OpenAI Plans to Buy Google’s AI Chips Report Says

OpenAI reportedly plans to buy Google’s AI chips known as Tensor Processing Units (TPUs) according to a report by The Information published on October 27, 2026. The Information reported on 27 October 2026 that OpenAI plans to buy Google’s AI chips known as TPUs. The report cites a person familiar with the matter and no price or chip count was shared. This matters most to investors who track Nvidia and Alphabet and to teams that pay for AI computing. A purchase would be a bigger step than the cloud rental OpenAI began in 2025. This guide explains what was reported and how TPUs compare with Nvidia GPUs. It also covers who gains and what is still unconfirmed. Treat it as a plan and not a signed deal.

What Did the Report Say About OpenAI Buying Google’s AI Chips?

OpenAI plans to buy AI chips from Google as it looks to expand its computing capacity and diversify its sources of processing power. The Information first reported the potential purchase. Other outlets subsequently covered the report. The purchase has not been publicly confirmed by either company.

Here is what we know:

• Terms unknown: The financial terms and other details of the planned purchase could not be learned.
• No final deal: Final terms of a potential deal have not been determined.
• A hiring signal: OpenAI posted a job for an engineer to lead the work of tuning its frontier models for TPUs.
• Early stage: The Information said the work of tailoring OpenAI models to TPUs appears to be at an early stage.
• No public reply: Neither company had responded to the report in the coverage I reviewed.

Why Is OpenAI Turning to Google’s TPUs?

OpenAI needs more AI computing power. Training AI models takes huge clusters. Running them for millions of users takes even more over time. Demand for AI computing keeps rising as more firms adopt tools like intelligent automation.

The Main Reasons

• Inference costs: OpenAI first rented Google TPUs in 2025 with the goal of reducing inference costs. Inference is the daily work of running a model for users.
• Chip supply diversification: Relying on one supplier is risky when demand is high. A second source gives OpenAI more room to grow and to negotiate.
• A cheaper alternative to Nvidia: Google designs TPUs for AI workloads but the relative cost and performance depend on the model workload infrastructure and software optimization.
• Inside knowledge: OpenAI named Richard Ho to run its hardware division. He previously led development of Google’s Cloud TPUs.

Renting vs Buying Chips

Renting gives speed and a low upfront cost. You pay as you use and you can leave. Buying gives more control and can lower the cost per chip over time. But you also take on power and data center space and engineering work. A move from renting to buying suggests OpenAI expects steady demand for TPU capacity.

OpenAI and Google TPUs: What Is the History?

The two companies compete in AI products and search. Yet their chip ties have grown step by step.

1. June 2025: Reuters reported that OpenAI had started renting Google Cloud TPUs to run ChatGPT and other products. It was OpenAI’s first meaningful use of chips not made by Nvidia.
2. Limits at the time: Google was not renting its most powerful TPUs to OpenAI.
3. A public denial: OpenAI said it had no specific plans to use TPUs at present.
4. Its own chip: OpenAI is developing a chip called Jalapeno with Broadcom to support its models later this year.
5. October 2026: The Information reports a plan to buy TPUs.

That denial now looks outdated. It shows how fast the chip plans of AI labs can change.

How Do Google TPUs Compare With Nvidia GPUs?

A graphics processing unit (GPU) is a flexible chip that handles many tasks at once. A Tensor Processing Unit (TPU) is a custom AI chip that Google designed for AI math. Here is a quick comparison.

Point	Google TPUs	Nvidia GPUs
Design	Custom chip built for AI workloads	Flexible chip adapted for AI
Seller	Google through cloud rental and limited direct sales	Nvidia sells widely to many buyers
Latest line	TPU 8t for training and TPU 8i for inference	Blackwell and Rubin
Supply	Google keeps most for itself	Broad but tight when demand spikes
Cost angle	Often described as cheaper	Premium priced
Software	Models need tuning for TPUs	Large existing software base
Best fit	Big steady workloads you can optimize	Flexible needs and fast setup

Google launched its eighth generation TPUs as TPU 8t and TPU 8i for training and inference. Google also trained its Gemini 3 model on TPUs. That shows they can handle frontier work.

What This Means for Buyers

TPUs may cost less per task. But the switch is not free. Engineers must tune models to run well on new hardware. That is why the OpenAI job listing matters. Savings only show up after that work is done.

