OpenAI Plans Direct Purchase of Google TPU AI Chips
The reported hardware deal deepens commercial ties between the two AI rivals as compute demands surge.
In short
- OpenAI plans to buy Tensor Processing Units directly from Google to expand computing capacity and diversify processing sources.
- The transaction follows OpenAI renting TPUs through Google Cloud since 2025 and announcing frontier model availability on Google Cloud.
- OpenAI is concurrently preparing its custom 'Jalapeno' silicon developed with Broadcom, led by former Google chip executive Richard Ho.
- The move comes as hyperscalers commit up to $600 billion combined to AI infrastructure this year.
OpenAI plans to purchase artificial intelligence chips directly from rival Alphabet's Google, according to a report published Tuesday by The Information. The deal would supply Google's Tensor Processing Units (TPUs) to expand OpenAI's computing capacity for training and running frontier artificial intelligence models.
Final terms of the transaction have not yet been determined, and both OpenAI and Google declined to comment, according to NDTV Profit. However, a direct hardware purchase would mark an important shift in OpenAI's supply chain, providing an alternative to Nvidia graphics processing units (GPUs) that currently dominate the market for machine learning compute.
Expanding Beyond Nvidia Compute
OpenAI remains one of Nvidia's largest hardware customers, relying heavily on its GPUs to train flagship models and support inference workloads for ChatGPT. Nvidia has also invested billions of dollars in OpenAI, as detailed by AIM Media House.
Despite that relationship, the ChatGPT maker has progressively diversified its infrastructure to secure lower-cost alternatives. Reuters previously reported that OpenAI began renting Google TPUs via cloud instances in June 2025. OpenAI has also been hiring an engineer specifically tasked with optimizing its frontier models on TPUs, according to AIM Media House.
Google began working on TPUs more than a decade ago, initially using the processors to speed up its web search engine and improve efficiency, according to NDTV Profit. Over the past year, Google has started selling the specialized processors directly to selected external customers rather than restricting access solely to Google Cloud instances.
Growing Commercial Ties Between Competitors
While OpenAI and Google compete in AI products and search, their commercial collaboration has widened in recent weeks. Just last week, OpenAI announced that it would make its frontier models available to customers through Google Cloud, as reported by BusinessWorld Online.
Google has also supplied TPUs to other competitors, notably Anthropic. Alphabet Chief Financial Officer Anat Ashkenazi previously stated that Google expects to recognize a small portion of revenue from direct TPU sales agreements before the end of this year, with the vast majority of revenue anticipated in 2027, according to BusinessWorld Online.
Google's semiconductor distribution is widening across third-party cloud platforms as well. On Tuesday, Oracle announced that Google TPUs will become available through Oracle Cloud Infrastructure starting in 2027, with capacity scaling over several years, as reported by NDTV Profit. In announcing that agreement, Google executive Amin Vahdat stated that the next generation of AI will require a diverse range of compute.
Custom Silicon and the Infrastructure Race
Purchasing external processors is only one part of OpenAI's hardware strategy. The company is actively developing its own custom processor family, codenamed Jalapeno, in collaboration with Broadcom, according to The Information via Investing.com.
OpenAI unveiled the first generation of the Jalapeno chip design in June and plans to deploy the silicon by the end of this year as part of a multi-generation processor roadmap. OpenAI's in-house semiconductor group is led by Richard Ho, a Google veteran who was previously involved in the development of TPUs, according to NDTV Profit.
These procurement and custom design efforts occur against the backdrop of massive capital expenditure across the sector. Amazon, Microsoft, and Alphabet plan to spend as much as $600 billion combined this year on AI infrastructure, according to BusinessWorld Online. As leading AI laboratories race to secure processing power, multi-vendor silicon architectures are fast becoming standard operating practice.
Questions people ask
What chips is OpenAI planning to buy from Google?
OpenAI plans to purchase Google's custom Tensor Processing Units (TPUs), specialized processors designed for training and running artificial intelligence models.
Has OpenAI used Google hardware before?
Yes. OpenAI began renting Google TPUs through Google Cloud in June 2025 before exploring direct hardware purchases.
Is OpenAI abandoning Nvidia GPUs?
No. OpenAI remains one of Nvidia's largest GPU customers; the planned TPU purchase and in-house chip development represent efforts to diversify suppliers and expand total computing capacity.
What is OpenAI's custom chip project?
OpenAI is developing custom processors codenamed Jalapeno in partnership with Broadcom, led by former Google chip executive Richard Ho, with initial deployment planned for later this year.
Sources
- OpenAI plans to buy Google AI chips to expand computing capacity - The Informationinvesting.com, Oct 27, 2026
- OpenAI plans to buy Google’s AI chips, The Information reportsBusinessWorld Online, Oct 28, 2026
- OpenAI Plans To Buy Google's Chips As Part Of Its AI Build-OutNDTV Profit, Oct 28, 2026
- OpenAI Plans to Purchase Google TPUs: ReportAIM Media House, Oct 28, 2026
How this story was made: the newsroom picked it up from Google Search, gathered the full text of the sources above, and drafted it with AI assistance. Every factual claim was then checked against those sources before publishing (31 claims checked). Illustrations marked as AI-generated are not photographs. Spotted an error? Tell us.
Published October 28, 2026 at 06:09 UTC
