On 22 August, the Reserve Bank of India published a number that stopped me.

In under eleven weeks, £47.9 billion ($65.4 billion) had flowed into foreign-currency FCNR(B) deposits at Indian banks under a special swap facility the RBI opened on 8 June, per the central bank’s 22 August release covering data to 21 August.

The number needs care. These are bank deposits, relatively liquid, hedged, carrying three-to-five-year tenors. They are not venture commitments and they are not private equity, and setting a bank deposit beside a ten-year commitment to a blind pool would tell us very little.

The number held my attention for a different reason.

India has spent thirty-five years learning how to make capital from Indians abroad easy to receive. The swap facility is only the latest instrument in a long sequence. India raised £1.2 billion through India Development Bonds in 1991, £3.1 billion through Resurgent India Bonds in 1998 and £4.0 billion through India Millennium Deposits in 2000, on the World Bank’s diaspora-bond research, and roughly £25 billion through deposit and bank-borrowing incentives during the 2013 currency pressure, per World Bank Policy Research Working Paper 7071. Now £47.9 billion has arrived in under eleven weeks.

India’s diaspora was wealthy long before any of this. What India built, decade by decade, was an increasingly sophisticated set of ways for that wealth to come home as investment.

That distinction has been on my mind because Africa holds the first half of the equation in abundance. The second half is what’s scarce.

Where does the diaspora’s money already go?

Sub-Saharan Africa received an estimated £41 billion ($56 billion) in remittances in 2024, on the World Bank’s Migration and Development Brief figures.

That number gets used carelessly in investment discussions. Most of it is never going to become venture capital. It pays school fees, medical bills, rent and the running costs of family businesses. Remittances and fund commitments do different jobs, and a manager who confuses them will build the wrong product.

But £41 billion establishes something that matters. A very large financial relationship already exists between Africans abroad and African economies. The money knows how to cross the border. The people sending it already hold an economic stake in the continent.

The open question is what happens after household transfers and direct personal investments have been exhausted.

Where does the African doctor in London put £100,000 if she wants professionally managed exposure to African private companies? Where does the Nigerian technology executive in California allocate £180,000 if he wants West African growth exposure without selecting the companies himself? What does the Ghanaian investment banker in New York do if she wants part of her alternatives book run by an African GP?

For a surprising number of people, there is no good answer. That is an infrastructure problem, in a population with no shortage of wealth.

£41bn Crosses the Border. Very Little Enters Africa Through Private-Capital Funds

Who actually funds African private capital?

Now look at the other side of the market.

Africa-focused private-capital funds raised £2.0 billion ($2.7 billion) across just 16 final closes in 2025, down 34% on the year, in the African Private Capital Association’s 2025 activity report.

More telling than the total is its composition. Development finance institutions supplied 64% of commitments. African investors supplied another 21%.

DFIs remain essential to African private markets. They did the difficult market-building work that commercial institutions would not do, and they still anchor funds that would otherwise never close. But a market in which nearly two-thirds of fund capital comes from development institutions is telling you something about the maturity of its LP base. African fund managers still depend on a narrow set of institutional providers.

There is movement underneath the headline. AVCA counts African investors as one-third of active participants in African venture deals for the second consecutive year. The continent recorded 506 venture deals worth £2.9 billion ($3.9 billion) across equity and debt in 2025, and first-time managers took a record share of the funds that closed. Developing capital markets are supposed to do exactly this. Slowly, they begin financing themselves.

One investor category remains hard to find anywhere in the African LP story: the diaspora.

Which is what makes India worth studying.

What did India look like before it had an LP class?

It is tempting to look at India’s private markets today and assume the ecosystem was always there.

One piece of venture history says otherwise, better than any macro chart.

When Blume Ventures raised its first fund in 2011, venture capital was still unfamiliar to wealthy Indian investors. The fund closed at ₹100 crore, about £7.7 million at today’s rate, raised entirely from domestic high-net-worth individuals and families. Blume’s own retrospective describes a conversion rate of roughly one committed investor for every six families approached. [

What interests me about that story is who had to do the work. There was no formed domestic LP class waiting for Indian venture managers. The managers were helping to create one, family by family, explaining what a venture fund even was before anyone could be asked to write a cheque into it.

Then the ecosystem began doing the persuading on its own. Companies created wealth, founders exited, executives accumulated stock and pattern recognition, family offices professionalised, and managers returned to market carrying track records. The relationship between Indian private wealth and Indian private markets became ordinary.

Blume’s retrospective carries one more line worth sitting with. By its estimate, Indian families captured only 2–3% of the roughly £220 billion ($300 billion) of value the startup ecosystem created over the preceding decade. [VERIFY: same Blume source] India was creating vast private-company value before Indian private wealth had learned how to participate in it.

That should sound familiar to anyone reading this from Lagos, Nairobi, Cape Town or London.

What did India actually build?

Everything up to this point is the problem: Africa has the wealth and not the machinery. What India built is the machinery — the diaspora instruments that turned three crises into capital, the NRI money now sitting inside regulated alternative funds, the ladder a professional climbs from a first small cheque to a full commitment, and what all of it means for the funds being raised on the continent right now.

That part is for paying subscribers.

