In May 2026, ahead of Y Combinator's Demo Day, the same offer landed in front of roughly 180 startup founders. OpenAI CEO Sam Altman delivered it himself. Instead of a cash investment, the company would pay in OpenAI API credits—tokens—and in return it would take equity in each startup. Every company in the batch, the same terms across the board.

TechCrunch called the scene a "mic drop moment." Like setting down the microphone and walking offstage, the structure made acceptance feel more natural than pushback. Among YC founders, the offer reportedly sparked days of discussion. For an early-stage startup, the chance to focus on building the product without worrying about infrastructure costs was the kind of proposal you had to actively prepare an argument against in order to turn down.

If there were a reason to say no, what would it be?

The Moment Investment Became Compute

The structure of the deal is simple. OpenAI pays in API credits rather than cash, and the startup books that as investment and hands over equity. The form of the investment shifts from cash to computing resources.

The approach itself isn't new. AWS has offered cloud credits to early-stage startups for years, and Google and Microsoft run similar programs. Counting services rather than cash as a form of investment is a strategy Silicon Valley has already road-tested.

But this case differs from the precedent in a few ways.

The scale is the first thing that stands out: a single set of terms offered to an entire YC batch at once. With one announcement, OpenAI created a structure that could establish equity ties with 180 startups simultaneously. This isn't an investor assembling a portfolio one company at a time—it's binding an entire ecosystem in a single move.

The timing is worth examining too. For a startup trying to build an AI product, early infrastructure costs are one of the biggest barriers. The more a product leans on LLM APIs, the faster its early operating costs climb. Any team that has lived through the pattern—costs surging before any real traction takes hold—understands just how persuasive the timing of this offer is. What position you hold in a negotiation when you show up with the solution at the exact moment the problem looks largest needs no further explanation.

Above all, what makes this deal hard to read as a simple investment is who is making it. The CEO of a company that supplies AI infrastructure stepped forward personally to propose swapping his own credits for startup equity. It's a scene where the language of an investor and the language of platform expansion merge in the same sentence. A structure like this becomes possible only when the company in question occupies the position of supplying essential infrastructure to the entire industry.

The Reasons to Be Skeptical

The view that this deal shouldn't be read only in a positive light surfaced immediately after the announcement.

The most direct criticism concerns dependency. A startup that designs its early service architecture on top of the OpenAI API faces substantial costs later if it wants to switch to Anthropic, Google DeepMind, or an open-source model. Rewriting the codebase, redesigning prompt architecture, changing how responses are handled, retraining the team—all of these costs grow the more tightly the system is tuned to one provider. The industry calls this switching cost. You start with free infrastructure, but in doing so you build, early and by your own hand, a structure you can't walk away from once a better alternative appears.

The asymmetry in valuation also draws attention. In this deal, OpenAI sets the unit price of the credits, while the market sets the future value of the startup's equity. One side's present value is fixed; the other's future value is undetermined. Anyone who reads deal structures will quickly spot which side holds the advantageous position in that asymmetry.

There's another angle. API credits are a consumable resource, and their use is confined to that one platform. Cash can be spent freely on hiring, marketing, office space—whatever the goal calls for; credits cannot. That's why some read this as trading away equity, an asset with no restrictions on use, in exchange for a resource whose purpose is fixed.

Of course, not every founder who accepted the offer signed without analysis. For a startup, the option of easing the burden of early infrastructure costs and concentrating on the product can be entirely rational. What matters, in my view, is whether they set the criteria for that calculation themselves, or whether they were simply persuaded by the appeal of the offer. The two paths can look identical in outcome, yet they produce entirely different positions at the negotiating table later on.

The Same Structure Has Already Reached Korean Founders and Operators

This episode belongs to the American YC ecosystem. But for the solo founder in Korea, the solo PM, the operator rolling out AI tools, the same structure has already arrived in a different shape.

AI SaaS that lures you in with a free plan, cloud partnerships offering early credits at no charge, contracts that bring you in at a pilot price and adjust the rate later, the process of building a service with a particular AI provider's API as the default layer—all of these are incentives that form a relationship of dependency with a specific platform early on. They simply don't ask for equity; the logic of the deal isn't much different.

Have you ever calculated the switching cost of the AI service you're using now? Redesigning prompts, reconfiguring workflows, retraining the team, converting data formats—add up these costs and you'll find that more than a few organizations have already entered a substantial relationship of dependency. The higher the switching cost climbs, the less room you have to respond to subscription price hikes, changes in terms, or shifts in service conditions.

It's also worth looking back at the point where the tool you're using now crossed over from "a tool you can choose" into "infrastructure that's hard to replace." A tool is something you can compare and select; infrastructure is the state in which the cost of replacement has grown larger than the cost of keeping it. Recognizing that boundary is itself the first step toward preserving room to negotiate in the future.

The principle that, when evaluating a deal, you should define the terms you can live with before assessing the other party's offer is one that recurs throughout the practice of sales and negotiation. The habit of moving aggressively while reading the deal structure first is what lets you keep your judgment intact even in the face of an attractive offer.

There's no way to know what kind of negotiating table the YC startups who accepted Altman's offer will be sitting at years from now. But the leverage they'll bring to that table depends, in large part, on how concretely they thought it through before accepting this offer.

Receiving AI without cash is settling in as a new language of investment. Without the eyes to read that language, the most attractive offer can end as the most expensive deal.