Tech Featured
  • Home
  • Apple
  • Apps
  • Automobiles
  • Gadgets
  • Gaming
  • Reviews
  • Tech Tips
No Result
View All Result
  • Home
  • Apple
  • Apps
  • Automobiles
  • Gadgets
  • Gaming
  • Reviews
  • Tech Tips
No Result
View All Result
Tech Featured
No Result
View All Result
Home AI AI News

OpenAI and Anthropic Are Buying 26 Million Google TPUs, and the Orders Stretch to 2029

Nov 16, 2026
in AI News
Reading Time: 4 mins read
0
OpenAI and Anthropic Are Buying 26 Million Google TPUs, and the Orders Stretch to 2029

The AI chip race has a new scoreboard. TF International analyst Ming-Chi Kuo wrote on X early Monday that OpenAI and Anthropic have placed orders for about 26 million Google TPUs between them, covering 2028 and 2029. The orders give Google chip shipment visibility through 2029, a runway Kuo says now runs further than Nvidia GPUs and rival AI accelerators.

The two labs building the most talked about AI models are now buying their compute from Google, not just renting it. That is the detail that matters, because it turns Google from a cloud landlord into a chip supplier for its own competitors.

12 million for OpenAI, 14 million for Anthropic

Kuo puts OpenAI at about 12 million TPU 8t chips, designed by MediaTek, bought outright. Anthropic accounts for about 14 million TPU 8i chips, designed by Broadcom, through a mix of leasing and outright purchases. Both labs are expected to grow their orders year over year.

The 8t and 8i are variants of Google’s eighth generation TPU, the same family as the TPU 8tx and 8i Max chips Google unveiled last week, which lock sensitive AI workloads inside a secure enclave. Anthropic’s own IPO filing points the same way in dollars, with additional TPU compute expected to be delivered between 2028 and 2031.

Put together, the two labs’ 26 million chips would be about 79 to 87 percent of Kuo’s estimated total TPU shipments for 2028 and 2029 combined. In other words, Google’s eighth generation TPU business is, for the next two years, mostly an OpenAI and Anthropic business.

Google’s runway now stretches to 2029

Kuo’s central claim is about visibility. “Strong new orders from OpenAI and Anthropic have extended Google TPU shipment visibility to 2029, ahead of Nvidia GPUs and other AI ASICs (mostly 4Q27 to 2Q28),” he wrote. The market generally expects AI capital spending to peak in 2028. Kuo’s numbers say TPU shipments keep growing past that point.

YearEstimated TPU shipments
20278 to 9 million units
202814 to 15 million units
202916 to 18 million units

From 2028, Kuo says the v9 generation becomes Google’s main compute platform, with the Humufish and Triggerfish projects at its core. The v10 line, now named Quillfish, is expected to reach mass production in 2029. MediaTek is the sole design supplier for both v9 and v10, and Kuo puts its share of total TPU supply at 45 to 50 percent in 2027, rising to 55 to 60 percent by 2029.

Nvidia’s share of future compute has already fallen to 39 percent

The orders land in a rough week for Nvidia. Jensen Huang resigned as chief executive last week, and the board named chief financial officer Colette Kress as interim chief while it investigates reports that Huang was targeted by a Chinese spy operation. Kuo says the uncertainty around Nvidia’s succession is helping Google discuss “post-2028 long-term plans” with customers.

The numbers behind that shift are stark. Michael Cembalest, chairman of market and investment strategy at J.P. Morgan Asset Management, wrote that Nvidia GPUs account for 39 percent of the computing infrastructure frontier AI labs have contracted going forward, down from 83 percent. Custom chips from Broadcom rank second at 24 percent, followed by Google’s TPUs at 16 percent, AMD GPUs at 15 percent, and Amazon’s Trainium at 6 percent.

That is a forward-looking share of contracted compute, not of chips already sold, and it is exactly the kind of number that explains why Google’s order book now stretches to 2029. The labs have already decided where their future compute comes from, and Nvidia is no longer the default answer. Our earlier report on the Nvidia board investigation has the full timeline of the succession crisis.

What this means for the AI chip race

Three things stand out. First, the TPU has crossed a line. It started as Google’s internal tool for search and YouTube recommendations. Now it is a merchant product, and its two biggest customers are the labs competing hardest with Google’s own Gemini models. Google is, in effect, selling shovels to its rivals in the gold rush.

Second, the winners beyond Google are MediaTek and Broadcom. MediaTek designed OpenAI’s 8t chips and holds the sole design orders for the next two TPU generations. Broadcom designed Anthropic’s 8i chips and already has a separate deal with OpenAI, announced in October 2025, to build 10 gigawatts of custom AI accelerators. OpenAI also agreed to receive 6 gigawatts of GPUs from AMD, which is how the 39 percent figure for Nvidia came to be.

Third, the contest is not one-sided. Tests cited by AITNT show Google’s Ironwood TPU leading Nvidia on tokens per dollar for some agent inference tasks, while an optimized Nvidia GB300 surpassed the TPU by 1.5 to 2.6 times in one model test. Nvidia still wins raw speed in places. Google is winning on supply, price, and the simple fact that its chips are available when the labs need them.

One caveat applies to all of this. Kuo’s unit counts are estimates from his industry survey, not figures disclosed by Google, OpenAI or Anthropic, and none of the three companies has confirmed them. The direction is consistent across every independent report of the post, from Stocktwits to FourWeekMBA, but the exact millions could move. What will not move is the message the orders send: the AI labs that once queued for Nvidia GPUs are now planning their 2029 compute around Google silicon, and Nvidia’s grip on the future of AI compute has loosened fast.

Sources: Ming-Chi Kuo, TF International Securities, via X, November 16, 2026; Stocktwits; FourWeekMBA; J.P. Morgan Asset Management; Seoul Economic Daily; AITNT.

Previous Post

Perplexity Teammates Arrive on the Mac, but Only the $200 Max Plan Gets Them

Emily Cooper

Emily Cooper

Emily Cooper writes about artificial intelligence, gadgets, and emerging technology. She follows new AI model releases, benchmarks, and developer tools closely, and focuses on what they actually mean for readers.

For any queries, you can reach us at [email protected]

Latest Posts

  • OpenAI and Anthropic Are Buying 26 Million Google TPUs, and the Orders Stretch to 2029
  • Perplexity Teammates Arrive on the Mac, but Only the $200 Max Plan Gets Them
  • ChatGPT Wallet Lets OpenAI’s Agents Buy Things for You, but Your Bank May Say No
  • iPhone Duo 2 Face ID Is a Toss-Up, Gurman Says, and the Clock Has Nearly Run Out
  • Samsung’s Gemini Smart Glasses Start at $395 and Ship December 10, and They Work With iPhones Too
  • Guest Posts
  • Advertise
  • Contact Us
  • Privacy Policy

© 2026 JNews - Premium WordPress news & magazine theme by Jegtheme.

No Result
View All Result
  • Home
  • Apple
  • Apps
  • Automobiles
  • Gadgets
  • Gaming
  • Reviews
  • Tech Tips

© 2026 JNews - Premium WordPress news & magazine theme by Jegtheme.