Anthropic pays xAI $1.25 billion monthly for compute access in massive AI infrastructure deal
May 20, 2026
Sam Altman made an audacious pitch to Y Combinator’s current startup class this week, offering $2 million worth of OpenAI tokens to every company in exchange for equity. The announcement came during a Tuesday night YC event in what partner Tyler Bosmeny called a “mic drop moment.”
The deal would see OpenAI invest in all 169 startups in the current cohort, not with cash but with AI tokens that companies can use to build their products. It’s an unprecedented move that highlights how AI infrastructure has become both a massive cost center for startups and a strategic battleground for tech giants.
The equity terms won’t be determined upfront. Instead, OpenAI is offering an “uncapped SAFE” – Y Combinator’s standard early-stage investment structure – that will convert when startups raise their first priced funding round, typically a Series A. The higher the startup’s valuation at that point, the smaller the equity slice OpenAI receives.
Industry observers speculate this could amount to roughly 2% equity if a startup hits a $100 million valuation, though the actual terms haven’t been disclosed publicly. YC managing director Jared Friedman confirmed the structure to TechCrunch, noting the conversion will happen “in the next priced round.”
For OpenAI, the strategy works on multiple levels. Beyond gaining equity stakes in promising early-stage companies, it encourages startups to build their businesses on OpenAI’s platform rather than competitors like Anthropic’s Claude. While this doesn’t necessarily lock companies in long-term, it certainly steers them away from rival AI providers during their formative stages.
The economics also favor OpenAI as inference costs continue dropping. What the company gives away in tokens today may cost very little to produce tomorrow, making the equity it receives in return increasingly valuable.
The startup community has already split into camps over whether this represents a good deal:
- Supporters argue it helps eliminate AI infrastructure bills, which can quickly spiral and consume disproportionate shares of early-stage budgets when cash is already scarce
- Skeptics worry about platform dependency and the risk that OpenAI could study startup ideas and incorporate successful features into its own offerings
- Others question whether token budgets from a single AI provider justify giving up additional equity
Seed investor Jason Calacanis, who runs his own competing accelerator, warned founders about the “classic platform playbook.” He posted on X: “If you take these tokens, there’s a non-zero chance that OpenAI will study exactly what your startup is doing, copy your idea and put your app into their free offering.”
The concern about AI giants absorbing startup innovations is real, but the counterargument is equally valid. OpenAI could monitor and potentially copy ideas from paying customers just as easily. Taking an equity stake might actually give the company more incentive to see startups succeed rather than cannibalize them.
The bigger strategic question centers on equity allocation. Y Combinator already takes a 7% stake for its $500,000 investment and access to its Silicon Valley network. Seed investors typically claim another 20% or so. Early employees also need equity compensation. For cash-strapped startups, surrendering additional equity for what amounts to operational credits represents a significant trade-off.
The real risk lies in startups burning through their OpenAI token budgets without sufficient progress to show for it, having given up equity in the process. Still, for many early-stage companies, trading future equity for immediate access to AI infrastructure may prove more sustainable than paying cash for those same tokens.
As Altman’s former role as Y Combinator’s president demonstrates, he already has extensive access to each cohort and their ideas regardless of any investment deal. The token-for-equity offer simply formalizes a relationship that could benefit both sides – if startups can resist the temptation to over-rely on a single AI platform as they scale.
