Read more

Company Overview

In plain English, Yellow Card lets a business hold US dollars, swap between stablecoins, and pay or collect money across borders without waiting three to five days for a wire or bleeding margin to correspondent banks. Its flagship product, Global USD Accounts, gives a company a dollar account that can hold and move stablecoins, manage treasury, and make and collect payments in more than 50 currencies across more than 50 countries. Think of it as the rails that move dollars into and out of markets where dollars are hard to get and slow to move.

On August 4, 2026, Yellow Card closed a $40 million strategic funding round led by SC Ventures, the innovation and investment arm of Standard Chartered, with participation from Sony Innovation Fund, Polychain Capital, and Blockchain Capital. That brings total equity financing past $120 million. The prior rounds were a $40 million Series B in 2022 and a $33 million Series C in October 2024. This is a company that has raised deliberately, not explosively.

The founding team matters here. CEO Chris Maurice and CTO Justin Poiroux started Yellow Card in 2016 and launched commercially in Nigeria in 2019 as a retail crypto exchange built to simplify cross-border payments across Africa. By 2022 they’d reached 16 African countries and claimed more than one million customers. Then, in late 2025, they did the hard thing: they walked away from the retail app that made their name and pivoted entirely to B2B infrastructure. Today the company reports licenses, authorizations, and registrations to operate in more than 20 jurisdictions across North America, Europe, and Africa, and network volume that has passed $10 billion, up from roughly $6 billion earlier in 2026. Named users of its infrastructure reportedly include large payment and financial-services players.

The Market

Start with the size of the problem. The global cross-border payments market processed roughly $190 trillion in transaction value in 2023, per industry data cited by Alphapoint. B2B flows, supplier payments, invoicing, payroll, and wholesale trade, account for around $10 trillion of that annually and are widely projected to become the largest cross-border category by 2030.

Now the friction. Most of that money still rides infrastructure designed in the 1970s. Swift wires often take three to five business days. Correspondent bank fees and FX markups can eat 2% to 7% of a transfer. In markets like Nigeria, Kenya, or across Latin America, dollar scarcity and volatility turn those inefficiencies into a daily operational tax.

Stablecoins are the technical fix, and adoption is real but early. McKinsey and Artemis Analytics pegged genuine stablecoin payment activity, stripping out trading and automated transfers, at about $390 billion in 2025, more than double 2024. B2B payments made up roughly $226 billion of that and grew 733% year over year. Total stablecoin supply crossed $300 billion, up from under $30 billion in 2020, and the US Treasury Secretary has publicly floated supply reaching $3 trillion by 2030.

Here’s the number that keeps me honest, and that you should hold onto: despite all of that, stablecoins remain roughly 1% of global payment flows, the same share reported in 2023 and 2024. Both stories are true at once: explosive absolute growth, and a tiny slice of the total. That’s not a contradiction. It’s the definition of early.

The timing case rests on regulation, not technology. The US GENIUS Act, the EU’s MiCA framework, and Hong Kong’s regime all moved stablecoins from legal gray zone to licensed activity across 2025 and 2026. That’s what turns a Standard Chartered venture arm from a spectator into a lead investor. And the geographic case is specific: in Latin America, an estimated 71% of stablecoin activity is already tied to cross-border payments, the highest share in the world. Yellow Card built its muscles in exactly this kind of market. Now it’s pointing them at LatAm and Asia Pacific.

Business Model and Moat

Yellow Card makes money the way infrastructure companies do: fees and spread on the flows it moves, plus account and treasury services layered on Global USD Accounts. The details of its take rate aren’t public, and I won’t guess at them. What I can tell you is where the durable advantage sits, and it isn’t the blockchain.

The blockchain part is commodity. Anyone can move USDC on a public network. The hard, slow, expensive part is the last mile: licenses to operate legally in each country, and local fiat rails to actually convert stablecoins into naira, or pesos, or rand, and back, at scale, without breaking a rule. Yellow Card’s stack of authorizations across more than 20 jurisdictions, paired with deep local payment rails, is the real barrier. A rival with better code still has to spend years assembling the same license map. That’s the moat.

The competition is stacked and getting stronger. On one side you’ve got crypto-native infrastructure players and stablecoin issuers. On another, big processors and banks are building their own rails. And on a third, Swift itself is upgrading. Yellow Card’s answer is to be the neutral, licensed, regulatory-first layer that any of them can plug into rather than rebuild. Standard Chartered’s venture arm leading this round is a signal that at least one global bank would rather partner than build the emerging-market license stack from scratch.

Spence’s Take

Two mental models from the book do the heavy lifting here.

The first is Disruption Theory, specifically new-market disruption. Clayton Christensen’s insight was that the most dangerous competitors don’t attack incumbents head-on. They start where the incumbent doesn’t care to serve, get good there, then move up. Yellow Card is a near-textbook case. It didn’t launch in New York trying to out-Swift Swift. It launched in Nigeria, in corridors the big banks treat as afterthoughts, solved a painful problem for people the incumbents ignored, and built licensing and rails as it went. Now it’s carrying that capability up-market toward LatAm, APAC, and, eventually, connecting banks directly. Incumbents rarely see this coming because the early market looks too small to bother with. That’s the trap, and it’s the bull case.

The second model is Commoditize Your Complement. The strategic move here is to make the thing next to you cheap so demand for your thing goes up. Yellow Card is commoditizing the correspondent-banking layer, driving the cost and delay of moving dollars toward zero. Great, unless you follow the logic all the way down. Because if moving dollars becomes free and instant, the value of being a middleman in that flow also trends toward zero. And Maurice said the quiet part out loud: he expects a near future where payments flow directly between banks on-chain, “without B2B payments companies or other payment service companies in the flow at all.” Read that twice. The CEO just described the endgame where his own category gets disintermediated. That’s not a gaffe. It’s clear-eyed. But it tells you exactly where the risk lives.

This is where building in regulated markets becomes the whole game, and it’s a pattern I see up close. At /mkt, where I’m CPO, the athlete-tokenization products run on a Reg A+ framework with tZERO as the trading infrastructure. The regulatory and licensing scaffolding is the slow, unglamorous, expensive part, and that’s precisely why it’s defensible. Same logic as Yellow Card: in regulated finance, the paperwork nobody wants to do is often the moat. Code gets copied in a weekend. A license map takes years.

So the bull case: Yellow Card becomes the licensed, neutral rails that banks and fintechs rent instead of build, the picks-and-shovels play for the emerging-market stablecoin era, validated by a strategic backer with skin in the game. The bear case: banks internalize the capability, stablecoin transfers commoditize the margin away, a key market reverses its regulatory stance, or a better-capitalized rival buys the license map faster. All of those are live. The disintermediation clock started the day the round closed.

Why It Matters

For investors and VCs. This is a deal worth watching, not for the round size but for the investor composition. A global bank’s venture arm and a Japanese conglomerate’s fund carry different mandates than the crypto-native money that backed every prior Yellow Card round. That shift is the signal. If you’re tracking the company, watch network volume (it’s moved from about $6 billion to past $10 billion this year), the count and caliber of bank partnerships, license additions in LatAm and APAC, and, most of all, whether take rate holds as volume scales. None of this is a recommendation. It’s a scorecard.

For potential customers. If you’re a CFO, treasurer, importer, exporter, fintech, or bank operating in or into emerging markets, the product coming your way is a dollar account that settles in minutes instead of days, holds and swaps stablecoins, and pays out in local currency across 50-plus countries. The problem it targets is concrete: dollar scarcity, multi-day settlement, and 2% to 7% FX and correspondent friction. If those are line items on your P&L, this is a category to pilot carefully, with your compliance team in the room.

For competitors and builders. The lesson is uncomfortable if you’re a pure technologist: regulatory licensing and local fiat rails beat clever code. Yellow Card’s edge is the boring, hard, regulated layer, not the chain. If you’re building in payments, go get the licenses and the last-mile rails while everyone else demos throughput. And heed Maurice’s own warning: own the customer relationship, not just the pipe, because the pipe is what gets commoditized. The builders who survive disintermediation become a system of record, not just a router.

The Bottom Line

Yellow Card is a disciplined, regulatory-first infrastructure company attacking a massive and genuinely broken market from the edge inward, with a moat made of licenses rather than code. The bull case is that it becomes the neutral rails everyone rents; the bear case is the one its own CEO named, a future where banks route around the middle. Watch whether the moat compounds faster than the disintermediation clock ticks.

This post is for informational and educational purposes only and is not investment advice, a recommendation, or a solicitation to buy or sell any security or asset. Funding figures, valuations, revenue projections, and company metrics are as reported by the company, its investors, or the cited outlets; private-company valuations are point-in-time snapshots that do not reflect public-market prices or predict future performance. Company performance claims and forward-looking projections are attributed to the company and are not independently verified. Strategic-investor participation does not imply any product, adoption, or supply agreement. Always do your own research and consult a licensed professional before making financial decisions.

