OpenAI plans to buy Google’s AI chips outright, according to Reuters, which reported the plan on Tuesday citing The Information and a person familiar with the matter. Neither company commented, and no financial terms were disclosed. But the direction of travel is unmistakable: the ChatGPT maker is moving from borrowing its biggest rival’s silicon to owning it.
The Rivalry That Turned Into a Supply Chain
OpenAI and Google compete head-on in AI products and search. Yet commercially, they keep getting closer. OpenAI is already one of the largest buyers of Nvidia’s graphics processing units, the chips it uses to train models and run inference. Now it wants Google’s Tensor Processing Units, or TPUs, on its own balance sheet rather than only through a cloud rental agreement.
TPUs have gained traction as an alternative to Nvidia’s market-leading chips as companies hunt for cheaper options and silicon better suited to running models at scale. Google only began selling TPUs directly to some customers this year.
From Renter to Buyer in 16 Months
The timeline is what makes this more than a procurement story. Reuters reported in June 2025 that OpenAI had begun renting Google’s AI chips. Last week, OpenAI said it would make its frontier models available through Google Cloud, letting customers buy them through the Google Cloud Marketplace. Now, reportedly, comes the third step: buying the hardware itself.
Each step has lowered the wall between the two companies. Renting proved the chips worked for OpenAI’s workloads. Selling models on Google Cloud made Google a distributor of its rival’s products. Owning TPUs would make Google a permanent fixture inside OpenAI’s own infrastructure, not just a vendor it can walk away from at renewal time.
Why TPUs, Why Now
The economics explain the urgency. Amazon, Microsoft, and Google parent Alphabet plan to spend as much as $600 billion combined this year on AI infrastructure, as investors increasingly ask when that spending turns into meaningful returns. In that climate, no lab can afford a single supplier with pricing power over its most important input.
Alphabet reports quarterly results on Wednesday, with investors watching for signs that its AI investment keeps driving cloud growth. Reuters notes that Alphabet expects to recognize only a small percentage of TPU sales revenue by the end of this year, with the vast majority expected in 2027. An OpenAI purchase would pull that future forward.
The Same Week, Oracle Went All In Too
OpenAI is not the only convert. Reuters reports that Oracle said Tuesday it will offer Google’s TPU chips on its cloud for the first time, alongside an expanded multi-gigawatt deal for Nvidia’s newest Vera Rubin servers and offerings from Cerebras and SambaNova. Oracle’s cloud chief put it plainly: different AI workloads have different compute needs, and customers want the flexibility to choose.
Nvidia still dominates training, but the inference market — the everyday work of answering user requests — is more competitive and diverse. Oracle is betting customers want a menu, not a monopoly. OpenAI buying TPUs is the same bet, made with its own money.
Why This Matters
The commodity version of this story is “OpenAI buys chips from Google.” The real story is the escalation ladder: rent in 2025, distribute models on Google Cloud last week, buy hardware this week — with Oracle joining the TPU camp on the very same day. In a single week, Google’s chips crossed from internal advantage to industry infrastructure, backed by two of the biggest spenders in AI.
What to watch next is scale. If the purchase is small, it is a hedge — leverage in negotiations with Nvidia. If it is large, OpenAI genuinely plans to split its fleet between Nvidia and Google, and every other lab will face the same question. Either way, the era when Nvidia was the only name on an AI purchase order is ending, and Google is the first rival to benefit.
Meanwhile in AI Tech: Microsoft’s Maia 300 chip is its own bid to loosen Nvidia’s grip, and the AI chip debt wave pulling in SpaceX, Broadcom, and Oracle shows how expensive this arms race has become.
