Sam Altman Pledges Eternal Loyalty to NVIDIA (While Building His Own Fabs)

Sam Altman has abruptly softened his tone — or, more precisely, moved to clear away the thick fog of rumors that something had soured between OpenAI and NVIDIA. In a recent post, he heaped praise on Jensen Huang's company, calling its products the best AI chips in the world. Given the endless chatter that OpenAI was supposedly unhappy about relying on a single supplier and was preparing an NVIDIA "killer" codenamed Tigris, the statement looks like a masterclass in corporate diplomacy.

Why now? All last year, the industry whispered that Altman chafed under NVIDIA's monopoly. When you plan to train next-generation models that will demand an order of magnitude more compute than today's, you need hundreds of thousands of accelerators. Waiting on H100 shipments became the biggest bottleneck for the entire industry, and OpenAI was no exception. It made sense for Altman to look for an exit, flying around the world to raise trillions of dollars for his own fabs. But building your own infrastructure from scratch takes more than hiring brilliant engineers: it means years of testing, failures, and billions in burned capital.

Altman said outright that OpenAI plans to remain a "super-customer" of NVIDIA for a very long time. That sounds like a sober admission of surrender to current reality: right now, there simply are no adequate alternatives. Even if custom silicon from Google or Amazon performs well on narrow tasks, NVIDIA holds the market thanks to its CUDA ecosystem — the software layer that has become the de facto standard for every developer. Switching to other hardware today would mean rewriting massive amounts of code and losing ground in the race against Anthropic and Google, which OpenAI cannot afford.

This game of thrones in the silicon world is fascinating to watch. On one hand, we hear about OpenAI partnering with Broadcom and reserving capacity at TSMC fabs. On the other, Altman publicly declares his love for the reigning monopolist. It's a classic diversification strategy in the spirit of "keep your friends close and your enemies closer." While OpenAI builds its backup runway, it badly needs Huang to keep shipping new Blackwell chips on a priority basis. Any hint of conflict could suddenly make OpenAI's place in line for scarce hardware a lot longer.

The upshot: Altman's words are not mere politeness but a calculated business move. NVIDIA benefits from having an ambassador who confirms its dominance at the highest level, and OpenAI benefits from keeping its status as a favorite customer. While the world waits for a real NVIDIA competitor to emerge, the Altman–Huang duo keeps setting the rules — even if, behind closed doors, each is already drafting a plan for solo world domination. In this industry, love lasts exactly until a more efficient process node comes along.

The bottom line: OpenAI's own chips are a dream of the future, while NVIDIA is a hard necessity of the present. Will Altman ever truly kick his Jensen Huang habit, or will this dependence prove fatal for the company?