ScaleSignals is the capital-route read written from inside the rooms where the money actually gets committed. One subscription, every briefing, and the data behind each one.

Subscribe now

Read more

The data says the assumption is broken at the top.

DocSend has spent years studying how founders raise and how investors actually behave with fundraising materials, combining founder surveys with proprietary data on how decks get read. Its seed-stage research found only a weak relationship between the number of investors a founder contacted and the number of meetings secured. The relationship between investors contacted and capital actually raised was weaker still. DocSend’s own conclusion was that reaching the right investors, rather than simply more of them, is what lets founders raise effectively. For an industry that has turned fundraising into funnel management, that is an uncomfortable finding, because it means the biggest number at the top of the funnel may be one of the least useful.

The 2022 market made the contradiction visible in a single year. Founders sent 30% more pitch decks than the year before and took 33% more investor meetings. Investor engagement with those decks fell 23%.

More fundraising activity did not produce more investor engagement

More activity, less attention. That does not mean outreach is pointless; nobody raises from an investor who has never heard of them. It means activity and progress are different variables, and the difference is expensive, because the people doing most of the activity are the people the company can least afford to lose: its founders. Every hour spent researching the wrong investor, chasing the wrong introduction and preparing for the wrong meeting is not merely unsuccessful fundraising. It is management time transferred out of the company into a process that never had a viable endpoint.

Which points at the distinction this whole piece turns on. A meeting with an investor who cannot invest, because of geography, stage, cheque size, sector, instrument, fund status, ownership requirements or a portfolio conflict, is still a meeting. It is just not a capital opportunity. A founder looks at forty investor meetings and sees a pipeline. The better question is how many of those forty were institutionally capable of writing the cheque at all.

Because a rejection and a mismatch are not the same thing. A rejection says: we could invest, and we chose not to. A mismatch says: this opportunity was never inside the investable set. Founders experience both as a no. Operationally they are completely different events, and the data suggests a large share of the second kind is being manufactured before the first email is ever sent.

The funnel nobody shows you sits in front of the one everybody manages. Here is what it looks like, and what it changes.

The funnel before the funnel

Read more

The Raise Readiness Sprint — my half-day workshop — runs its next cohort on Friday 4 September. You score your own company on the evidence investors assemble before you pitch, map the gaps, and leave with a 30-day plan to close them. £295, capped at 25 → maven.com/benbotes/raise-readiness-sprint

—

Every startup now has access to the same models, the same …

Read more

---

On 9 July, a Paris fintech called Aria announced rough…

Read more

Run a thought experiment with me.

Take the four people who gave the startup world its operating system. Don Valentine, whose conviction at Sequoia that the market matters more than the team became the physics of venture capital. Andy Rachleff, who studied that conviction at Benchmark and gave it a name. Alexander Osterwalder, who put an entire business on one page. Ash Maurya, who rebuilt that page for founders working at speed.

Now move them to 2026 and hand them the same brief: give builders a tool for knowing whether they are winning.

None of them would build what they built.

Not because the originals were wrong. Because each of those tools was a precise answer to the scarcities of its own era, and every one of those scarcities has since inverted.

What the tools were actually solving

The history takes one paragraph, and it earns its place only because it reveals the design constraints. In 2007, when Marc Andreessen published the essay that carried Rachleff’s term into every accelerator on earth, building software was slow and expensive. A funded competitor needed a year to ship an alternative. A customer encountered a handful of new options a quarter. Under those constraints, the fatal unknown was singular: does the market exist at all? Product-market fit named that unknown and gave founders a target. The canvases that followed in 2008 and 2010 solved the era’s second problem, the opacity of the business model itself, by making the whole machine visible on a single page. A map you could fill in, in a workshop, once.

Notice the shared design assumption underneath all three tools. Each produces a completed artifact: a claimed state, a filled canvas. Each lets you finish. And finishing was rational, because the world you had described held still long enough for the description to stay true. The map and the territory stayed aligned for years at a time.

The inversion

Every constraint that made those tools correct has flipped.

Building is no longer the hard part. Products that took a funded team a year now get assembled in weeks, by teams that would not have qualified as teams five years ago. “Can you satisfy the market” has stopped being a filter, because nearly everyone who tries can satisfy the market within a quarter. The question that once separated the living from the dead is now table stakes.

Copying has collapsed with it. The distance between a novel capability and a commodity capability used to be measured in years. It is now measured in release cycles. Whatever you demonstrate to win your position, a competitor can demonstrate before your next board meeting.

And the customer has been repriced. The person you serve is shown alternatives at a volume no buyer of 2007 experienced, with switching cheaper to consider and more aggressively invited than at any point in commercial history. Your customer is not sitting where you found them. They are being recruited away from you daily, by products that did not exist when you earned their signature.

Put simply: the frameworks did not decay. The shelf life of their output did. A completed snapshot of your business was durable intelligence in 2007. In 2026 it is a photograph of a moving object, aging from the moment it is taken.

What they would build instead

So return to the room. Valentine, Rachleff, Osterwalder, Maurya, 2026, same brief. What do first principles hand them?

