Google to sell its TPUs to some customers, who also fancy big-G GPUs
AI is driving more searches and ads
Google Cloud will start selling its custom tensor processing units to some customers, because they want them and the search giant wants to diversify its revenues.
News of the chip sales emerged on Wednesday during the Q1 2026 earnings call for Alphabet, Google’s parent company.
CEO Sundar Pichai said Google has observed growing demand for TPUs from “AI labs, capital markets firms and high-performance computing applications” and will therefore “begin to deliver TPUs to a select group of customers in their own data centers.” The CEO said some customers have shown “massive interest in our GPU offerings as well.”
Chief Financial Officer Anat Ashkenazi said Google will record some revenue from TPU sales this year, but that the balance sheet impact will be more marked in 2027. She also warned “It is important to keep in mind that revenues from TPU hardware sales will fluctuate from quarter to quarter depending on when TPUs are shipped to customers.”
Even if revenue isn’t enormous or steady, Pichai thinks selling chips will pay off by helping to fund research on next-gen silicon and helping to create economies of scale that make it easier for Google to build kit for its own use.
Amazon Web Services recently teased the possibility of selling its home-grown chips to third-party customers. Google has now beaten it to market – although demand for AI silicon is so high the market is probably big enough for both companies.
Google Cloud is doing just fine without selling TPUs: revenue for Q1 2026 came in at just over $20 billion, 63 percent beyond the $12.26 billion for the same quarter in 2025. Pichai said the cloudy biz could have done better if it were able to build enough infrastructure to satisfy customer demand. But he sees good days ahead because the G-Cloud now has a $460 billion backlog of undelivered contracts, and that represents a near-doubling quarter over quarter.
Ashkenazi said Google expects to recognize just over half of the backlog as revenue in the next 24 months. If Google can achieve that, it will mean annual revenue hits $130-billion plus – not far behind AWS’s $150 billion annual revenue run rate.
Capital expenditure for the quarter reached $35.7 billion. Ashkenazi said the overwhelming majority of that cash went on “technical infrastructure to support the AI opportunities we see across the company” and added “Approximately 60 percent of our investment in technical infrastructure this quarter was in servers, and 40 percent was in data centers and networking equipment.”
The CFO also updated capital expenditure forecasts to between $180 billion and $190 billion, up from the company’s previous forecast of $175 billion to $185 billion. She said the increase reflects the needs of energy and infrastructure biz Intersect, which Google acquired earlier this year.
When generative AI burst onto the scene, pundits warned that consumers could turn to chatbots instead of search engines.
That didn’t happen in this quarter as Pichai proudly pointed to “AI experiences driving usage, queries at an all-time high, and 19 percent revenue growth.”
- Google Cloud Next proves what we suspected: Everything is AI now
- Google Meet or Google Mute? Even CEOs get borked sometimes
- Forget one chip to rule them all: With TPU 8, Google has an AI arms race to win
- Trump-branded datacenter project fails to make itself great, again
AI has also helped Google to improve the relevance of ads it serves to users, which in turn is increasing engagement from advertisers.
Revenue for the quarter was $109.9 billion, up 22 percent year-over-year. Net income landed at $62.6 billion.
Those numbers impressed investors, who in after-hours trading sent Alphabet shares 3.7 percent higher. ®
More about
More about
Narrower topics
- AIOps
- Android
- App stores
- Chrome
- Chromium
- DeepSeek
- Gemini
- Google AI
- Google Cloud Platform
- Google I/O
- Google Nest
- Google Project Zero
- GPT-3
- GPT-4
- G Suite
- Kubernetes
- Large Language Model
- Machine Learning
- MCubed
- Neural Networks
- NLP
- Pixel
- Privacy Sandbox
- Retrieval Augmented Generation
- Star Wars
- Tavis Ormandy
- TOPS
- Waymo
Broader topics
More about
More about
More about
Narrower topics
- AIOps
- Android
- App stores
- Chrome
- Chromium
- DeepSeek
- Gemini
- Google AI
- Google Cloud Platform
- Google I/O
- Google Nest
- Google Project Zero
- GPT-3
- GPT-4
- G Suite
- Kubernetes
- Large Language Model
- Machine Learning
- MCubed
- Neural Networks
- NLP
- Pixel
- Privacy Sandbox
- Retrieval Augmented Generation
- Star Wars
- Tavis Ormandy
- TOPS
- Waymo