Read more

On Monday, Corma emerged from stealth with $60 million in seed funding led by Sequoia Capital, with Khosla Ventures and Coatue joining. The company, founded in 2025 and split between Tel Aviv and San Francisco, is building what it calls the first foundation model purpose-built for defensive cybersecurity. CEO Alon Pluda is ex-DeepMind, and the founding team pairs AI researchers with veterans of Israel’s Unit 8200. A $60 million seed is a big number on its own. For a company that shipped its first model about six weeks ago, it’s the kind of check that says investors think they’re buying a category, not a product.

Corma’s argument is structural. General-purpose models got great at offense almost by accident, because finding bugs and writing exploit code lean on the exact reasoning and coding skills those models were trained on. Defense is a different job: sifting oceans of audit logs, catching faint signals over weeks, and making thousands of consistent decisions without slipping once. Corma says its agents, deployed at Fortune 100 and Fortune 500 firms across healthcare, finance, energy, and critical infrastructure, cut threat-response times by more than 94% and widened security coverage 15 times in early use. Those are company-reported figures from its own simulations and early deployments, so treat them as claims, not verdicts. The round didn’t disclose a valuation.

Here’s the model from my book worth naming: Asymmetry.

Most competition isn’t a fair fight, and the sharp move is finding where the payoff is lopsided. Cybersecurity is the textbook case. The attacker has to find one door left open. The defender has to lock every door, every time, forever. Same tools, wildly different odds. When AI supercharged both sides at once, it didn’t level that asymmetry. It widened it, because speed and automation help the side that only has to win once.

Here’s my take.

Most of the AI security noise is about smarter attacks. Corma is betting the bigger opportunity is the unglamorous side of the ledger, the defense that has to be right on the thousandth try, not the first. I spent years as an Army Reserve officer, and asymmetry is one of the first things you learn: the cheaper, faster side sets the tempo, and you don’t beat it by matching it move for move. You beat it by changing the shape of the fight. Whether Corma can actually change that shape is the open question. But its read on where the asymmetry sits looks right to me.

Find the lopsided fight. Then pick the side everyone else is ignoring.

This is for informational and educational purposes only. It isn’t investment, financial, legal, tax, or regulatory advice, and it isn’t an offer or solicitation to buy or sell any security or financial instrument.

If you want the mental models behind breakdowns like this, my book, Mental Models: How to Think, Act, and Win, is on Amazon now.

Get it Here

Share Startup Spotlight

It’s a bank.

Erebor, the startup bank founded by Palmer Luckey, is reportedly nearing a $1.5 billion funding round at an $8 billion valuation. Andreessen Horowitz, Lux Capital and SV Angel are among the investors participating, alongside existing backers including 8VC and Haun Ventures.

The numbers get more interesting.

Erebor received final regulatory approval in February. Since then, deposits have reportedly climbed to $4.6 billion, while annualized recurring revenue has surpassed $100 million.

That’s an unusually fast ramp for a newly approved bank.

And the customer thesis is even more interesting.

Erebor was created partly in response to the collapse of Silicon Valley Bank in 2023. But it isn’t simply trying to recreate the old model. It’s targeting companies operating in capital-intensive sectors such as AI, defense, energy and virtual currencies.

That matters because the economics of these companies don’t look like traditional SaaS.

An AI company might need enormous computing capacity before revenue catches up. A defense startup can spend years getting through procurement. An energy company may need hundreds of millions of dollars before its first commercial project. These businesses need financial infrastructure built around long development cycles and unusual capital requirements.

The mental model: Inversion

One of the mental models in Mental Models: How to Think, Act, and Win is Inversion.

Most founders ask, “How do we win?”

Inversion asks, “How do we lose?”

For a startup bank, the obvious answer is becoming interchangeable with every other financial institution.

So flip it.

What happens if you build the bank specifically for companies that traditional banks struggle to understand?

That’s the bet Erebor appears to be making.

And there’s a broader startup lesson here.

Some of the biggest opportunities aren’t necessarily new products. They’re new infrastructure for industries that are changing faster than the institutions serving them.

We’ve seen versions of this at /mkt. Building in a regulated market means you can’t treat compliance as something to bolt on after the product works. The regulatory structure has to be part of how you build the business from day one.

That’s harder.

It can also create differentiation.

My contrarian take: the next generation of billion-dollar startups may look less like apps and more like infrastructure.

Banks for new industries. Exchanges for new asset classes. Financing platforms for capital-intensive businesses. Compliance systems for regulated markets.

The opportunity isn’t always to invent the next thing.

Sometimes it’s to build the financial and operational rails that make the next thing possible.

This is for informational and educational purposes only. It isn’t investment, financial, legal, tax, or regulatory advice, and it isn’t an offer or solicitation to buy or sell any security or financial instrument.

If you want the mental models behind breakdowns like this, my book, Mental Models: How to Think, Act, and Win, is on Amazon now.

Get it Here

Share Startup Spotlight

Today, MStack AI is reportedly in advanced talks to raise at least $30 million, potentially expanding the round to $40 million or $50 million, at roughly a $230 million pre-money valuation. At the same time, generative AI startup Simplismart is discussing a $50 million round at a valuation somewhere around $120 million to $140 million.

That’s meaningful capital. But the bigger signal is what sits underneath it.

Investors aren’t just chasing another chatbot. MStack is building AI-native chemical R&D and manufacturing technology. Simplismart is focused on generative AI infrastructure. And the investors circling these deals include Alpha Wave, Lightspeed, Prosperity7, Peak XV and potentially Nvidia.

Then there’s the fund side.

Aum Ventures announced the first close of a new ₹750 crore, roughly $80 million deeptech fund, with more than 65% of the initial commitments coming from international LPs. The fund plans to back 25 to 30 companies across AI, semiconductors, space, robotics, defense, energy transition and advanced manufacturing. Initial checks are expected to be $750,000 to $2 million.

That’s a different kind of AI story.

The capital isn’t only flowing toward model companies. It’s moving into businesses where AI intersects with hard science, specialized infrastructure and regulated or difficult-to-replicate industries.

The mental model: Inversion

In Mental Models: How to Think, Act, and Win, I call Inversion one of the simplest ways to improve a decision.

Instead of asking, “How do we win?”

Ask: “How do we definitely lose?”

For AI startups, one obvious answer is building something that can be copied by the next model release.

If your entire moat is a better wrapper around someone else’s model, you should assume that advantage has an expiration date.

Invert the problem.

What would make the business difficult to replace?

Proprietary data. Deep domain expertise. Distribution. Physical infrastructure. Regulatory approvals. Workflow integration. Specialized research. Switching costs.

That’s where today’s funding is getting interesting.

At /mkt, we’re building in a regulated market using Reg A+ offerings and tZERO’s trading infrastructure. The lesson is similar: regulation can look like friction, but the right structure can become part of the product. The key is building the business around the rules, not pretending the rules aren’t there.

My contrarian take: the next great AI companies won’t necessarily look like AI companies.

They’ll look like chemical companies, defense companies, chip companies, financial infrastructure companies and industrial businesses that happen to have AI as a core capability.

The winners won’t just have the best model.

They’ll have the best reason the model matters.

This is for informational and educational purposes only. It isn’t investment, financial, legal, tax, or regulatory advice, and it isn’t an offer or solicitation to buy or sell any security or financial instrument. Funding figures and valuations are based on public reporting and may change or remain unconfirmed.

If you want the mental models behind breakdowns like this, my book, Mental Models: How to Think, Act, and Win, is on Amazon now.

Get it Here

Share Startup Spotlight

On Thursday, LifeMine Therapeutics announced a $188 million Series E led by Milky Way Investments, with new money from Bezos Expeditions and Gates Frontier and returning backers including GSK and ARCH Venture Partners. The company disclosed it alongside a previously raised $75 million Series D, for $263 million total. CEO Greg Verdine has said the Series E came in three to four times larger than the company originally planned for the round.

Now rewind about a year. LifeMine had cut staff and put its fungus-based drug-discovery platform on ice to pour its remaining resources into a single asset: LIFE-001, an experimental immunosuppressant for organ transplant patients, derived from fungi. That’s the same family that gave medicine statins, penicillin, and cyclosporine. The company says the drug has completed dosing in more than 100 healthy volunteers so far. Every forward-looking clinical timeline here is the company’s own, not settled fact.

Here’s the mental model from my book worth applying: Inversion.