Their originals were maps, and maps are the correct tool when the territory holds still. Today the territory redraws itself between board meetings, and the tool that matches a moving territory is an instrument. A map tells you where you are. An instrument tells you what is happening to you, continuously, while it happens.

Which means the 2026 designers would refuse to name a state. States expire too fast to carry information. They would name a motion, because in a market that never holds still, only rates of change tell you the truth. Rachleff’s 2026 question would no longer be whether you are in a good market with a product that satisfies it. Nearly everyone clears that bar now. His question would be whether the market is doing more of your work this quarter than it did last quarter, and whether your balance sheet outlives the answer. Osterwalder would hand you an instrument panel to read weekly, and would consider a canvas completed once and framed on the wall to be a record of the past.

The central instrument on that panel measures one thing.

Product-market pull

Product-market pull is the degree to which the market carries the product forward faster than the company pushes it.

A motion, running right now or not running, and readable from both sides of the table. I will describe it from both, because both is where I sit.

From the investor’s chair, pull is visible before the founder says a word. The diligence is half done before the first call, because demand has already arrived in my inbox from people I trust who were not asked to send it. Customers are recruiting customers. Users show the product to colleagues with no referral incentive in sight. Renewal happens without a save motion. The pipeline holds names the company has never contacted.

From the founder’s chair, the difference declares itself in where the effort goes. A company with mere fit satisfies the customers it has, while the market stays indifferent to the customers it lacks; sales carries the growth, marketing carries the awareness, the founder carries the conviction. A company with pull finds the market performing some of each of those jobs on its behalf, and the share the market performs is rising.

That last mechanism is why pull matters now in a way it could not have mattered in 2007. Pull is the only market position that renews itself at the speed the market moves. Every advocate re-earns your position daily, in conversations you are not in, against alternatives you have not seen. Fit defends nothing while you sleep. Pull is a workforce that never clocks off.

The three forces it has to outrun

Pull is a motion, and motions face resistance. Three forces, and naming them is how you manage them.

The customer force. The people you serve are being changed continuously, by everyone else selling to them and by the tools reshaping their work. Serve the customer you signed and you drift from the customer they are becoming. Pull survives this force through contact alone: the unglamorous, continuous customer conversation that companies so often retire the quarter after the fit certificate arrives.

The capital force. A market can be pulling your product on a timeline your balance sheet does not survive. Financial-market fit, whether your capital structure outlives your market’s clock, sits underneath every claim in this essay. Pull shortens the timeline capital must survive, because carried growth is cheaper than pushed growth. It does not remove the question, and a founder mid-raise should answer it before answering anything else.

The category force. AI compresses the window between novel and commodity, and no volume of customer affection protects a feature the whole category now ships. Pull built on a capability decays with the capability. Pull built on accumulated context, the data a customer gave you, the workflow they built around you, the trust they extended to you, decays far more slowly, because context is the one thing a rival cannot ship in a release cycle.

That is the argument, and the argument is free. What follows is the instrument: the loop that outruns the three forces, the honest places the framework breaks, and the gauge that shows which direction your company is already moving. The paid half is the part you will use on Monday morning.

Read more

The Raise Readiness Sprint — my half-day workshop — runs its next cohort on 4 September. You score your own company on the exact evidence investors assemble before you pitch, map the gaps, and leave with a 30-day plan to close them. £295, capped at 25 → maven.com/benbotes/raise-readiness-sprint

Read more

The most consequential rooms in capital are the ones the money never sees. Every institution that allocates at scale, the credit committee, the investment committee, the board, does its real deciding behind a closed door, in a meeting the applicant does not attend, under rules the applicant was never shown. Venture capital compresses that whole apparatus into a few hours on a Monday morning, and it is remarkable how little of what founders are taught about fundraising describes what happens inside it.

The last piece in this series ended at that door. The first meeting, it argued, is not where a fund evaluates you; it is where you recruit the partner who will argue for you in a room you will never enter. This piece is about the room.

Start with what the numbers say about its weight. In the Gompers, Gornall, Kaplan and Strebulaev survey of 885 venture investors, for every hundred companies a firm considers, roughly ten reach the partner meeting and about one gets funded. But the number that matters more is the conditional one. First Round, one of the few firms to publish an insider’s account of its own process, notes that firms commonly issue term sheets to somewhere between a quarter and 60% of the companies that reach a partner meeting. Hold those two figures side by side. Overall odds: about one in a hundred. Odds once you are in the room: as high as one in two. Almost the entire distance between rejection and a term sheet is crossed in a single meeting you do not attend.

What founders imagine happens there is a presentation. What actually happens is closer to a trial with local rules, and the rules differ more between firms than almost anyone outside the industry realises. In some partnerships one voice can kill your deal. In others one voice is all your deal needs. The same company, with the same numbers and the same champion, clears one firm’s Monday meeting and dies in another’s, and the difference was never the company.

Those rules are not published. But they are knowable, and they change what a founder should do in every meeting that comes before the one that counts.

Read more

Some of the most consequential capital decisions in the world are shaped in meetings that last half an hour. Trillions of dollars of eventual enterprise value route through first conversations between a founder and a partner, and the decisions those conversations produce are made faster, on less information, and by a different logic than almost anyone on the founder’s side of the table believes. The market imagines a deliberate evaluation. What actually happens is closer to a recruitment.

