OpenAI is shifting its business model by experimenting with equity-for-compute deals, a practice dubbed tokenmaxxing. Instead of charging cash for API tokens, the company provides large compute grants, such as $2 million in tokens, to early-stage startups in exchange for equity stakes.
This strategy follows a series of mathematical breakthroughs by OpenAI. On September 8, 2026, the company proposed a partial solution to the Navier-Stokes existence and smoothness problem using 10,000 agents and 2.7 million messages over 88 hours. In August 2026, it solved 10 mathematical problems for $2,000 in compute.
On October 6, 2026, OpenAI released 722 mathematical manuscripts covering 372 result families in fields like algebra and topology. Each result required approximately three hours of compute, equivalent to ChatGPT Pro processing. These rapid publications have sparked academic disputes regarding credit and research priority.
The shift away from metered token sales suggests that AI labs may view selling compute as underpricing their models. By taking equity, labs gain a share of the value created by startups. However, this model creates risks for founders, as their compute supplier may simultaneously become a direct competitor.