Most founders ask, “How do we win?” Inversion flips the question: “How do we lose, and how do we avoid that?” A year ago, LifeMine’s most likely way to lose wasn’t picking the wrong molecule. It was running out of cash before any molecule could prove itself. So the team did the unglamorous thing. It narrowed, cut, and preserved enough runway to still be alive when the biotech funding market thawed.

That’s the whole game in a downturn. You don’t have to be brilliant. You have to still be standing when conditions change. LifeMine froze the exciting part, a platform the company says spans roughly 1,200 potential drug targets, precisely so it could survive to unfreeze it later. This week, with fresh capital, unfreezing it is exactly the plan management laid out.

My contrarian take: in a boom, the market rewards the companies that push the hardest. In a bust, it rewards the ones that refuse to die. Those are different skills, and the second one is wildly underrated. Plenty of “better” biotechs with flashier stories aren’t around to raise this round. LifeMine is, because it inverted the question early and asked how it might not make it, then removed that path.

Survival isn’t the consolation prize. Sometimes it’s the entire strategy.

If this was useful, share it with someone who builds things. And if you want the full toolkit of 50 mental models, you can grab my book, Mental Models: How to Think, Act, and Win, on Amazon right now.

Compliance and attribution: Figures reflect LifeMine’s August 6, 2026 announcement, corroborated by BioPharma Dive, BioSpace, and Endpoints News. LifeMine did not disclose a post-money valuation. Statements about the size of the round relative to plan, platform scope, drug candidate progress, and future clinical timelines are company-reported and forward-looking, not settled fact or independent projection. Nothing here is medical advice or a characterization of any treatment’s safety or efficacy. This post is for informational and educational purposes only. It is not investment advice, research, or an offer or solicitation to buy or sell any security.

If you want the mental models behind breakdowns like this, my book, Mental Models: How to Think, Act, and Win, is on Amazon now.

Get it Here

Share Startup Spotlight

Earlier this week, on August 5, Moove announced a $250 million Series C at a company-reported $2.1 billion valuation. Mubadala led the round, with Woven Capital (Toyota’s growth fund) and Ion Pacific co-leading. The company started in Lagos in 2020 by financing cars for ride-hailing drivers who couldn’t get bank loans, underwriting them on earnings data instead of credit scores. Today it says it runs about 42,000 vehicles across 29 cities in 13 countries, reports roughly $420 million in annual recurring revenue, and already manages autonomous fleets for Waymo in Phoenix and Miami, with London next.

Here’s the part worth sitting with. Moove doesn’t build the self-driving stack. It owns and operates the fleets that self-driving companies need: the vehicles, the charging, the maintenance, and the round-the-clock depot operations it calls “Nests.” Co-founder Ladi Delano’s framing is that five years of running human-driven fleets was the training ground for running autonomous ones. The company says it plans to grow its autonomous workforce from around 150 people today to roughly 500 by year-end.

That’s a clean example of a mental model from my book: Circle of Competence.

The idea is that your edge lives inside the boundary of what you actually know how to do, and the discipline is refusing to step outside it just because the outside looks exciting. Moove didn’t wake up and decide to out-engineer Waymo on autonomy. It looked at what it had already gotten good at (financing, uptime, maintenance, capital efficiency across tens of thousands of vehicles) and extended that competence into the next wave. Same muscle, new market.

Most companies do the opposite in a hype cycle. They abandon the thing they’re great at to chase the thing everyone’s talking about. Moove ran toward the unglamorous, capital-heavy operations layer the frontier players would rather not own.

My contrarian take: in every gold rush, the flashiest layer gets the headlines and the hardest layer gets the margins. Someone still has to own the depreciating asset, keep it charged, and get it back on the road by morning. That’s not a lesser business. In a market this capital-intensive, it might be the better one.

I’ve felt a version of this at /mkt, where we build in a regulated market using Reg A+ offerings and tZERO for trading infrastructure. The edge was never a flashier pitch. It was the operational and regulatory competence to do the unglamorous work correctly, at scale, in a domain where “move fast and break things” earns a very different kind of attention.

Know your circle. Then push its edges. Don’t abandon its center.

If this was useful, share it with someone who builds things. And if you want the full toolkit of 50 mental models, you can grab my book, Mental Models: How to Think, Act, and Win, on Amazon right now.

Compliance and attribution: Figures reflect Moove’s August 5, 2026 announcement (Business Wire), corroborated by TechCrunch, Bloomberg, and Disrupt Africa. The $2.1 billion valuation and the $420 million annual recurring revenue figure are company-reported and point-in-time; private-company valuations and revenue figures can change materially and are not independently verified here. Statements about fleet expansion, workforce growth, and new market launches reflect the company’s stated plans, not independent projections or fact. This post is for informational and educational purposes only. It is not investment advice, research, or an offer or solicitation to buy or sell any security. Author disclosure: Spencer Gareiss is Chief Product Officer at /mkt, referenced above solely as an illustrative example of building in a regulated market.

If you want the mental models behind breakdowns like this, my book, Mental Models: How to Think, Act, and Win, is on Amazon now.

Get it Here

Share Startup Spotlight

On Thursday, Hadrian announced $1.37

billion in Series D financing at a company-reported valuation of $7.87 billion. That’s roughly five times what investors valued it at a year ago, when it raised a $260 million Series C. The round was co-led by WCM Investment Management, Washington Harbour Partners, Valor Equity Partners, 137 Ventures, and Baillie Gifford, with JPMorgan Chase’s Strategic Investment Group anchoring as co-lead. Total capital raised to date sits near $2 billion, per PitchBook estimates.

Here’s what Hadrian actually does. It builds highly automated factories that mass-produce precision parts for submarines, munitions, and aerospace systems the military already relies on. Call it “Factories-as-a-Service,” powered by a software platform named Opus. Four sites, nearly 3 million square feet, including a facility in Muscle Shoals, Alabama tied to Virginia-class and Columbia-class submarine work.

CEO Chris Power put the thesis plainly to Axios: “This has been a five-year slog.” The company got early to a problem the rest of the market only recently woke up to. America can invent world-class defense technology. It just can’t build enough of it, fast enough.

That’s the whole story, and it’s a textbook case of a mental model from my book: the Theory of Constraints.

The idea is simple. Every system has one bottleneck that caps its total output, and improving anything other than that bottleneck is wasted motion. You can pour billions into designing better weapons, but if the factories can’t produce parts at volume, the design work doesn’t ship anything. The constraint isn’t invention. It’s capacity.

Hadrian’s investors just made a $1.37 billion bet on exactly that read. They didn’t fund the flashiest layer of the stack. They funded the choke point everyone else skipped because it’s slow, physical, and unsexy. Submarines don’t ship faster because you drew a better blueprint. They ship faster because someone built the factory.

Here’s my contrarian take. In a market obsessed with the frontier, new AI, new autonomy, new hardware, the biggest edge is often sitting one layer down in the boring part nobody wants to own. The winning move isn’t always a better product. Sometimes it’s removing the single bottleneck that caps everyone’s output at once.

I’ve lived a smaller version of this at /mkt, where we build in a regulated market using Reg A+ offerings and tZERO for trading infrastructure. The hard part was never the idea. It was the plumbing that makes the idea legal and operational at scale. Power’s “five-year slog” line landed for me, because that’s what building through a constraint actually feels like.

Watch where the smart capital flows next. It’s moving toward the bottleneck.

If this was useful, share it with someone who builds things. And if you want the full toolkit of 50 mental models, you can grab my book, Mental Models: How to Think, Act, and Win, on Amazon right now.

Compliance and attribution: Figures reflect Hadrian’s August 6, 2026 announcement (PRNewswire), corroborated by TechCrunch, Bloomberg, CNBC, and Axios. The $7.87 billion valuation is company-reported and point-in-time; private-company valuations can change materially and are not independently verified here. Statements about factory expansion, future production lines, and total capital deployed reflect the company’s stated plans and third-party estimates (PitchBook), not independent projections or fact. This post is for informational and educational purposes only. It is not investment advice, research, or an offer or solicitation to buy or sell any security. Author disclosure: Spencer Gareiss is Chief Product Officer at /mkt, referenced above solely as an illustrative example of building in a regulated market.

If you want the mental models behind breakdowns like this, my book, Mental Models: How to Think, Act, and Win, is on Amazon now.

Get it Here

Share Startup Spotlight

That’s the story. Capital that spent the last few years chasing apps and models is now chasing things that plug into the grid, come out of the ground, or run an operation that used to need a room full of people.

Here are the five that actually matter, with a hot take on each. Then I’ll go long on the one I’d circle.