The numbers set the stage. When Gompers, Gornall, Kaplan and Strebulaev surveyed 885 institutional venture investors for the most comprehensive study of VC decision-making ever run, they found that of all the opportunities a firm considers, roughly one in a hundred becomes an investment. For every hundred companies that cross a partner’s desk, around twenty-eight get a meeting with management, roughly ten make it to the partner meeting, about five enter due diligence, and one gets funded. Read those numbers as a founder and one thing jumps out: the steepest cliff is not at diligence, where the spreadsheets come out. It is in the stretch between the first meeting and the partner meeting, where more than half of the companies that earned a conversation disappear.

Which means the first meeting is doing something other than what founders think it is doing. It is not primarily an information exchange; the partner has usually seen the deck, and increasingly they have seen your data before you knew the firm existed. It is not the decision either; almost no fund can say yes in the room, because the yes belongs to a partnership, not a partner. What the first meeting actually decides is narrower and more human: whether the person across the table is willing to spend their own credibility carrying your company into a room you will never enter.

That is the piece founders consistently miss. You are not pitching the fund. You are auditioning your advocate. Every serious firm runs some version of the Monday partnership meeting, and your company will live or die in that room, retold by someone else, in your absence, in about three minutes. The first meeting is where the partner decides whether they can make that retelling work, and whether they want their name on it.

Everything a good partner does in those thirty minutes, the questions that feel oddly specific, the interruptions, the places they lean in, is in service of that test. Here is what they are actually listening for.

Read more

A venture investor spends about four minutes on the average pitch deck, according to DocSend. Founders spend months building the thing that gets four minutes, and almost none on the signals that draw the real scrutiny — because those signals are not in the deck at all. They sit in the behaviour of a company’s customers, the efficiency of its spending, a…

Read more

The most consequential decisions in venture capital are made before a founder ever walks into the room, and they are not about any particular founder. They are decisions about fund size, ownership targets and how many companies a partnership can afford to back, taken months or years earlier, and together they quietly decide which kinds of companies can get funded at all. A fund’s structure is a filter set in advance, and much of what looks like taste in a pitch meeting is that filter doing its work. The shape of which companies get built is settled upstream of the pitch, by arithmetic the founder never sees.

Here is that arithmetic in the form a founder actually experiences. When you pitch, you are answering one question: is this a good company. Inside the fund, the partner is answering a second one: can this company make our fund math work. You are being evaluated on the first and funded on the second, and the distance between them is where most good companies quietly lose.

The gap is not about effort or belief. It is structural, and it comes from the one feature of venture capital that governs everything else. Returns do not spread evenly across a portfolio. They concentrate, brutally, in a handful of companies. The Horsley Bridge data that Sebastian Mallaby popularised found that 6% of deals generated 60% of all returns across three decades, and Correlation Ventures, looking at more than twenty thousand investments, found that around 65% of venture deals return less than the money put in. A fund does not survive on a spread of respectable outcomes. It survives on one or two companies that become very large, and everything in its construction is bent toward finding them.

That single fact rewrites the question in the room. A partner is not really asking whether your company is good. Good is necessary and nowhere near sufficient. They are asking whether your company could plausibly become one of the rare outcomes that carries the entire fund. And most companies, including most good ones, cannot, through no fault of their own.

Read more

More than two hundred venture funds now pay for software whose only job is to find founders before they show up in any database. The tools have names — Evertrace, Harmonic, a handful of others — and they spend their day crawling trade registries, GitHub commits, patent filings, research grants and freshly registered domains for the first faint signs of …

Read more

For most of the past decade, the most valuable thing a company could build was a faster way to move money, attention or goods across a screen. Capital organised itself around that fact. The transaction layer — payments, lending rails, marketplaces, the software that sits between a buyer and a seller — absorbed the equity, set the valuations, and trained a generation of investors to recognise opportunity by how close it sat to a transaction.

That era is now meeting its limit, and the clearest evidence is in where the smart money is going next. In developed markets, AI has absorbed the top of the equity stack — roughly 63.5% of all US venture equity value in 2025. Underneath it, lenders are doing something quieter and more telling: rotating out of software and toward the physical economy. Pinegrove Credit Partners, spun out of SVB, now puts close to half its capital into defence, aerospace and hardware, with software lending down to around a fifth of deals from a historical 40–50%. The capital is migrating to the parts of the economy where the revenue is physical — a satellite, a grid, a manufacturing line, a food system. After a decade of funding the rails that move value, capital is relearning that the things being moved have to exist first, and that most of them are underbuilt.

Emerging markets are where that lesson is sharpest. The same structure now becoming visible in the West has governed frontier capital for years, and it shows up in a single, repeatable pattern: capital funds the transaction rails before it funds the real-economy rails. Fintech absorbs the dollars. Food systems, healthcare delivery, energy access and physical infrastructure absorb the unmet demand. The money and the opportunity sit in different places, and the sectors carrying the largest demand against the smallest capital are the most underfunded opportunities in the emerging-market landscape entering 2026. This briefing names them, ranks them, and shows where the gap is widest.

