Investor pressure forces Nvidia to shrink its OpenAI bet just as Anthropic's numbers defy bubble warnings

Nvidia has significantly reduced its financial commitment to OpenAI's Ohio data center project following investor pressure about capital allocation risk. The chip maker cut its guarantee from $250 billion to just under $120 billion, a nearly 50 percent reduction that signals growing concern among shareholders about the pace and scale of AI infrastructure spending.

The cut matters because it reveals a fundamental tension in the AI boom. Nvidia has committed roughly $110 billion in GPU purchases and infrastructure support to OpenAI's "Stargate" project, which aims to build massive compute facilities. Investors questioned whether such concentrated bets on a single company made sense given execution risks, timeline uncertainties, and the unproven business models of frontier AI labs.

OpenAI and Microsoft, which is backing the project with capital, now face a smaller financial cushion from Nvidia. The company originally pledged to guarantee a portion of costs to reduce perceived risk. Halving that guarantee shifts more burden onto the founders and their financial partners to deliver on promises and manage cost overruns.

The move contradicts the narrative that investors have abandoned AI spending entirely. Rather, it reflects a shift toward discipline and risk management. Limited partners and fund managers increasingly ask whether AI companies can generate returns that justify valuations and capital intensity. A $250 billion commitment to one project, with no clear path to profitability, became harder to justify as fiduciary duty.

Anthropic, meanwhile, is complicating the "AI bubble" thesis by posting extraordinary revenue growth. The company reported revenue jumped from $4.7 billion to $11.5 billion in a single quarter, nearly 2.5 times the prior level. That trajectory suggests at least one major AI lab is finding paying customers willing to commit significant budgets to its Claude models and services.

Anthropic's performance matters because it validates the demand side of the equation. While Nvidia's retreat suggests caution on infrastructure spending, Anthropic's numbers show enterprises and developers are actually purchasing access to frontier AI capabilities. That creates a rare data point: a large AI company demonstrating revenue acceleration, not just user growth or hype.

The contrast between Nvidia's pullback and Anthropic's revenue jump frames the real debate. The question is not whether AI has value. Anthropic's numbers confirm it does. The question is whether the infrastructure spending required to build and train next-generation models can sustain returns on the scale investors expect. A $250 billion data center bet requires sustained compute demand and pricing power that remain unproven.

OpenAI, Google, Meta, and others continue building massive facilities because they believe frontier AI capabilities require scale. The reduced Nvidia guarantee does not kill that strategy. It simply makes the path to profitability more visible to investors. OpenAI now owns more of the risk and reward.

What happens next depends on whether labs like Anthropic and OpenAI can convert their technical capabilities into durable business models. Anthropic's revenue numbers suggest the answer is yes, at least for near term. Whether that scales to justify hundreds of billions in compute spending remains the unresolved question driving both investor caution and continued capital deployment.