1. Base Power — $1B Series D at a $13B valuation

Austin’s Base Power raised $1 billion in a Series D on August 3 at a $13 billion post-money valuation, its second billion-dollar round in about ten months, per the company. Ribbit, Addition, Valor Equity Partners, and JPMorganChase’s Strategic Investment Group led it, and Base says total funding now tops $2.5 billion. The same day, it started producing Base Core, a roughly 39 kWh home battery built at its Austin factory.

Spence’s take: Base isn’t selling batteries. It’s selling grid capacity one garage at a time and pricing it like infrastructure.

2. Valar Atomics — $1B Series B

Nuclear startup Valar Atomics announced a $1 billion Series B on August 3, led by Sequoia, plus a separate $200 million credit facility. The company didn’t disclose a valuation; Bloomberg reported it at $6 billion. Founder Isaiah Taylor, who started the company in 2023, is building small, factory-made reactors aimed at AI data centers, and says one of its reactors recently powered an Nvidia chip off a live fission reaction.

Spence’s take: When the pitch is “reactors as a product, not a construction project,” physics was never the hard part. The assembly line is.

3. Mariana Minerals — $310M Series B

Mariana Minerals raised $310 million on August 3, led by Khosla Ventures with a16z, Breakthrough Energy Ventures, and Mitsubishi joining, bringing total capital to roughly $400 million per the company. Fortune reported a $1.5 billion valuation. The plan: AI-run mines and refineries (copper in Utah, lithium in Texas) to cut U.S. dependence on foreign critical minerals.

Spence’s take: Mining is the least “software” business there is, which is exactly why software eating it would matter.

4. HappyRobot — $150M Series C at $1.2B

Madrid’s HappyRobot closed a $150 million Series C on August 4 at a $1.2 billion post-money valuation, led by Prysm Capital and co-led by Eurazeo. Founded in 2022, it says revenue is up 5x since its $44 million Series B late last year, with 150-plus enterprise customers including DHL, Uber Freight, and Kuehne+Nagel.

Spence’s take: The AI agents that win won’t be the ones that draft the email. They’ll be the ones that make the call and book the load.

5. Decade — $85M seed

Brazil’s Decade came out of stealth on August 4 with an $85 million seed, which the company calls the largest ever for a Latin American startup, backed by Greenoaks, Benchmark, and Diffusion. The founders, ex-Nubank CTO Vitor Olivier and Hyperplane’s Felipe Meneses, are building an AI-native wealth advisor that pairs every client with a senior human advisor plus software.

Spence’s take: An $85M seed for a pre-product company is a bet on the founders’ last ten years, not their next twelve months.

——————————————————————————————————————

The one I’d circle: Base Power

Here’s the mental model. Second-order thinking.

First-order read: home batteries are selling, so the company’s worth more. Second-order read: every battery Base installs becomes a node it can dispatch, so the fleet quietly turns into a power plant that grows without Base ever building one. That’s why a $13 billion price on what looks like a hardware company doesn’t read as crazy to its backers. They aren’t pricing battery margins. They’re pricing what a few hundred thousand coordinated batteries do to the grid.

And notice how Base makes money: selling power directly in retail-choice states like Texas, and in regulated markets handing utilities control of the batteries, who then pay for the access. That regulated-markets piece is the part most builders miss. The moat isn’t only the hardware. It’s the permission to operate. Companies that treat regulation as terrain to master rather than a wall to climb tend to compound quietest and longest. It’s the same reason we built /mkt inside Reg A+ instead of around it. When the rules are the hard part, clearing them is the advantage.

——————————————————————————————————————

Next week

Next week’s Startup Spotlight deep-dive goes long on one of these, and I’m leaning Valar Atomics: what “factory-built nuclear” actually has to prove before the valuation makes sense, and what the bull and bear cases look like from here. Paid subscribers get the full breakdown.

If this was useful, share it with someone who builds things. And if you want the full toolkit of 50 mental models, you can grab my book, Mental Models: How to Think, Act, and Win, on Amazon right now.

Share

If you want the mental models behind breakdowns like this, my book, Mental Models: How to Think, Act, and Win, is on Amazon now.

Get it Here

Share Startup Spotlight

Base Power, the Austin company founded by Zach Dell, closed a $1 billion Series D on Monday at a $13 billion post-money valuation, according to the company. The round was led by Ribbit, Addition, Valor Equity Partners, and JPMorganChase’s Strategic Investment Group, part of the bank’s Security and Resiliency Initiative, with a long list of returning backers including Thrive Capital, a16z, and Lightspeed. That’s a 3.25x jump from the $4 billion valuation Base set at its Series C just nine months ago, and it takes total capital raised past $2.5 billion in under three years. Alongside the raise, Base launched Base Core, a 39.2 kWh home battery built at its Austin factory. Today the company operates in two states, Texas and Illinois, with utility partnerships covering more than 200 megawatts.

Here’s the part the valuation is actually pricing.

Optionality. It’s the value of having the right, but not the obligation, to act when conditions swing your way. A single home battery gives one family backup power during an outage. But a fleet of them, dispatchable on command, gives Base the right to push capacity into the grid exactly when the grid is most stressed and power is most expensive. Each battery is a cheap call option on grid scarcity. And options are worth the most when volatility is high. The AI buildout is manufacturing grid volatility at industrial scale: surging demand, interconnection queues that run six to ten years, and capacity prices that have spiked hard in some markets. Base isn’t really selling hardware. It’s assembling one of the largest portfolios of grid-scarcity options in the country, and it’s doing it while the volatility keeps climbing.

I spent years around market structure, so here’s what I’d actually watch. An option is only worth something if you can exercise it. Base’s value depends on dispatching that distributed capacity into markets that pay for it, and those markets are regulated, fragmented, and different in every state. A $13 billion valuation on a company operating in two states isn’t priced for 200 megawatts of backup batteries. It’s priced for Base becoming national grid infrastructure, for the whole option portfolio to actually clear. The hardware is proven. The exercise is the bet.

So track one number, and it isn’t the valuation. Track how fast Base can turn “installed batteries” into “contracted, dispatchable capacity” in each new market it enters. That conversion rate is the real business.

If this was useful, share it with someone who builds things. And if you want the full toolkit of 50 mental models, you can grab my book, Mental Models: How to Think, Act, and Win, on Amazon right now.

Spence

This post is for informational and educational purposes only and is not investment advice, a recommendation, or a solicitation to buy or sell any security or asset. Funding figures, valuations, and company metrics are as reported by the company or the cited outlets; the valuation is a company-reported, point-in-time private figure that does not reflect public-market prices or predict future performance. Always do your own research and consult a licensed professional before making financial decisions.

If you want the mental models behind breakdowns like this, my book, Mental Models: How to Think, Act, and Win, is on Amazon now.

Get it Here

Share Startup Spotlight

Base Power, the Austin company founded by Zach Dell, closed a $1 billion Series D on Monday at a $13 billion post-money valuation, according to the company. The round was led by Ribbit, Addition, Valor Equity Partners, and JPMorganChase’s Strategic Investment Group, part of the bank’s Security and Resiliency Initiative, with a long list of returning backers including Thrive Capital, a16z, and Lightspeed. That’s a 3.25x jump from the $4 billion valuation Base set at its Series C just nine months ago, and it takes total capital raised past $2.5 billion in under three years. Alongside the raise, Base launched Base Core, a 39.2 kWh home battery built at its Austin factory. Today the company operates in two states, Texas and Illinois, with utility partnerships covering more than 200 megawatts.

Here’s the part the valuation is actually pricing.

Optionality. It’s the value of having the right, but not the obligation, to act when conditions swing your way. A single home battery gives one family backup power during an outage. But a fleet of them, dispatchable on command, gives Base the right to push capacity into the grid exactly when the grid is most stressed and power is most expensive. Each battery is a cheap call option on grid scarcity. And options are worth the most when volatility is high. The AI buildout is manufacturing grid volatility at industrial scale: surging demand, interconnection queues that run six to ten years, and capacity prices that have spiked hard in some markets. Base isn’t really selling hardware. It’s assembling one of the largest portfolios of grid-scarcity options in the country, and it’s doing it while the volatility keeps climbing.

I spent years around market structure, so here’s what I’d actually watch. An option is only worth something if you can exercise it. Base’s value depends on dispatching that distributed capacity into markets that pay for it, and those markets are regulated, fragmented, and different in every state. A $13 billion valuation on a company operating in two states isn’t priced for 200 megawatts of backup batteries. It’s priced for Base becoming national grid infrastructure, for the whole option portfolio to actually clear. The hardware is proven. The exercise is the bet.

So track one number, and it isn’t the valuation. Track how fast Base can turn “installed batteries” into “contracted, dispatchable capacity” in each new market it enters. That conversion rate is the real business.

If this was useful, share it with someone who builds things. And if you want the full toolkit of 50 mental models, you can grab my book, Mental Models: How to Think, Act, and Win, on Amazon right now.