The map itself uses the two axes that matter most when capital is scarce and mispriced. The horizontal axis is capital intensity: how much money is currently flowing into a sector. The vertical axis is opportunity size: how large and how fast the underlying market is growing. Cross them and four quadrants appear. Three of them hold most of the capital deployed across emerging markets today. The fourth holds the sectors most underfunded relative to their market — where the returns of the next decade are most likely to come from, and the one quadrant almost nobody is mapping.

Where has the capital already gone?

Start with what is over-funded, because it frames everything else. In every emerging-market region, financial services absorb a share of venture capital out of all proportion to the rest of the economy. In Africa, fintech took roughly $1.49 billion in 2025, about 37% of all tech funding (Partech Africa). In MENA, it absorbed around 58% of reported startup funding (MAGNiTT). In Latin America, fintech attracted close to 61% of all venture dollars in 2024 (LAVCA). In Southeast Asia, financial services have run between 45% and 74% of capital in recent periods (Cento Ventures).

None of this is a case against fintech. The markets are real. Billions of people remain underbanked, and digital financial infrastructure genuinely enables much of what gets built on top of it. But a sector that takes between a third and two-thirds of all available capital in every region is, by definition, the crowded quadrant. Valuations sit high, winner density is high, and distribution is contested. A generalist fintech founder raising in 2026 is entering the most capital-saturated category in the emerging-market landscape. The opportunity is large, and so is everyone else who already found it. E-commerce and broad cleantech occupy the same quadrant for the same reason: large, real markets that capital has already arrived at. African cleantech alone reached $1.18 billion in 2025, up 99% year on year (Partech). The money is there, and it is competing with itself.

Which sectors are most underfunded?

Now the quadrant that matters. These are the sectors where the underlying market is large and growing but capital stays thin — underfunded not because the demand is unproven, but because capital has not yet arrived. The distance between the size of the market and the size of the cheque is the entire point. It is where price has not caught up to value, and the only condition under which an early position can be underwritten cheaply.

Four sectors lead the list, and the gap in each is stark enough to state in a single line. In agrifood, African agritech and food processing raised roughly $116 million across 39 deals in 2025, against an African food and agribusiness market the African Development Bank projects will reach $1 trillion by 2030. A trillion-dollar market is drawing nine figures of venture capital. The same shape repeats in Latin America, which produces around 14% of global agricultural value while its agritech sector raised roughly $119 million in 2024, and in MENA, which imports about half its food while foodtech drew only tens of millions.

In healthcare delivery, African healthtech raised $224 million in 2025, its record year, up 232%, against an African Development Bank estimate of roughly $26 billion in health-infrastructure investment needed annually. The strongest funding year on record still leaves a gap measured in orders of magnitude. In energy access, the aggregate cleantech figure conceals the shortfall: grid connection, last-mile and off-grid access — the sub-sectors that actually reach the nearly 600 million Africans without electricity — need around $15 billion a year to deliver universal access, while access-specific funding sits far below that, hidden inside the broad cleantech total. And in connectivity, African networks drew about $101 million across nine deals in 2025, against an estimated $100 billion requirement to reach universal affordable broadband by 2030. The applications keep getting funded ahead of the networks they depend on.

Those four — agrifood, healthcare delivery, energy access and connectivity — are the most underfunded sectors visible on the map. The demand in each is measurable, institutional, and in most cases growing. What is missing is the capital, and that absence is the opportunity. The question is which side of that line you are building or allocating on.

Read more

Why Emerging-Market Fintech Runs on Physical Trust, Not Just Software

The market prices these companies as software. The thing that makes them work, and makes them hard to displace, is a physical system that turns trust into transactions. Here is how to read the trust layer, as a founder, GP or LP.

The prevailing belief in finance is that money is becoming pure software. Wallets, rails, stablecoins, banking APIs, neobanks: the story is that the physical apparatus of finance is dissolving into code, and trillions in market value are being allocated on that conviction. The belief is mostly right in the markets that wrote it. It is half wrong in the markets that will add the next billion users.

The half it misses is the one that decides whether any of it works. Somebody, somewhere, has to make a person believe that a number on a screen is really their money, and will still be their money tomorrow. In markets where banks have spent decades earning distrust, that belief is not produced by a better interface. It is produced in the physical world: by an agent who hands over cash, a shopkeeper who vouches for the product, a field officer who fixes the failed transaction. Call it the trust layer. It is the part of an emerging-market fintech that does not appear in the pitch, and it is often the part that makes the company.

The visible company is digital. The deck talks about transaction volume, active users, merchant acquisition, platform expansion. The load-bearing company is frequently physical, and it sits underneath all of that: agents, merchants, cash-in and cash-out points, repayment behaviour, the everyday places where people already move money. Some emerging-market fintechs scale faster than they should on paper because they are not only building software. They are building on top of trust that already exists, or they are building the physical trust layer themselves.

Mobile money is the clean example. The GSMA reports that merchant payments through mobile money grew by almost half in 2025, to $155 billion, the fastest-growing use case in an ecosystem that now moves more than $2 trillion a year. Read from the screen, that is a digital-finance story. But mobile money has always run on an agent and distribution network that lets people convert between cash, balances, merchants and trust. The app did not remove the physical layer; it made the physical layer more valuable.