What Does This Mean for Nvidia and the AI Chip Market?

Nvidia remains a major supplier of AI computing hardware but the potential purchase could give OpenAI another source of computing capacity. The real risk for Nvidia is pricing power. A top customer with a second supply line has more leverage.

Anthropic shows the pattern. Its filing lists 4.7 GW of Nvidia GPUs and 4.5 GW of TPUs and 1.2 GW of Trainium and 0.8 GW of AMD GPUs. Nvidia still leads but the gap is small. You can see how Anthropic is using that compute in our post on the Claude Duet AI agent.

Other players are also moving:

• Broadcom (AVGO): It co designs TPUs and built OpenAI’s first custom chip.
• Microsoft Maia and Amazon Trainium: Big cloud firms are building their own chips.
• Cerebras: Its core cloud revenue grew 287% in the second quarter of 2026 with help from a major OpenAI deal.

The result is a multipolar AI chip market. Many labs now mix suppliers instead of picking one.

Does This Affect Nvidia Stock?

The reports I reviewed gave no clear price reaction. One outlet said the plan puts strain on the Nvidia relationship. Until terms are known any move is short term news.

What Does This Mean for Alphabet (GOOGL) and Google Cloud?

For Alphabet this is a proof point. Google is trying to turn its in house chips into a product line.

• Direct sales began: Sundar Pichai said Google would deliver TPUs to a select group of customers in their own data centers.
• First deliveries: CFO Anat Ashkenazi said Google delivered to customer data centers for the first time in Q2 and booked a small amount of revenue.
• 2027 is the big year: Most TPU revenue is expected to land in 2027.
• Cloud growth: Google Cloud revenue grew 82% to $24.8 billion in Q2 2026 and its backlog reached $514 billion.
• Oracle OCI: Oracle will offer TPUs on its cloud starting in 2027.

Morgan Stanley estimated that selling 500,000 TPUs to a third party data center could add about $13 billion to 2027 revenue. There are limits though. Google still keeps most of its TPUs for itself. Hardware also tends to earn lower margins than cloud services.

Google is also reshaping search with AI. If that is your area see our guide to AI rank tracking for Google AI Overviews.

What Has Been Confirmed and What Is Still Unverified?

Confirmed

• The Information published the report on 27 October 2026.
• Reuters reported in June 2025 that OpenAI rented Google TPUs.
• Google has started selling TPU systems to customers.
• Oracle plans to offer TPUs on OCI in 2027.
• OpenAI has a public job listing for TPU optimization work.

Still Unverified

• The price and the number of chips.
• Whether OpenAI would buy hardware or rent more cloud capacity.
• Which TPU generation OpenAI would get.
• Whether Google will release its most powerful TPUs to OpenAI.
• Any public comment from OpenAI or Google.

The only source is one unnamed person. Wait for official word before you treat it as a done deal.

Key Takeaways

• The OpenAI Google AI chips report says OpenAI plans to buy TPUs but the terms are unknown.
• OpenAI wants lower inference costs and a more diverse chip supply.
• TPUs may cost less than Nvidia GPUs but they need extra engineering work.
• Nvidia stays central because OpenAI still plans huge Nvidia spending.
• Alphabet gains proof that TPUs can serve outside customers and most revenue comes in 2027.
• Watch for official confirmation and the TPU generation involved.

Frequently Asked Questions

Who will make Google AI chips?

Google designs its TPUs. Broadcom is confirmed as Google’s TPU design partner through 2031. Google uses these chips in its own services and rents them through Google Cloud.

What is the best AI chip stock to buy?

No single stock fits everyone and this is not financial advice. Nvidia sells the most AI chips and Broadcom co designs custom chips for Google and OpenAI and Anthropic. Alphabet now sells TPUs too. Compare the risks of each and speak with a licensed advisor.

Who is the largest manufacturer of AI chips?

Nvidia is the largest AI chip seller by most counts. One analysis puts its share at roughly 70% of AI chips. Custom chip makers like Google and Broadcom are growing fast.

What is Google’s 2026 plan?

Google raised its 2026 capital spending guidance to $195 billion to $205 billion. It is also shipping eighth generation TPUs and selling TPU systems to select customers. Most of that revenue is expected in 2027.