 B300-class GPUs initially referenced in some reporting).

- $950 million over five years with an **unnamed global technology firm** with major Asia-Pacific presence.

These are take-or-pay structures, providing strong revenue visibility once the clusters are online. Current revenue remains small (~$1.5M range in recent periods) because the big deployments are still ramping. The backlog represents the bridge from tiny to material scale.

Recent $1.6B Strategic Financing & Capital Position

On June 17, 2026 (closed ~June 29), SharonAI announced and closed an oversubscribed $1.6 billion private placement financing:

- ~$900 million in Class A ordinary shares and pre-funded warrants.

- $700 million aggregate principal amount of 4.75% Convertible Senior Notes due 2032.

Anchored by:

- Situational Awareness L.P.

- Funds managed by Oaktree Capital Management

- Plus other new and existing institutional and strategic investors.

Proceeds are earmarked primarily for accelerating the NVIDIA AI factory deployment (the 40k GB300 GPUs and supporting infrastructure) and broader expansion across Australia and APAC.

This financing materially de-risks the capex-heavy buildout and gives the company a strong balance sheet to execute.

Management, Location Advantages & Institutions Buying In

Management & Governance:

- Chairman: Former CEO of Telstra (Australia’s largest telecommunications company). This is a significant advantage for navigating local regulations, power/infrastructure partnerships, and enterprise/government relationships in Australia.

- CEO and founding team with deep experience in the sector.

- High insider alignment via voting control structures (common in founder-led growth companies).

Location Moat:

Australia’s combination of land, energy profile, stability, and growing sovereign AI demand is hard to replicate quickly. Building large AI campuses elsewhere faces power constraints, land scarcity, or regulatory hurdles.

Institutions:

The $1.6B raise brought in heavyweight names (Oaktree, Situational Awareness, others). This is not retail-driven speculation; sophisticated capital is backing the execution plan.

Challenges, Risks & Execution Realities

No high-growth AI infrastructure story is without risk. Key ones to monitor:

- Execution on deployment: Building and bringing online 72 MW + 40,000 GPUs on schedule is a major operational lift. Delays in power, construction, or supply chain would push revenue recognition.

- Customer concentration & credit risk: The ESDS contract has drawn scrutiny in some short reports regarding the counterparty’s size and ability to meet large annual payments (letters of credit are in place as protection). The second unnamed contract helps diversify this.

- Dilution: The recent equity component + potential future convert note conversions will increase share count. Valuation must account for this.

- Competition: Hyperscalers expanding capacity, other neoclouds, and potential new entrants.

- Technology & pricing: GPU generations evolve quickly; utilization rates and realized pricing will determine margins.

- Macro: Any broad slowdown in AI capex spending would pressure the entire sector.

SharonAI’s response so far: massive contracted backlog + top-tier capital + NVIDIA partnership + experienced local leadership. Execution is now the main variable.

Current Financial Picture

- Revenue: Still minimal (~$1.5M range recently) as the company was pre-scale.

- Losses: Widened as it invested in platform and prepared for growth (typical for this stage).

- Balance sheet: Significantly strengthened post $1.6B raise. Convertible notes provide long-dated capital with equity optionality for investors.

- Path forward: Revenue ramp expected to accelerate materially from late 2026 as contracted clusters come online and utilization builds. Gross margins should expand with scale; operating leverage follows.

The story is classic high-growth infrastructure: heavy upfront investment and losses, followed by rapid revenue inflection and margin expansion if execution succeeds.

Revenue is expected to rise from 1.6 million in 2025 to 2,431.7 million in 2028, which is roughly a 1,519x increase, while operating profit is expected to swing from losses to 240.4 million by 2028.

Institutional-Grade Valuation Model

What follows is a rigorous, transparent, scenario-based valuation framework built specifically for a high-growth, pre-scale AI infrastructure company. I have modeled SharonAI using a hybrid approach, cross-checked against backlog NPV, EV-per-GPU, and growth-adjusted comps.

The goal is not only to give you a single price target but also to show you exactly how I think about value creation, what moves the needle, and where the asymmetry lies.

Let’s dive in…

Subscribe now

Read more