Spence

This post is for informational and educational purposes only and is not investment advice, a recommendation, or a solicitation to buy or sell any security or asset. Funding figures, valuations, and company metrics are as reported by the company or the cited outlets; the valuation is a company-reported, point-in-time private figure that does not reflect public-market prices or predict future performance. Always do your own research and consult a licensed professional before making financial decisions.

If you want the mental models behind breakdowns like this, my book, Mental Models: How to Think, Act, and Win, is on Amazon now.

Get it Here

Share Startup Spotlight

Valar Atomics closed a $1 billion Series B led by Sequoia Capital, with partner Shaun Maguire joining the board, alongside a separate $200 million credit facility (Erebor Bank and J.P. Morgan, per reporting). That’s $1.2 billion in fresh financing for a company that didn’t exist three years ago. Bloomberg reported the round valued Valar at roughly $6 billion; the company didn’t confirm the figure, so treat it as reported, not stated. Even so, it’s about triple the $2 billion valuation Valar carried after its roughly $450 million round back in April.

The trigger was a July milestone. Valar’s Ward 250 reactor became the first in US history to power an Nvidia AI chip directly off a fission reaction, one of only four companies to start a self-sustaining chain reaction this year, per reporting. Valar and Nvidia are now planning a 30-megawatt, waterless, nuclear-powered AI facility in Utah.

So why is a nuclear company suddenly priced like software? One mental model explains the whole thing.

Theory of Constraints. Every system has exactly one binding constraint at a time, and improving anything other than that constraint is wasted motion. For three years, the industry treated compute as the constraint: buy more GPUs, win. But you can’t run a GPU you can’t plug in. S&P Global estimates the industry wants to bring roughly 50 gigawatts of US data centers online over the next three years, and that only about half of that can actually connect to the grid. Interconnection queues run six to ten years. PJM’s capacity prices jumped 833% in a single delivery cycle. The constraint moved. It’s electrons now, and Sequoia’s billion dollars is a bet on who owns the one input everyone else is short on.

Here’s what most people are reading wrong. The headline isn’t “nuclear is back.” It’s that a chipmaker is now underwriting reactors, because the fastest way to sell more chips is to manufacture the power to run them. Silicon and electricity stopped being two separate industries the moment a reactor lit up a Blackwell.

The risk is exactly as real as the thesis. Valar still operates under a Department of Energy pilot authorization and needs a full Nuclear Regulatory Commission license before it can sell a single commercial kilowatt-hour. A $6 billion price assumes it clears a regulatory gap that has humbled better-funded reactor companies. That’s priced-in perfection meeting one of the slowest approval processes in American industry. If you want to track one thing, track the licensing timeline, not the physics.

The physics already worked. The question a billion dollars just asked is whether the paperwork can move at the speed of an AI roadmap.

If this was useful, share it with someone who builds things. And if you want the full toolkit of 50 mental models, you can grab my book, Mental Models: How to Think, Act, and Win, on Amazon right now.

Spence

This post is for informational and educational purposes only and is not investment advice, a recommendation, or a solicitation to buy or sell any security or asset. Funding figures and financing terms are as reported by the company, its investors, or the cited outlets. Valar Atomics’ valuation is attributed to Bloomberg’s reporting and was not confirmed by the company; it is a point-in-time, private-company figure that does not reflect public-market prices or predict future performance. Regulatory status is as reported and subject to change. Always do your own research and consult a licensed professional before making financial decisions.

Read more

Company Overview

In plain English, Eliyan builds the connective tissue between chips. Modern AI accelerators stopped being single slabs of silicon years ago. They’re now clusters of smaller chips, called chiplets, stitched together inside a package, then packages stitched into racks, then racks stitched into clusters. Every one of those stitches is a place where data can get stuck. Eliyan makes the interconnect, the physical and electrical link, that moves data between those pieces faster and, critically, with less power per bit.

On July 29, Eliyan announced it completed a Series C totaling $145 million at a $1 billion valuation, crossing into unicorn territory. The oversubscribed round was led by Seligman Ventures, whose managing partner Umesh Padval joins the board, with two notable new strategic backers: Cisco Investments and Lumentum. Worth flagging up front: strategic investment is capital, not a purchase order. No joint products or supply agreements with either company have been disclosed, so read their participation as a signal of technical fit, not confirmed adoption.

The team is the tell. CEO and co-founder Ramin Farjadrad was CTO and VP of networking at Marvell, one of the deepest benches in high-speed connectivity anywhere. He co-founded Eliyan in 2021 with Syrus Ziai and Patrick Soheili. This isn’t a first rodeo. Eliyan raised a $40 million Series A in 2022 and a $60 million Series B in 2024, and separately took $50 million in strategic investment in January 2026 from a roster that reportedly included AMD, Arm, Coherent, and Meta. A quick compliance note for anyone doing the math: the company describes the Series C as “completed at a total of $145 million” and hasn’t said whether the January strategic money sits inside that figure, so don’t simply stack the numbers into a single total.

Its products carry the house branding: NuLink PHYs and NuGear chiplet families, the silicon-proven building blocks. Days before the raise, Eliyan unveiled NuLink-XD, a 224G interface built on TSMC’s 3nm process. The company claims its interconnect delivers up to four times the performance at half the power of conventional approaches. That’s a company claim, not an independent benchmark, so hold it accordingly.

The Market

Here’s the problem almost nobody outside the industry talks about. The world’s most expensive AI chips spend a large share of their lives waiting. Industry estimates put GPU utilization in large clusters at roughly 30% to 40%, and it’s often not a software problem. The chips are idle because they’re waiting for data to arrive. When you can’t feed silicon fast enough, you’ve bought a Ferrari to sit in traffic.

That problem is about to get worse in a very specific way. Inference, the work of running trained models in production, is projected to grow from roughly one-third of AI compute in 2023 to about two-thirds by the end of 2026. Inference is bandwidth-hungry and latency-sensitive. As the industry shifts from training a handful of giant models to serving billions of queries, the value migrates from raw compute toward moving data between chips without stalling. The bottleneck stopped being how fast a chip thinks and became how fast you can talk to it.

Now connect that to the thing I wrote about last week. Interconnect efficiency isn’t only a speed story, it’s a power story. Every wasted bit of data movement is wasted electricity, and electricity is the input the entire AI buildout is now short on. S&P Global estimates only about half of the roughly 50 gigawatts of US data centers the industry wants over the next three years can actually plug into the grid. When power is the binding constraint, the company that moves more data per watt isn’t selling a nice-to-have. It’s selling headroom inside a fixed power budget. That’s where silicon and electricity become one problem, and it’s why an interconnect startup belongs in the same conversation as a reactor company.

Timing is the last piece. The industry is mid-transition from monolithic dies to chiplet architectures, and from purely electrical links toward co-packaged optics for the longer runs. Transitions are when incumbents’ advantages reset and new standards get set. Eliyan is positioning to span that whole continuum, from electrical die-to-die inside a package out to optical rack-to-rack fabrics.

Business Model and Moat

Eliyan makes money two ways: licensing its interconnect IP, and selling chiplet products, to hyperscalers, chip developers, memory suppliers, and optical module makers. That model has a specific shape. Design cycles are long and painful to win, but once your IP is designed into a customer’s chip, you’re locked in for years. The company’s Chief Strategy and Business Officer, Patrick Soheili, has said Eliyan expects revenue to grow from low single-digit millions in 2025 to hundreds of millions by the end of 2027, with initial chiplet shipments beginning in 2026. Treat that as a company projection, not a fact on the ground. It’s the number the whole thesis rides on, and it hasn’t happened yet.

The competitive field is crowded and includes the most powerful company in the industry. Nvidia has NVLink for chip-to-chip and, through its Mellanox acquisition, InfiniBand for cluster networking. Both are excellent and both are proprietary, which is precisely the opening. There’s also UCIe, the Universal Chiplet Interconnect Express, an open industry standard backed by a consortium that wants die-to-die connection to become a commodity. And there are the SerDes incumbents like Broadcom and Marvell, plus a wave of optical interconnect startups such as Ayar Labs, Celestial AI, and Lightmatter chasing the same co-packaged-optics future.

So what’s defensible? Three things, in order of durability. First, silicon-proven PHY IP and the patents around it. Farjadrad’s team has been taping out real interconnect on leading-edge nodes since 2022, and interconnect physics is genuinely hard to copy. Second, neutrality. Eliyan is foundry-agnostic, memory-agnostic, and increasingly optics-agnostic, which makes it the natural choice for any hyperscaler or chip developer who wants to mix best-of-breed parts without being taxed by a single vendor’s ecosystem. Third, design-win stickiness. Once you’re in the chip, you’re in for the product’s life.

