NVIDIA Is About to Write OpenAI a $30 Billion Check. Here's What It Really Means.

Five months ago, NVIDIA and OpenAI announced a $100 billion partnership to much fanfare. It was supposed to be a landmark moment -- the world's most valuable chipmaker pouring an unprecedented sum into the company behind ChatGPT. Markets loved it. NVIDIA's stock surged past $5 trillion. That deal is dead. And the one replacing it tells a very different story about where the AI industry is actually headed. ## The New Deal NVIDIA is in the final stages of a $30 billion equity investment in OpenAI, according to the Financial Times. The deal could close as early as this weekend. It's part of a broader funding round that's expected to raise more than $100 billion from a consortium that includes Amazon, SoftBank, and Microsoft, valuing OpenAI at roughly $730 billion before the new money flows in. That would make OpenAI the second most valuable private company on the planet, trailing only SpaceX. But here's what matters: this isn't the same arrangement they announced in September. The original $100 billion commitment was structured as a "letter of intent" -- NVIDIA would invest money, and OpenAI would turn around and spend it buying NVIDIA hardware. It was, as several analysts pointed out at the time, a circular deal. NVIDIA was essentially financing its own sales. The new $30 billion investment is a straight equity stake. NVIDIA gets OpenAI stock. There's no attached requirement for OpenAI to buy NVIDIA chips with the proceeds. That's a significant distinction, and it changes the relationship between these two companies in ways that won't be obvious from the headline number alone. ## What Happened to the $100 Billion? The original deal started falling apart in early February. A Wall Street Journal report revealed that NVIDIA CEO Jensen Huang had privately expressed doubts about the transaction and criticized what he called a "lack of discipline" in OpenAI's business approach. He also flagged concerns about growing competition from Google and Anthropic. Huang later called those claims "nonsense," but NVIDIA's stock still dipped on the news. The truth is probably somewhere in between. The $100 billion commitment was always more aspirational than contractual -- a letter of intent, not a binding agreement. When market sentiment shifted and tech stocks dropped 17% from their January peaks, the calculus changed. Investors started asking harder questions about circular funding in AI, and the optics of NVIDIA bankrolling its own biggest customer became harder to defend. Sarah Kunst, managing director at Cleo Capital, noted the original announcement was deliberately vague. "There wasn't a strong 'It will be $100 billion,'" she told CNBC. "It was, 'It will be big. It will be our biggest investment ever.' And so I do think there are some question marks there." Fair point. ## NVIDIA's Investment Playbook The $30 billion deal isn't happening in a vacuum. NVIDIA has spent the last two years building an investment portfolio that looks less like a chipmaker's balance sheet and more like a venture fund with a very specific thesis: back the companies that will buy your GPUs. This strategy has drawn sharp criticism. Tech critic Ed Zitron has argued that NVIDIA "seeds companies and gives them the guaranteed contracts necessary to raise debt to buy GPUs from NVIDIA, even though these companies are horribly unprofitable and will eventually die from a lack of any real demand." That's a harsh read, but it's not entirely wrong. The circularity problem in AI funding is real. When the biggest chip supplier is also the biggest investor in chip buyers, it creates a feedback loop that can inflate demand signals and mask the true economics of the industry. It's the kind of thing that looks brilliant during a boom and catastrophic during a bust. The shift to a clean equity stake in OpenAI suggests NVIDIA is aware of this criticism and is trying to put some distance between its investment strategy and its sales pipeline. Whether the market buys that framing is another question. ## OpenAI Is Hedging Its Bets The other side of this story is equally telling. OpenAI isn't sitting around waiting for NVIDIA to dictate terms anymore. In October, OpenAI signed a deal with AMD for 6 gigawatts of GPU capacity. In January, it struck a $10 billion agreement with Cerebras for dedicated low-latency inference computing through 2028. It's also working with Broadcom to develop a custom AI chip that could eventually reduce its dependence on NVIDIA altogether -- though Broadcom's CEO, Hock Tan, told investors in December that the company "does not expect much in 2026" from that partnership. OpenAI even explored deals with inference chip startups Groq and Cerebras. NVIDIA effectively killed the Groq option by acquiring the company in a $20 billion licensing deal last December, hiring its founder and CEO in the process. That's the kind of move that tells you exactly how seriously NVIDIA takes the threat of customers walking away. ## What This Means for AMD and Intel If you're AMD, this deal is complicated. On one hand, OpenAI diversifying away from NVIDIA is good news -- it validates AMD's MI300 series as a real alternative and suggests the monopoly grip is loosening. On the other hand, a $30 billion NVIDIA investment in OpenAI means there's still enormous financial gravity pulling the two companies together, even without explicit chip purchase requirements. For Intel, the picture is bleaker. The company has been largely absent from the AI training arms race, and its Gaudi accelerators haven't gained meaningful traction against NVIDIA's H100 and Blackwell architectures. OpenAI hired Sachin Katti away from Intel in November to lead its compute infrastructure team -- which says something about where the talent sees the future. NVIDIA still commands roughly 85% of the GPU market for AI workloads. That number will come down as custom silicon and alternative architectures mature, but it won't happen quickly. Training large language models still overwhelmingly runs on NVIDIA hardware, and the CUDA software ecosystem creates switching costs that money alone can't overcome. ## The Bigger Picture This deal reveals an AI industry in transition. The era of blank-check hype spending is winding down. OpenAI's market share for consumer AI has dropped from 86.7% to 64.5% over the past year, with Anthropic gaining ground in enterprise. The company has started testing ads in ChatGPT -- a move that drew pointed criticism from Anthropic and raised questions about OpenAI's path to profitability. NVIDIA moving from a $100 billion circular commitment to a $30 billion equity bet is a sign that even the most bullish players in AI are starting to demand clearer returns. The investment is still massive. But it's grounded in ownership, not sales guarantees. That's a more honest arrangement, and probably a more stable one. Whether $730 billion is the right valuation for a company that's burning cash and losing market share -- well, that's the $30 billion question.