Which sets up the question the market keeps underwriting too lightly. The transaction is digital. The trust, in these markets, is not. That gap is where the value hides, and where the mispricing happens.

Read more

For a long time I filed development finance in the side column. Venture capital was the market; development capital was the well-meaning layer around it. That now reads as too simple.

In many emerging markets, development finance does not sit beside venture capital. It helps make the market legible enough for venture capital to enter — giving commercial investors a way to read risk, back first-time managers, and justify capital moving into places where they would otherwise hesitate. That does not make the capital soft or charitable. It makes it structured. And once you see the structure, a funding round starts to look different.

The visible story is clean. A startup raises, a venture firm leads, a press release announces the round, and the market reads it as private capital moving into an emerging market. That is usually the story the founder wants — it shows momentum, confers legitimacy, and tells future investors, customers and talent that the company has crossed a threshold. Often it is true. It is rarely the whole structure.

In March 2026, the South African fintech littlefish announced a $9.5 million Series A led by Partech, with TLcom, Flourish Ventures and Proparco participating. On the surface, a familiar venture story: a company building merchant infrastructure for Africa’s banks, with real traction, raising from serious investors.

Then the next layer appears. Proparco, the French development finance institution, disclosed that its participation was made under the EU-backed Choose Africa VC programme. The cheque the market saw was venture capital. One of the cheques inside it was development finance, carrying a mandate.

That is the point of this piece. The visible cheque is only one layer of the capital. The rest is structure — and structure always has a job.

Read more

A founder sat across from me last quarter and walked me through her growth numbers like she was handing me something I had never seen. I had seen most of it before she opened the laptop. We had never met, never spoken; her company had simply been throwing off signal for eighteen months, and the tools a fund can buy had been reading it the whole time.

What I want to describe is not a scandal and not a secret. It is the ordinary mechanics of raising money now, and it is hardest on the founders who still believe the meeting is where the fund learns who they are.

The story you are told about fundraising is a story about persuasion. Build traction, sharpen the deck, get the meeting, and in the room reveal what you have built well enough that someone commits. The model hands the founder control of the information — you decide what to show, and when. The pitch is the moment of disclosure.

That model is a decade out of date.

A layer of software now tracks private companies in real time, long before they raise, and in some cases before they exist. SignalFire has run an internal platform called Beacon since 2013; by its own account it tracks more than 80 million companies and 600 million people across 40 datasets, and it does not wait for founders to come to it — it tells the firm’s partners which founders to contact. EQT built Motherbrain to run the same play across Europe. A product called Evertrace advertises that it finds founders before they appear in any startup database, reading domain registrations before there is a company to register.

So before you send a deck, a fund can hold a file on you: a continuously updated picture of how you hire, how you grow and who your founders are, scored against a few hundred companies that resemble yours. The file is not a reference call or a rumour. It exists whether or not you have spoken to anyone — and it is usually richer than the deck you are about to send.

The fund already has your numbers. The rest of this piece is what it does with them — and how to walk in and overturn the score. Founding-member access to The Forensic Read closes 12 July.

Read more

How Palantir Built Software Institutions Couldn’t Easily Remove

Palantir’s real advantage was not complexity. It was becoming part of how institutions make decisions.

Most people explain Palantir through the things they can argue about: government contracts, intelligence work, controversial deployments, Alex Karp, Peter Thiel, defence budgets and, more recently, artificial intelligence. That version is not wrong. It is just too narrow to be useful for anyone trying to understand how software becomes difficult to remove from a serious institution.

The more useful question is not whether Palantir built complex software. It clearly did. The more useful question is how the company built software that could enter environments where data, decisions, permissions, security, workflows and institutional accountability were already tangled together. In those environments, software does not win because it looks elegant in a demo. It wins when it becomes useful inside the mess the customer cannot avoid.

Viewed through that lens, Palantir begins to look less like a conventional software company and more like an institutional embedment company. Its strongest position is not simply that customers use the software, or that the software connects data sources, or that the company has large public and private sector clients. Its strongest position appears when the platform moves closer to the point where information becomes judgment, judgment becomes coordination, and coordination becomes action. The product matters, but the operating position around the product matters more.

That distinction matters because most founders still misunderstand enterprise defensibility. They think stickiness comes from contract length, procurement friction, integration pain or feature depth. Those things can matter, but they are not the same as removal resistance. A product is easier to replace when it is only used for a task. It becomes harder to replace when the customer starts depending on it to understand what is happening, decide what to do next, and coordinate work across teams that do not naturally operate from the same view of reality.

Palantir’s real advantage was not one platform, one contract, one customer segment or one founder. It was institutional embedment. The company entered hard environments, learned from difficult deployments, and built itself into workflows where the cost of removal was not only technical. It was operational. Data became workflow, workflow became decision support, decision support created trust, trust increased switching cost, and switching cost created room for expansion. That is the loop competitors could describe but could not simply copy.

Why this matters

A founder raising capital does not need to copy Palantir. A GP assessing a company does not need to become an expert in government procurement. An LP judging a manager does not need another heroic software story. The useful lesson is more transferable: durable advantage often appears when software moves from being a product the customer buys to being part of how the customer operates.