The moat, honestly stated, is “be the neutral standard in a market defined by one dominant proprietary player.” That’s a real position. It’s also a position that lives or dies on whether an open standard commoditizes it first.

Spence’s Take

Two mental models here, one for the bull case and one for the bear.

First Principles for the bull case. The lazy way to scale an AI chip is to make the die bigger. Physics says no: you hit the reticle limit, yields collapse, costs explode. Eliyan didn’t accept the inherited assumption that a processor should be one big monolithic thing. They went back to the physics and asked a cleaner question: if the future is many small chiplets, what’s the best possible way to connect them? Rebuild from that question and interconnect stops being an afterthought bolted onto a chip and becomes the architecture. That’s why the “half the power” claim matters more than the “four times faster” claim. Faster wins benchmarks. Lower power per bit wins in a world where every data center is running into a power wall. A silicon company whose core value proposition is fewer watts is, quietly, an energy company. That’s the whole reason it belongs in this week’s deep-dive.

Inversion for the bear case. Don’t ask what makes Eliyan win. Ask what has to be true for it to fail, then check how likely each of those is. It fails if UCIe matures fast enough that die-to-die connection becomes a free commodity and the IP premium evaporates. It fails if Nvidia’s proprietary NVLink plus its captive ecosystem simply keeps the volume, because customers would rather buy the whole stack from one vendor than assemble best-of-breed. It fails if the design wins slip and that low-millions-to-hundreds-of-millions revenue ramp stalls out in the gap between “designed in” and “shipping.” And it fails if the optical transition rewards a fundamentally different architecture than the one Eliyan is building toward. None of those is far-fetched. At least two of them are actively being pursued by companies with more money than Eliyan will ever raise. The bull case is real. So is the list of ways it ends.

Why It Matters

For investors and VCs. This is a deal worth watching, not for the valuation, which is a company-reported, point-in-time private figure, but for what it signals about where AI capital is rotating: away from models and toward the plumbing. If you’re tracking the story rather than the stock, the signals that actually matter are concrete. Watch for design-win announcements with named customers. Watch whether initial chiplet shipments actually land in 2026 as promised. Watch whether Cisco and Lumentum convert from investors into customers, because that’s the difference between a strategic bet and a strategic relationship. And watch the UCIe roadmap, because the pace of that standard is the single biggest external variable in the thesis. None of this is a recommendation to do anything. It’s a list of what would confirm or break the story.

For potential customers. If you’re a hyperscaler, a chip developer, or a system builder, what’s coming is an alternative to buying your entire scale-up architecture from one vendor. Eliyan is building a foundry-neutral, memory-agnostic interconnect that lets you combine the best chiplets from different sources and squeeze more usable work out of the same power envelope. If GPU utilization in your clusters is stuck in the 30s and 40s, the problem you’re feeling is very likely the wire, not the chip, and the fix is better interconnect, not more accelerators you can’t fully feed.

For competitors and builders. The lesson here is portable and it’s the same lesson as last week’s reactor. Find the constraint everyone is ignoring and own it. Compute got all the attention while interconnect and power quietly became the real limits. Eliyan’s second lesson is about positioning: in a market with one dominant proprietary player, the durable move is often to be the neutral layer everyone else can build on, not to fight the giant head-on. And the third lesson is temperamental. Eliyan sells the boring connective tissue, the wires and the PHYs, in an industry obsessed with the glamorous part. Boring, load-bearing, and hard to replace is a very good place to build a business.

The Bottom Line

Eliyan is a credible, well-led bet on the least glamorous and most load-bearing problem in AI infrastructure, and the fact that its core advantage is measured in watts is what makes it matter beyond semiconductors. The bull case is that it becomes the neutral standard for connecting a chiplet world that’s also power-constrained. The bear case is that an open standard or a proprietary giant gets there first, and the revenue ramp everyone is underwriting never fully arrives.

This is Startup Spotlight. Every week I break down a company worth watching. Subscribe for $7.99/mo to get the full analysis.

Spence

This post is for informational and educational purposes only and is not investment advice, a recommendation, or a solicitation to buy or sell any security or asset. Funding figures, valuations, revenue projections, and company metrics are as reported by the company, its investors, or the cited outlets; private-company valuations are point-in-time snapshots that do not reflect public-market prices or predict future performance. Company performance claims and forward-looking projections are attributed to the company and are not independently verified. Strategic-investor participation does not imply any product, adoption, or supply agreement. Always do your own research and consult a licensed professional before making financial decisions.

Read more

In April, Valar Atomics was worth $2 billion. This week it’s worth $6 billion, according to Bloomberg. The number of reactors it has running went from one to one.

That’s the setup for the biggest venture story of the week. On Monday, Valar closed a $1 billion Series B led by Sequoia Capital, with partner Shaun Maguire joining the board. The company also lined up a $200 million credit facility led by Erebor Bank and J.P. Morgan, per Valar’s own announcement. Add it up and a three-year-old startup founded by Isaiah Taylor is sitting on roughly $1.2 billion in fresh financing.

Here’s what makes it interesting. Valar builds small modular reactors, the high-temperature, gas-cooled kind. In June, its Ward 250 unit hit self-sustaining criticality and powered an Nvidia Blackwell chip, and Valar signed a deal with Nvidia to develop a waterless 30MW AI facility. Its earlier rounds were smaller and quieter: $130 million last November, backed by Palmer Luckey and Palantir’s Shyam Sankar, then $450 million in April at that $2 billion mark.

So why does one working reactor command a reported $6 billion? Use the model.

Theory of Constraints.

Every system has exactly one binding constraint. Improve anything else and total throughput doesn’t budge. For two years, the loud constraint on AI was compute: who has the chips. That constraint quietly moved. Training and inference need round-the-clock power that solar and wind can’t reliably supply on their own, and AI electricity demand is projected to top 200 terawatt-hours a year by the end of the decade. The bottleneck is now megawatts.

Valar’s pitch isn’t “better reactors.” It’s reactors as a manufactured product instead of a custom megaproject. In the company’s words, a project can build one reactor, but a fleet requires manufacturing. If power is the constraint, the winner isn’t whoever builds the best plant once. It’s whoever can stamp out standardized units fast enough to keep up. That’s what Sequoia is pricing.

At /mkt, we build in a regulated market too, and the lesson rhymes: the technical constraint is rarely the last one standing.

Which is the contrarian part. Theory of Constraints has a punchline people skip. Break one constraint and it doesn’t disappear. It moves. Valar’s binding constraint isn’t physics anymore. It’s a commercial license from the Nuclear Regulatory Commission that it doesn’t have yet, in an industry famous for over-budget, over-schedule projects. A billion dollars doesn’t buy that approval. It buys a factory.

So the real bet here isn’t on the reactor. It’s on whether a manufacturing mindset can compress a regulatory timeline nobody has compressed before. If you want to know whether this round ages well, don’t watch the valuation. Watch the license.

If you want the mental models behind breakdowns like this, my book, Mental Models: How to Think, Act, and Win, is on Amazon now.

Get it Here

Share Startup Spotlight

Everybody’s building an “AI security” startup right now. Only one of them has run 310,000 real attacks inside live production networks. That’s the whole story.

On August 3, Horizon3 announced a $250 million Series E at a company-reported valuation of more than $2 billion. Per the company, that’s triple the $650 million valuation it carried after its Series D in June 2025. NightDragon and NEA, both existing backers, co-led the round. Seven new investors and five returning ones came in, and the company says it was oversubscribed. (Valuation figures here are company-reported and reflect a single point in time.)

Horizon3’s product, NodeZero, does autonomous penetration testing. It attacks your own network the way a real adversary would, finds the exploitable path, tells you how to fix it, then verifies the fix. No consultant, no annual engagement, no PDF that’s stale the day you open it. TechCrunch reported the raise landed the same week two AI labs disclosed their models had breached systems outside their intended scope. The timing wasn’t subtle.

Here’s the mental model: Cornered Resource.

A cornered resource is a single valuable asset your competitors can’t get, no matter how much money they raise. It’s not the algorithm. Algorithms get copied. It’s the thing underneath. For Horizon3, that’s 310,000 authorized attacks run inside real production environments across roughly 7,000 organizations the company reports serving, including banks, hospitals, and federal agencies. CEO Snehal Antani called it a “data moat,” and he’s right to. You can’t buy that dataset. You can’t scrape it. You can only earn it, one signed authorization at a time, over six years, by being trusted enough to attack somebody’s live bank.

That’s the part founders miss. Everyone obsesses over the model. The defensible asset is usually the boring, permission-gated data that took years to accumulate and can’t be recreated on a whiteboard.

