Google to Offer Its TPUs Through Oracle, Anthropic Forecasts $114.5 Billion

Key Takeaways

1Oracle will offer access to Google TPUs starting next year, a first outside of Google's cloud.

2Anthropic expects $114.5 billion in spending at Google between April 2026 and July 2033.

3Projections estimate TPU revenues between $100 billion and $138 billion in 2028.

Why it matters : the advancement of multichip technology could reduce Nvidia's pricing power.

Google will make its TPUs accessible to Oracle customers starting next year, marking a first outside of its own cloud. Anthropic plans to spend $114.5 billion with Google between 2026 and 2033, while analysts anticipate TPU revenues could exceed $100 billion by 2028.

Double-digit billion TPU projections starting in 2026

Mizuho Bank forecasts that Google’s TPU-related revenues will reach $28 billion in 2026 and $67 billion in 2027. For 2028, Mizuho estimates this figure could rise to $100 billion, while John Blackledge from TD Cowen suggests $138 billion, representing 34% of Google Cloud's revenues. In comparison, Nvidia reported $96.2 billion in revenue in the previous quarter. On Wednesday, the market reacted little to these announcements, with the stock price remaining stable.

Oracle to provide access to Google’s TPUs, a first for another hyperscaler

Google and Oracle have announced an agreement that will allow Oracle customers to access Google’s TPUs starting next year. This is the first time Google’s TPUs will be available on the platform of a hyperscaler other than Google itself.

Anthropic commits $114.5 billion to Google and diversifies its chips

In its IPO documents, Anthropic indicates it plans to spend $114.5 billion on computing services from Google between April 2026 and July 2033. The company does not specify what portion of this amount will be allocated to TPUs but estimates its TPU capacity spending with Google and other providers will be between $125 billion and $150 billion. Anthropic also mentions using Amazon's Trainium chips and AMD GPUs, in addition to those from Nvidia and Google.

Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisissez votre rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

OpenAI shows interest in TPUs, Nvidia remains dominant

OpenAI has posted a job listing for an engineer to optimize its models for TPUs and plans to purchase TPUs from Google. Despite these signals, Nvidia's chip business remains significantly ahead of its competitors, with Google among its major GPU clients. Google’s AI chips, developed over more than a decade and initially designed to enhance search, reflect this long-term investment.

Multi-chip strategy could diminish Nvidia's pricing power

Several analysts anticipate an acceleration in Google’s TPU business, driven by the growing adoption of multi-chip strategies by companies. If this trend continues, Nvidia's pricing power could be weakened. Some analysts believe the market undervalues Google as companies seek to reduce their dependence on Nvidia GPUs. On October 18, Mizuho highlighted that Google’s TPU hardware business represents a larger opportunity than current estimates reflect, and that direct TPU sales could alleviate Google’s capital expenditure. Bank of America considers Google’s spending relatively efficient compared to its peers, with TPUs supporting its AI efforts. Mizuho also believes this business could cushion any potential impacts on Google’s search capabilities in the face of rising AI agents. Google declined to comment, while Nvidia and OpenAI did not respond to inquiries.