Founders lose time when they mistake enterprise complexity for enterprise defensibility. Investors lose money when they mistake long sales cycles for deep customer commitment. Managers lose credibility when they describe portfolio companies as mission critical without showing where the product sits inside the customer’s real operating system. Palantir is a valuable case because it shows how difficult customers, deployment knowledge, trust and workflow integration can reinforce each other over time.

If you only have three minutes, here is the point. Palantir did not only build data software. It built an institutional embedment loop. Hard customers created pressure to solve problems that ordinary software could not absorb. Deep workflow integration moved the product closer to the decision layer. Difficult deployments created operating knowledge. Operating knowledge created trust. Trust made replacement harder. Once the product sat near the decision layer, adjacent workflows became easier to reach.

That does not mean Palantir is impossible to remove. No company is. Palantir’s own filings say many customer contracts contain termination-for-convenience provisions, which is a useful reminder that contractual language alone should never be confused with true customer dependency. [S13] The point is not contractual lock-in. The point is that software becomes more defensible when the customer’s practical cost of removal rises because the product has become part of how work is understood, governed and acted upon.

For founders, the useful question is not whether the product sounds enterprise-grade, but where it actually sits in the customer’s operating reality. For investors, the useful question is not only whether the company has large customers, but whether each deployment makes the company more trusted, more embedded, more knowledgeable and harder to replace. The public version of the Palantir story is easy to find: government contracts, data analytics, controversy, AI, Alex Karp and Peter Thiel. The useful part sits underneath that story, in the operating pattern that turns difficult deployments into institutional learning.

For paid subscribers, the rest of this deep dive breaks down that build pattern: how hard customers created harder proof, how workflow integration became more important than adoption, how deployment knowledge compounded quietly, how trust changed the retention equation, and how founders, GPs and LPs can use the same lens to judge whether enterprise software is merely being used or becoming embedded. The lesson is not to copy Palantir. The lesson is to understand how software becomes part of institutional decision-making.

Read more

Most people explain SpaceX through the things they can see: reusable rockets, dramatic booster landings, NASA contracts, Starlink satellites and, inevitably, Elon Musk. That version is not wrong. It is just too shallow to be useful for anyone trying to make a better capital deci…

Read more

A founder walked into a conversation with me last week with a plan she’d built nine months earlier. Vertical B2B software, $4M of ARR, growing 60% year-on-year, gross margins north of 80%. The plan was a clean Series A: $10M at a $40M post, twenty points, one board seat, back to building. The same shape of round she’d watched her peers close eighteen months ago. She wanted to know how aggressive to be on the post.

I had to tell her the round she was describing doesn’t really exist for a company like hers anymore. Not at that size, not at that valuation, not at the conversion rate she was modelling. Not because anything is wrong with her business — by every measure I care about it’s a better business than it was a year ago — but because the stack underneath the question has moved.

If you read the December companion to this piece, you’ll know the world she’d planned in. Equity owned the stack at the end of 2025. Debt was the backstop, blended capital was somebody else’s category, and the whole literacy founders had built was around dilution — how much, at what price, to whom. That December read was a look-back; I wrote it in 2026 to put the before on the record while it was still clearly in view. This is the after. Six months on, the stack got rebuilt around her, and around almost every founder who isn’t sitting inside the AI fraction. Here is what it now looks like — and, more practically, what to do with the new picture.

What’s changed since December?

The default flipped. That’s the GP read in one line, and it’s the line everything else in this piece flows from.

A year ago you raised priced equity by reflex, and the only question worth arguing was the percentage you sold. Today, equity is concentrating into a sliver of the market — the AI names — and for the rest of the market the working stack is no longer equity with debt as a backstop. It’s debt, revenue-based and blended capital sitting where equity used to sit, with equity as the enabler around them. The verb has shifted. Founders who negotiated dilution a year ago are negotiating structure now. Which kind of capital, in what order, against what.

The data supports the read squarely. US venture debt hit a record $68.8 billion in 2025, the highest year ever, and Q1 2026 set a decade high for late-stage venture debt deals — median deal $10.8 million, average $68.2 million, per the PitchBook-NVCA Venture Monitor. Growth-stage startups captured 67% of all US venture debt dollars in that single quarter, roughly $13.3 billion. On the equity side, AI absorbed about 63.5% of all VC equity deal value in 2025, and in Q1 2026, 51.7% of every megadeal involved an AI company. The headline venture numbers are at all-time highs precisely because a handful of names are absorbing them — strip out the top five deals from Q1 2026 and total deal value falls more than 70%.

So the part of the stack the December map showed at the top — priced equity — didn’t disappear. It narrowed. It became the route for a particular kind of company (large model labs, AI infrastructure, GPU-scale compute) and stopped being the default container for the rest of the market. Everything else in the stack absorbed the slack. Venture debt rose hard. Revenue-based finance kept climbing. Blended and DFI capital moved from peripheral to load-bearing for the kinds of businesses commercial capital won’t touch alone. The June map you’ll have seen alongside this piece reflects exactly that — priced equity dropped a band, and venture debt, blended and revenue-based each gained one.