I’ve watched this up close at /mkt, where we build in regulated markets using Reg A+ offerings and tZERO for trading infrastructure. The edge there isn’t a clever feature either. It’s doing the slow, compliance-heavy infrastructure work most teams route around. Regulated trust compounds. It’s a cornered resource too.

The contrarian take:

Horizon3’s $250 million isn’t really a bet on AI. Plenty of companies have AI. It’s a bet that six years of accumulated, permissioned attack data is a wall nobody can climb fast enough to matter. If that thesis holds, the moat was dug long before this round. The money just pays to defend it.

So when you’re sizing up your own company, or somebody else’s, ask the harder question. Not “do they have AI.” Everyone has AI. Ask what they have that no amount of funding can buy. That answer, if it exists, is the whole business.

My takeaway is simple.

The founders who’ll stand out over the next decade won’t necessarily be the smartest in the room. They’ll be the ones whose incentives let them keep showing up after everyone else has been given a reason to leave.

That’s harder to measure than ARR.

It may also be more valuable.

This post is for informational purposes only and is not investment advice, a recommendation, or a solicitation to buy or sell any security. Figures are as reported by the companies involved and have not been independently verified. Company names and products are referenced for illustration only. Always do your own research and consult a licensed professional before making financial decisions.

If you want the mental models behind breakdowns like this, my book, Mental Models: How to Think, Act, and Win, is on Amazon now.

Get it Here

Share Startup Spotlight

Every startup pitch used to end with the same question: How big can this become?

Now there’s a new one.

Are you actually going to stick around?

That’s not a hypothetical. It’s becoming one of the defining themes in venture capital. As AI labs and large tech companies compete aggressively for top talent, venture investors are increasingly worried that founders will sell early, accept an acquihire, or simply leave for a multi-million dollar compensation package before building an enduring company. According to reporting today, U.S. AI startup acquihires have climbed sharply, reaching 245 transactions by mid-2026. Some investors say promising deals have already fallen apart after founders accepted offers from frontier AI labs instead.

Think about how unusual that is.

For years, venture investing focused on market size, product quality, customer growth, and capital efficiency. Those still matter. But when the opportunity cost of remaining a founder skyrockets, commitment itself becomes a competitive advantage.

One mental model from Mental Models: How to Think, Act, and Win fits perfectly here: Incentives Drive Behavior.

People don’t respond to speeches. They respond to incentives.

Today’s incentives are powerful. A talented founder can spend years building an uncertain startup or accept an enormous guaranteed package from an AI company with nearly unlimited compute, distribution, and capital. That’s a rational decision for many people.

The interesting part is what happens next.

Investors aren’t just asking whether a founder is technically capable. They’re trying to understand whether that founder has reasons to keep building when the easy exit appears.

That’s a much harder question to diligence.

It also changes how founders should think about building companies. Durable businesses aren’t created by people chasing the highest short-term offer. They’re built by people whose incentives remain aligned with solving a problem over many years.

You can see similar dynamics in regulated industries. At /mkt, for example, building infrastructure around Regulation A+ offerings and secondary trading requires patience, compliance, and long development cycles. Those markets don’t reward shortcuts. They reward consistency and execution over time.

None of this means founders should reject acquisition offers or employment opportunities. Every situation is different, and there isn’t a universal right answer. But it does mean investors are paying closer attention to founder motivation because incentives often predict behavior better than pitch decks do.

My takeaway is simple.

The founders who’ll stand out over the next decade won’t necessarily be the smartest in the room. They’ll be the ones whose incentives let them keep showing up after everyone else has been given a reason to leave.

That’s harder to measure than ARR.

It may also be more valuable.

This post is for informational purposes only and is not investment advice, a recommendation, or a solicitation to buy or sell any security. Figures are as reported by the companies involved and have not been independently verified. Company names and products are referenced for illustration only. Always do your own research and consult a licensed professional before making financial decisions.

If you want the mental models behind breakdowns like this, my book, Mental Models: How to Think, Act, and Win, is on Amazon now.

Get it Here

Share Startup Spotlight

The hardest part of selling a blockbuster weight-loss drug in 2026 isn’t inventing the molecule or marketing it. It’s filling enough injector pens, fast enough, to keep up with demand. That unglamorous problem is where a lot of money is quietly moving.

On July 30, the contract manufacturer Resilience and Eli Lilly announced a combined $750 million investment to expand pharmaceutical manufacturing in the Cincinnati area, per the companies’ announcement. The expansion adds production of Lilly’s KwikPen injectable device, the delivery system for its diabetes and obesity medicines, at a site that’s been making Mounjaro and Zepbound since the two companies first partnered in 2023. The companies say the project adds at least 400 skilled jobs, pushes Resilience’s Ohio headcount past 1,400, and reaches full operations in early 2027.

Notice what this isn’t. It isn’t a new drug, a new trial, or a flashy AI round. It’s capacity. Specifically, it’s domestic capacity to fill and finish the devices patients actually inject, which has been one of the tightest chokepoints in the entire obesity-drug boom.

Here’s the model I’d reach for. Cornered Resource: when a company secures preferential access to a scarce, valuable asset that rivals can’t easily replicate, it builds an edge no amount of marketing can copy. Everyone in this category has a molecule. Far fewer have guaranteed, at-scale, U.S.-based fill-finish capacity locked in years ahead of demand. By funding dedicated capacity through a partner instead of fighting for slots on the open market, Lilly is turning a shared bottleneck into something closer to a private one.

That reframes how to read the whole GLP-1 story. The headlines track prescriptions, sales, and which drug posts the best trial data. But the constraint that actually caps growth right now is physical: how many pens roll off a line each week, and where those lines sit. Demand for these drugs isn’t the scarce thing. Supply is. And supply is a manufacturing and logistics problem, not a marketing one.

So here’s the contrarian close. In a gold rush, everyone studies the gold. The durable winners study the supply chain. The companies pulling ahead in obesity drugs aren’t only the ones with the best science, they’re the ones who saw two years ago that capacity would be the fight and started building while competitors were still counting demand.

One caution, because it cuts both ways. Capacity you build today is a bet that today’s demand holds. Concentrated, dedicated capacity is also concentrated exposure if pricing, policy, or a next-generation oral pill changes the picture. A cornered resource is powerful right up until the thing it corners stops being scarce.

If this was useful, share it with someone who builds things. And if you want the full toolkit of 50 mental models, you can grab my book, Mental Models: How to Think, Act, and Win, on Amazon right now.

Spence

This post is for informational purposes only and is not investment advice, a recommendation, or a solicitation to buy or sell any security. Figures are as reported by the companies involved and have not been independently verified. Company names and products are referenced for illustration only. Always do your own research and consult a licensed professional before making financial decisions.

If you want the mental models behind breakdowns like this, my book, Mental Models: How to Think, Act, and Win, is on Amazon now.

Get it Here

Share Startup Spotlight

Give a new hire the keys to every room in the building on their first day. Now let them work at machine speed, around the clock, and never ask permission twice. That’s roughly what a lot of companies are doing with AI agents right now, and this week the money started betting on the cleanup.

On July 28, a Tel Aviv startup called Act Security walked out of stealth with $60 million and a blunt thesis: the problem isn’t that companies can’t find their security risks, it’s that AI can now exploit those risks faster than any human can patch them. The funding came in two rounds, a $20 million seed led by Team8 and Bessemer Venture Partners and a $40 million Series A led by Notable Capital, per the company’s launch announcement. The founders previously built Medigate, the medical-device security company Claroty acquired for $400 million. Act’s pitch is to shrink the access surface itself. The company says close to 97% of cloud permissions sit dormant and unused, quietly waiting to be inherited by a workload or an agent that shouldn’t have them.

Act wasn’t alone. The same day, Hush Security added a $30 million Series A, backed by Akamai, Battery Ventures, and YL Ventures, to govern non-human identities and enterprise AI agents. Per Crunchbase, startups at the intersection of AI and security have raised around $855 million in 2026. The pattern is hard to miss: as agents get deployed, someone has to hold their keys.

Here’s the model I’d apply. Gall’s Law: a complex system that works is almost always found to have evolved from a simple system that worked. You don’t design a working complex system from scratch. You grow one from a small thing that already does its job.

That’s the lens for this whole category. Cloud security spent a decade building the comprehensive dashboard that surfaces every possible risk, and AI just made those dashboards impossible to act on in time. The teams getting funded now aren’t promising to govern everything at once. They’re starting from one narrow, almost boring primitive: kill the access nobody uses, enforce least privilege, remove the path instead of watching it.

So here’s the contrarian close. Over the next year, you’ll see a stack of startups pitch the all-in-one “AI agent control plane,” the single platform that governs every agent everywhere. Gall’s Law says be skeptical of the grand version. The winners in this wave probably won’t be whoever promises the most. They’ll be whoever nails the smallest thing that actually works, then earns the right to grow from there. In security, boring and deterministic tends to beat ambitious and sprawling.

