‘Google news’ that reportedly made Nvidia CEO Jensen Huang call OpenAI CEO Sam Altman

A single news report about OpenAI testing Google's AI chips sent Nvidia CEO Jensen Huang scrambling to his phone in June, triggering an urgent call to OpenAI CEO Sam Altman that would reshape their partnership and culminate in one of the tech industry's largest infrastructure deals.

The episode exposed Nvidia's vulnerability despite its dominance in AI chips, revealing how even a hint of competition from Google could rattle the chipmaker's relationship with its most high-profile customer.

What followed was a dealmaking blitz that would bind the two companies together with up to $100 billion in commitments—and unprecedented financial risk.The June 27 article by The Information, reporting that OpenAI had begun renting Google's tensor processing units (TPUs) to power ChatGPT, caused "something of a stir" inside Nvidia's Santa Clara headquarters, according to The Wall Street Journal.

Huang immediately reached out to Altman, asking if the report was true and signaling his willingness to revive stalled partnership negotiations.

Nvidia's fear of losing its biggest AI customer

The urgency of Huang's response reflected the high stakes involved. OpenAI is Nvidia's most prominent AI customer, and the prospect of losing ground to Google represented a significant competitive threat. During the call, Huang realized that Nvidia could strengthen its position by making a direct investment in OpenAI, a person who spoke to him told the Journal.

OpenAI quickly moved to clarify its position. A spokesperson said the company had no active plans to use Google's TPUs at scale, though it acknowledged early testing with the chips. Nvidia's official X account then amplified an article in which OpenAI denied plans to deploy Google's in-house chip.

One phone call leads to $100 billion deal

The conversation catalysed negotiations that had previously stalled, ultimately producing a landmark $100 billion agreement announced in September.

The deal commits Nvidia to investing progressively in OpenAI as the AI startup deploys at least 10 gigawatts of Nvidia systems—representing up to 5 million GPUs—for next-generation infrastructure. By current standards, the chip lease arrangement could cost OpenAI $350 billion, the Journal calculated.The partnership includes an extraordinary provision previously unreported: Nvidia is discussing guaranteeing loans OpenAI plans to secure for building its own data centers.

This arrangement could saddle the chip giant with billions in debt obligations if OpenAI fails to repay, according to the Financial Times.

OpenAI's diversification strategy continues

Despite the Nvidia deal, OpenAI has continued pursuing its multi-supplier strategy. The startup announced a 6-gigawatt partnership with AMD in October, with AMD offering warrants for up to 10 percent of its stock as incentive. OpenAI also unveiled a 10-gigawatt deal with Broadcom for custom chip development, matching Nvidia's commitment.Meanwhile, the Google TPU report ultimately proved prescient for the broader market. In late October, AI startup Anthropic—founded by former OpenAI leaders—announced a multibillion-dollar deal with Google for access to up to 1 million TPU chips, Bloomberg reported, validating Google's chip technology as a credible alternative to Nvidia's dominance.