Five Days

The trial ended Monday. The jury took less than two hours.

While everyone was watching an Oakland courthouse, OpenAI spent five days quietly building infrastructure. Not the compute kind. The capture kind.

Three announcements. Five days.

On Friday, OpenAI launched a personal finance tool for ChatGPT Pro subscribers. Connect your accounts through Plaid. See your balances, transactions, investments, subscriptions, and upcoming payments in one dashboard. Ask questions. Get answers grounded in your actual financial history.

Twelve thousand financial institutions. Chase. Fidelity. Schwab. Robinhood. American Express. Capital One. Intuit coming soon — which is the one that will let ChatGPT estimate the tax impact of selling a stock or calculate your odds of getting approved for a specific credit card.

Two hundred million people already ask ChatGPT financial questions every month, OpenAI noted. They're not building this for a niche audience.

The technical layer is Plaid, which has been connecting fintech apps to bank accounts for a decade. You've used it before. Probably without thinking much about it.

The data OpenAI can access: balances, transactions, investments, liabilities. What it cannot access: full account numbers, transfer mechanisms. Read-only. It cannot move your money.

If I gambled, I'd imagine the prediction markets would be laying odds on them wanting to add that eventually.

What happens when you disconnect: OpenAI says they remove synced data from ChatGPT within 30 days. "Remove." I'm sure that's fine.

OpenAI built the Hiro team's personal finance expertise in-house when they acquired the Hiro startup in April. This wasn't a weekend project. It's been in development. It launched four days before the verdict.

The company that had two sets of revenue numbers for different audiences. The company whose CFO went to the Wall Street Journal saying it wasn't ready for public reporting standards. The company whose CEO and president held undisclosed stakes in a vendor they signed a $10 billion deal with.

That company now has a direct line to your transaction history.

Read the terms. They're more honest than the marketing.

The day after the verdict, OpenAI announced Guaranteed Capacity. They didn't wait 24 hours.

The structure: enterprise customers can commit to one, two, or three-year contracts for guaranteed access to OpenAI compute. Longer commitments get steeper discounts. The capacity can be drawn down across OpenAI's entire product portfolio — one contract, multiple endpoints, multiple workflows.

Sam Altman framed it as meeting customer demand. "Customers are increasingly asking us for certainty on capacity. As models get better, we expect that the world will be capacity-constrained for some time."

Worth keeping in mind: yes, compute resources are real and finite. Also worth keeping in mind: OpenAI sets the caps. They control the limits. The scarcity they're selling you certainty against is scarcity they define. If that sounds familiar, it should — it's the same logic as old-school broadband caps or, if you're old enough to remember, the minutes on your cell phone plan. The carrier created the constraint and then sold you a premium tier to worry less about it.

The translation for the IPO story: multi-year committed revenue is a fundamentally different asset than month-to-month subscriptions. It's predictable. It's auditable. It's the kind of thing that makes an S-1 look considerably cleaner to institutional investors who lived through the WeWork roadshow.

Enterprise already makes up more than 40% of OpenAI's revenue. It's on track to reach parity with consumer by the end of 2026. Turning that into locked three-year contracts before the IPO roadshow is not a coincidence.

The Deployment Company — the PE-backed captive enterprise demand mechanism documented here — gets you in the door. Guaranteed Capacity locks you in for three years. Available, per Altman, until the current allocation sells out.

He's not wrong. He's also marketing a multi-year enterprise infrastructure contract like it's the McRib — limited time, act now, and yes, they plan to bring it back after the allocation sells out. Exactly what McDonald's does. FOMO as a procurement strategy.

Tuesday night Sam Altman walked into a Y Combinator event and offered every startup in the current batch $2 million worth of OpenAI tokens in exchange for equity.

169 companies. Every single one.

Let's be clear about what's actually happening here. He's paying with something he has unlimited access to, at a cost he controls, with limits he sets — for everyone else. Tokens aren't scarce to OpenAI. OpenAI decides what tokens cost. OpenAI decides how many you get. The founders are receiving something that is genuinely valuable to them and costs Altman approximately nothing to conjure. In exchange they're handing over a stake in something real.

The deal structure: an uncapped SAFE, converting at the next priced round — typically the Series A. No ceiling on valuation, meaning the higher the Series A, the smaller OpenAI's slice. Numbers floating around on X suggest roughly 2% equity at a $100M valuation, though the actual terms aren't public.

YC already takes 7% for $500,000 in cash. Now OpenAI wants a piece too. In exchange for tokens.

That's not an investment. That's the cheapest equity acquisition strategy in Silicon Valley history dressed up as one.

Jason Calacanis — who runs a competing accelerator, conflict noted — posted the platform playbook warning: OpenAI could study what each startup is building, absorb the good ideas, and ship them as features. The fear is documented and real.

The lock-in is the quieter story. Every one of these 169 companies is now building on OpenAI. Their codebases will call OpenAI APIs. Their teams will get comfortable with OpenAI's tooling. When the free tokens run out, switching has a cost. And by the time they're paying customers, they've already built their product around the platform.

That's how you capture an ecosystem at its most vulnerable.

The Deployment Company gets enterprise customers in the door. Guaranteed Capacity locks them in for three years. Plaid connects consumer users' bank accounts. Tokens for equity locks 169 early-stage startups onto the platform before they've ever had the chance to choose otherwise.

Five days. Consumer capture. Enterprise lock-in. Startup ecosystem capture.

The trial ending was real news. But OpenAI didn't take the day off.

I've said it throughout this series. The product works. The court case is over. The Oakland record still exists.

What's happening now is the IPO story being written in real time. The valuation needs a narrative. The narrative needs locked revenue, captive users, and an ecosystem of early-stage companies building on your platform.

Five days. Three mechanisms.