Andrew Ziperski named the shape of it on X in April, writing from the operator-analyst seat. His read: the equity markets have effectively closed for all but a very small group of leaders, and the way through from here for everyone else turns on the willingness to think creatively about how the capital is structured. Or as he put it himself: “Finance is now a first-class citizen in Silicon Valley.” That sentence is the macro of this entire piece.

In the more granular view, Shaun at Third Sphere put the mechanism on the record this week:

He names what the map shows — companies borrowing against revenue, offtakes and assets, with private credit thriving inside tech precisely because the old equity machine has stopped delivering for the median company outside the AI cluster. That’s the shift the data describes, told from a different seat than mine.

But why are lenders rotating away from software?

Here is the second-order line, and it’s the one you won’t see anywhere except from a GP watching this happen in real time on cheques.

Venture debt is rising — but inside the rise, the shape of what lenders are willing to underwrite has moved hard. The lenders themselves are rotating away from software and toward hardware, defense, aerospace, energy and the physical economy. Pinegrove Credit Partners — the venture lending platform spun out of SVB and now backed by Brookfield — today puts roughly 45% of its capital into defense, aerospace and hardware. Their software lending, historically 40–50% of deals, is down to around 20%. Cut roughly in half. They are not alone. Marc Lipschultz, co-CEO of Blue Owl Capital, said it cleanly to the Journal: “There are certain parts of software that are vulnerable.”

The fear underneath the rotation is obvious once you stop to look at it. Venture debt is mechanically a bet on the recurring revenue of the business you lend to. Software’s whole appeal as a lending asset was the cleanness of that recurring revenue — SaaS contracts you could underwrite against, churn curves you could model, retention you could trust. AI is right now the credible threat to that recurring revenue across an unknowable number of software categories. So the lenders who priced loans against SaaS multiples three years ago are pulling back from the exact asset class they used to favour, and reaching toward things AI cannot disrupt as easily — a satellite, a drone, a battery, a fuel cell, a defence platform, a manufacturing line. The debt is migrating to the parts of the economy where the revenue is physical.

If you’re a software founder reading this, that is the part you need to know that the broad numbers will not tell you. “Venture debt is up” hides the fact that, inside the category, your specific door has narrowed. The lender will still see you — but the questions they ask have changed, and the terms they’ll write against your business are tighter than they would have been at the end of last year.

The rest of this piece is for paid subscribers — the part that turns the map into a decision.

So which stack actually fits your model and your market now?

Read more

I’ve been thinking about why some founders close their rounds quickly and others spend six months on a pipeline that never quite moves. (You’d think it would be the obvious things — better company, better deck, more meetings. It mostly isn’t.)

Here’s what I’ve started noticing from my side of the table. The number that tells me whether a raise is real is…

Read more

The figure is almost never the story. What a round actually tells you is buried one level down, in how it was built — which instruments, which investors, which risk sat with whom. Two companies can announce the same headline number and have done two completely different things with it. A founder who copies the number without reading the structure copies the wrong thing, and pays for it with a quarter they will not get back.

May produced an unusually clean set of teaching cases. Four rounds, four structures, and between them almost the entire grammar of how private capital is priced right now — from a strategic land-grab in Silicon Valley to a Shariah-compliant debt facility in West Africa. I have sat on both sides of these conversations: the GP deciding what a structure signals, and the founder trying to read what an investor actually means. From that seat, here is what these four rounds are really saying.

When the cap table is the product

Start with the round everyone has already read about. Hark, a roughly twelve-month-old AI lab, raised over $700 million in a Series A at a $6 billion post-money valuation, led by Parkway Venture Capital. The participant list is the part worth your attention: Nvidia, AMD Ventures, Intel Capital, Qualcomm Ventures, ARK Invest, Brookfield, Greycroft, Salesforce Ventures and more.

Read that list again, slowly, because it contains a thing that does not normally happen. Nvidia, AMD, Intel and Qualcomm are direct competitors. They do not share cap tables as a rule, because backing the same horse undercuts the whole point of a proprietary bet. Four of them on one Series A is not four investors who independently liked the company. It is the chip industry collectively buying an option on what might become the next hardware interface — and none of them wanting to be the one who sat it out. The structure is a hedge dressed as a financing.

This is what I mean by reading a round from the inside. A founder looking at Hark sees the lesson “raise enormous money pre-product if your vision is big enough.” That lesson is false, and acting on it is expensive. The truth is that this round was not priced on Hark’s fundamentals — there is barely a product to price. It was priced on a serial founder who had already built two companies worth more than a billion dollars each, who put in $100 million of his own capital first, and on a strategic land-grab among chipmakers who can each afford to be wrong. Strip any one of those three conditions away and the round does not exist.

Strategic capital prices differently from financial capital. A financial investor asks what the company is worth. A strategic investor asks what it is worth to them to have a seat — to secure supply, to deny a rival, to learn. Those are not the same question, and the cheque that follows from the second one tells you almost nothing transferable about your own raise unless you can offer a strategic reason of your own. The most common way founders waste a quarter is chasing strategic money with a financial pitch, never understanding that the investor was solving a different equation the whole time.

The full Deal Read — the three remaining structures, and the framework for reading which one applies to your own raise — continues below for paid subscribers.

Read more