Firms like OpenAI and Meta claim a compute shortage while also exploring selling compute capacity. This isn't a contradiction but a strategic evolution. They are buying all available supply to secure their own needs and then arbitraging the excess, effectively becoming smaller-scale cloud providers for AI.

AI companies with the foresight to sign long-term, multi-year compute contracts gain a significant margin advantage. They lock in prices based on past valuations, while competitors are forced to buy capacity at much higher current market rates driven up by the increasing value of new AI models.

In a significant strategic misstep, Google sold a large volume of its custom TPU accelerators to rival Anthropic. Immediately after, demand for Google's own Gemini model surged, leaving Google compute-constrained and trying to secure more capacity from a sold-out TSMC.

Top AI labs like OpenAI and Anthropic engage in a 'Cournot Equilibrium' by competing on the supply of compute and data centers, not by undercutting each other on price. This strategy aims to create high barriers to entry and maintain high prices for access to frontier models.

Dario Amodei highlights the extreme financial risk in scaling AI. If Anthropic were to purchase compute assuming a continued 10x revenue growth, a delay of just one year in market adoption would be "ruinous." This risk forces a more conservative compute scaling strategy than their optimistic technical timelines might suggest.

For leading AI labs like Anthropic and OpenAI, the primary value from cloud partnerships isn't a sales channel but guaranteed access to scarce compute and GPUs. This turns negotiations into a complex, symbiotic bundle covering hardware access, cloud credits, and revenue sharing, where hardware is the most critical component.

OpenAI's aggressive partnerships for compute are designed to achieve "escape velocity." By locking up supply and talent, they are creating a capital barrier so high (~$150B in CapEx by 2030) that it becomes nearly impossible for any entity besides the largest hyperscalers to compete at scale.

Despite a $380 billion valuation, Anthropic's CEO admits that a single year of overinvesting in compute could lead to bankruptcy. This capital-intensive fragility is a significant, underpriced risk not present in traditional software giants at a similar scale.

Dario Amodei reveals a peculiar dynamic: profitability at a frontier AI lab is not a sign of mature business strategy. Instead, it's often the result of underestimating future demand when making massive, long-term compute purchases. Overestimating demand, conversely, leads to financial losses but more available research capacity.