OpenAI Reportedly Wants Google’s AI Chips. Is Nvidia’s Biggest Customer Hedging?

The biggest buyer of Nvidia chips may be shopping at Google. That is the report that shook the AI chip market this week.

But note the word “reported.” The story rests on a single anonymous source, and neither company has confirmed it. We will keep that line in view all the way through.

What was reported

The Information reported on October 28, 2026 that OpenAI plans to buy Google’s tensor processing units, or TPUs. TNW, which relayed the report, says it did not verify it. The report cites one person familiar with the matter.

No numbers came with it. There is no price, no chip count and no delivery date. Decrypt’s write-up says the same: neither company has confirmed the plan.

Reuters, as cited by TNW, says OpenAI declined to comment. Google did not immediately respond.

TPU versus GPU, in plain words

A GPU is a general-purpose chip that Nvidia dominates. A TPU is a chip Google designed for one job: the matrix maths inside neural networks. Google describes its Cloud TPUs as custom chips built to accelerate machine learning.

Think of a GPU as a skilled all-round cook. A TPU is a machine built to make one dish very fast. Which one wins depends on the dish.

This is not a surprise move

OpenAI has circled Google’s chips before. Reuters reported in June 2025 that OpenAI had begun renting them. OpenAI later said it had no active plans to use them in its products, according to TNW.

There are more signs. Decrypt reports an OpenAI job listing in San Francisco for an engineer to optimise frontier models for TPUs. TNW adds that OpenAI said last week its frontier models would come to Google Cloud.

So the direction was visible. What is new is the word “buy.” Renting capacity is flexible. Buying chips is a commitment.

Why Google would sell

Google said in April it would sell TPU hardware directly to a small group of customers that run it in their own data centres, per TNW. Alphabet’s finance chief, Anat Ashkenazi, expects a small share of TPU sales revenue by year-end and most of it in 2027, according to the same report.

That is a real shift. For years, TPUs were mostly a reason to rent Google Cloud. Selling them turns Google into a chip vendor as well.

There is a tension, too. TNW notes that TPU demand is high, partly because of Google’s deals with Anthropic, and that Google’s own researchers have reportedly had to wait for access. A new large buyer would add to that queue.

The twist: OpenAI is already spreading its bets

Here is the part that makes this less dramatic than it sounds. OpenAI is not leaving Nvidia. It is adding suppliers.

Decrypt reports that Nvidia signed a letter of intent in September 2025 to invest up to $100 billion in OpenAI, tied to at least 10 gigawatts of Nvidia systems. Nvidia ultimately invested $30 billion in OpenAI’s funding round, well below that letter.

Two weeks after the Nvidia deal, AMD gave OpenAI the right to buy up to 160 million AMD shares, tied to 6 gigawatts of AMD chips. OpenAI is also preparing its own Broadcom-built inference chip, with initial deployment planned for the end of 2026.

Add Google TPUs and you get four routes to compute. That looks less like a divorce and more like a portfolio.

We made a similar point about another lab in Anthropic’s custom chips are a margin play. And our earlier piece, How Big Tech Is Fighting Over AI Chips, Not AI Models, explains why the chip layer now decides who can scale.

What this could mean for the market

If the report is right, three things follow. First, Nvidia keeps a huge customer but loses some pricing power over it. A buyer with options negotiates better.

Second, Google gains a showcase customer for TPU sales. Winning OpenAI would be a strong signal to other buyers.

Third, software gets harder. Models tuned for one chip family do not move to another for free. The TPU job listing hints that OpenAI knows this.

None of this is certain. We have no volumes and no dates. A plan can change before a contract is signed.

There is also a competitive angle. OpenAI and Google compete in AI products and search, according to TNW, yet they are deepening ties. In this market, rivals are also suppliers. That mix makes every deal harder to read.

We would add one caution. A report about plans is not a report about purchases. Treat it as a signal about direction, not as proof of a signed order.

Why every lab wants more than one supplier

Training and running large models needs huge amounts of compute. Chips are only part of it, but they set the pace. If one supplier is late, a lab waits.

TNW cites Reuters on the scale of the spending. Amazon, Microsoft and Alphabet plan to spend as much as $600 billion combined on AI infrastructure this year. At that size, a small gain in price or supply is worth a lot.

So a second source is a form of insurance. It also gives a buyer something to point at when it negotiates. Even a chip that is never bought in volume can change the price of the chips that are.

What builders should take from this

If you run models yourself, the lesson is practical. Do not tie your whole stack to one chip family unless you must.

Ask three questions. Which chips does your serving software support today? How much work is a move to another family? And what would you pay to keep that option open? The answers decide how much leverage you have when prices change.

What to watch next

Watch for a confirmation from either company, or for filings that list TPU purchases. Alphabet’s commentary on TPU sales may also say more.

Also watch OpenAI’s own chip. If the Broadcom design ships on time, the mix of suppliers will shift again.

For now, the safe reading is narrow. A report says OpenAI wants a second chip supplier. Compute is the scarce resource, and every large lab is acting like it.

Sources & Credits

Text credit & news reporting source: TNW (relaying The Information and Reuters). TNW: ‘OpenAI reportedly plans to buy Google’s TPU chips alongside Nvidia’

Text credit & news reporting source: Decrypt (relaying The Information). Decrypt: ‘OpenAI Plans to Buy Google’s AI Chips as It Looks Beyond Nvidia: Report’

Product documentation: Google Cloud. Google Cloud: Cloud TPU

Enjoyed this breakdown? Stay ahead of the curve with our technical analyses and news. Join our community by subscribing to the CVisiona Newsletter for weekly deep dives, and subscribe to our CloudMelon Vis YouTube Channel for video breakdowns of Cloud-Native, AI, and Tech Enterprise news.

Question for you: Is a second chip supplier a smart hedge for OpenAI, or a sign that GPU supply is still too tight? Tell us in the comments.