The agents are already inside the building. The question every operator should be asking isn’t how smart they are. It’s how many doors they can open.

That’s it for today. If you want the full set of 50 mental models I use to read moves like this one, you can grab my book, Mental Models: How to Think, Act, and Win, on Amazon right now.

Spence

This post is for informational purposes only and is not investment advice, a recommendation, or a solicitation to buy or sell any security or asset. Funding figures and company claims are as reported by the companies and cited outlets and have not been independently verified. Always do your own research and consult a licensed professional before making financial decisions.

If you want the mental models behind breakdowns like this, my book, Mental Models: How to Think, Act, and Win, is on Amazon now.

Get it Here

Share Startup Spotlight

Here are the five that actually matter.

1. Atoms raised $1.7 billion to automate the physical world

Travis Kalanick is back. His industrial AI company, Atoms, closed a $1.7 billion round led by Andreessen Horowitz, with Ben Horowitz joining the board and, in a plot twist, Uber joining as an investor nine years after pushing Kalanick out. Per TechCrunch, the round did not disclose a valuation. Atoms merges CloudKitchens and the automation firm Pronto into three divisions, food, mining, and transport, all aimed at automating heavy industry.

Spence’s take: When a16z writes a check this size before there’s a disclosed valuation, they aren’t buying a product. They’re buying a track record.

Here’s the model I keep coming back to on this one: Circle of Competence. Kalanick has a real, demonstrated edge in one specific thing, taking a messy physical industry and wrapping software around it. He did it to transportation at Uber and to food logistics at CloudKitchens, and the bet is that the same competence carries into mining and heavy transport. The contrarian read: circles of competence are narrower than the decks that sell them. “Digitize the physical world” sounds like one skill, but it might be three brutally hard problems wearing a single slogan. Proven edge in food delivery doesn’t automatically transfer to autonomous mining, so watch whether Atoms ships in one vertical before you believe it ships in all three.

2. Etched hit a $10.3 billion valuation for a chip that runs one thing

AI chip startup Etched closed a $300 million Series C led by Sequoia, with the company reporting a post-money valuation of roughly $10.3 billion. That’s double its December figure in about seven months, and the company says it’s the highest valuation Sequoia has ever led at Series C. The catch: Etched’s Sohu chip runs only transformer models, and the company reports more than $1 billion in booked orders. (A private valuation is a point-in-time snapshot set by investors, not a market price.)

Spence’s take: Betting everything on one architecture is either the sharpest move in the room or the most fragile. There’s no middle.

3. Antares raised $470 million to put reactors on military bases

Nuclear startup Antares closed a $470 million Series C, $370 million in equity and $100 million in debt, co-led by Paradigm and Caffeinated Capital, with Point72 Ventures participating. The company builds small modular reactors and, per its announcement, plans deployments to U.S. military installations by 2028. Per TechCrunch, investor appetite for advanced nuclear is being driven by the AI data center building boom and its enormous power demands.

Spence’s take: The AI story isn’t just chips and models. It’s who generates the electricity to run them.

4. Meshy raised nearly $400 million to turn text into 3D

Meshy closed a roughly $400 million Series B at a $1.5 billion valuation, which the company says is its first publicly disclosed valuation and the largest round yet in AI 3D generation. It turns a text prompt or a single image into a usable, printable 3D model in about a minute. The company reports annual recurring revenue growing about 12x year over year with more than 12 million registered users. (Company-reported metrics; the valuation is a time-sensitive snapshot.)

Spence’s take: Everyone funded text and image models. The teams quietly winning built for the dimension nobody was looking at.

5. Augustus became a $1 billion clearing bank with an actual charter

Augustus raised a $180 million Series B at a $1 billion valuation, led by Tiger Global. The interesting part isn’t the round. It’s that Augustus received conditional approval from the OCC for a U.S. national bank charter, one of only a handful granted since 2010 (conditional approval is not final authorization). The company provides dollar access to international fintechs and says it already processes billions for customers including the crypto exchange Kraken.

Spence’s take: In regulated markets, the license is the moat. Anyone can write code. Almost nobody can get the charter.

That last one is close to home. At /mkt, building in a regulated corner of finance, I’ve learned the hardest part isn’t the technology. It’s earning the right to operate inside the rules. Augustus is a reminder that regulatory approval, slow and painful as it is, can end up being the most defensible asset a company owns.

Next week

The AI power crunch gets the full deep-dive. Antares is one bet. Etched is another. I’m pulling the thread on why the two hottest categories in venture right now, silicon and electricity, are really the same story, and what that means for anyone building on top of AI. Paid subscribers get the whole breakdown.

If this was useful, share it with someone who builds things. And if you want the full toolkit of 50 mental models, you can grab my book, Mental Models: How to Think, Act, and Win, on Amazon right now.

Spence

This post is for informational purposes only and is not investment advice, a recommendation, or a solicitation to buy or sell any security or asset. Funding figures, valuations, and company metrics are as reported by the companies, their investors, or the cited outlets; private-company valuations are point-in-time snapshots that do not reflect public-market prices or predict future performance. References to digital assets and platforms that handle them are informational only; digital assets are volatile and carry risk of loss. Spencer Gareiss is CPO at /mkt; any mention of /mkt is illustrative and not a solicitation. Always do your own research and consult a licensed professional before making financial decisions.

If you want the mental models behind breakdowns like this, my book, Mental Models: How to Think, Act, and Win, is on Amazon now.

Get it Here

Share Startup Spotlight

Everyone’s pouring money into smarter AI. Almost nobody’s asking where the electricity comes from. This week, one company raised $470 million on exactly that question.

Antares Nuclear closed a $470 million Series C, co-led by Paradigm and Caffeinated Capital, with Point72 Ventures, Shine Capital, and Industrious Ventures joining. The round breaks down to $370 million in equity and $100 million in debt, and it brings the three-year-old company’s total to $604 million. Not bad for a startup that closed a $96 million Series B just seven months ago.

Antares builds microreactors. Small modular nuclear units that put out between 100 kilowatts and 1 megawatt, enough to power up to 750 homes, per TechCrunch. The company says its Mark-0 reactor reached criticality earlier this year under the Department of Energy’s reactor pilot program, in partnership with Idaho National Laboratory. Paradigm called it the first private advanced reactor criticality in decades. Antares is one of three finalists in the Pentagon’s advanced nuclear program, and it’s aiming to put working reactors on U.S. military bases by 2028.

Here’s the model I keep coming back to: the Theory of Constraints. Every system has one bottleneck that caps the whole thing, and improving anything except that bottleneck is wasted motion. You can add lanes to a highway all day. If one toll booth is jammed, traffic still doesn’t move.

Right now the AI story is all about the model layer and the chips. But every one of those chips needs power, and the grid wasn’t built for this. Data centers running AI workloads are pushing electricity demand up faster than utilities can add supply. That’s the toll booth. All the GPUs in the world don’t matter if you can’t plug them in. Investors are starting to see it, which is why $470 million just went to a company that makes electrons instead of tokens. It’s the same pattern I see building in regulated markets at /mkt. The flashy front end gets the attention. The boring infrastructure underneath decides whether anything actually ships.

Now the contrarian part. The obvious read is “AI needs power, so power is the trade.” Fine. But look at where this money’s going. Not to the grid. Not to utility-scale solar. To reactors small enough to sit on a single military base, sized for one customer at a time. That’s a bet that the winning move isn’t fixing the grid. It’s routing around it.

If that’s right, the next decade of energy doesn’t look like bigger plants and longer transmission lines. It looks like power generated right next to where it’s used, one site at a time. Centralized systems getting quietly replaced by distributed ones, the same way mainframes gave way to personal computers. The constraint doesn’t just get relieved. It gets designed out.

I built financial infrastructure for a living, and the lesson holds everywhere. Whoever owns the bottleneck owns the market. Right now, the bottleneck is a wall socket.

If this was useful, share it with someone who builds things. And if you want the full toolkit of 50 mental models, you can grab my book, Mental Models: How to Think, Act, and Win, on Amazon right now.

This post is for informational and educational purposes only. It is not investment advice, nor an offer or solicitation to buy or sell any security. Funding figures and technical milestones are as reported by Antares, its investors, and cited outlets, and have not been independently verified. Spencer Gareiss is Chief Product Officer at /mkt; the reference to /mkt is illustrative only and is not a solicitation. Views are his own.

If you want the mental models behind breakdowns like this, my book, Mental Models: How to Think, Act, and Win, is on Amazon now.

Get it Here

Share Startup Spotlight