JP — ITmedia

Oracle Cloud to Adopt Google TPU, Starting Service in 2027

ChipsOctober 28, 20261 min read

Oracle says it will bring Google's AI chip, the Google TPU, to Oracle Cloud and start offering it as a service in 2027. The news was shared at Oracle's yearly event in Las Vegas.

What happened

Oracle announced that it will add Google TPUs to Oracle Cloud Infrastructure (Oracle Cloud). Google TPUs are special chips that Google designed for AI and machine learning work. Oracle said it will provide the power, cooling, and data center networking needed to run these chips at a large scale. The service is planned to start in 2027. The news was made at the keynote speech of Oracle AI World 2026 in Las Vegas.

Why Oracle is doing this

Oracle is trying to catch up with the big cloud companies AWS, Microsoft Azure, and Google Cloud. These companies now make their own AI chips, not only use Nvidia's GPUs. By offering Google's chips, Oracle gives customers another choice besides Nvidia. Google also wants its TPUs to be used more widely, so having them on Oracle Cloud helps Google too.

Key facts

Service start: 2027
Event: Oracle AI World 2026, Las Vegas
Announced at the keynote speech
Oracle will provide power, cooling, and data center networking

New software support

Last week, Google Cloud released TorchTPU as open source. It lets PyTorch, a popular AI software library, run natively on Google TPUs. This could allow users to run many kinds of AI tools on Google TPUs, not only Google's own AI services.

Why it matters

This deal gives companies more choice in AI computing power. It may also weaken Nvidia's strong position in the AI chip market. For Oracle, it is a way to show that it can offer strong AI cloud services. For Google, it means its TPUs can reach more customers.

Glossary

Google TPU: An AI chip designed by Google for machine learning and AI processing.
Oracle Cloud Infrastructure: Oracle's cloud computing service that lets companies rent computers, storage, and networks over the internet.
TorchTPU: Google Cloud's open-source tool that lets the PyTorch AI library run natively on Google TPUs.
PyTorch: A popular free software library that developers use to build and train AI models.
AWS: Amazon's cloud computing service, one of the largest cloud providers.

Infineon to Build First High-Volume 300mm Gallium Nitride Semiconductor Fab

Infineon Technologies Austria is retrofitting an existing cleanroom in Villach, Austria, to produce gallium nitride power semiconductors on 300mm wafers at volume, with production start planned for the end of 2027. The project requires about 1.1 billion euros, of which up to 319 million euros may come from the EU's European Chips Act, though the formal allocation by the Austrian economics ministry is still pending. Infineon says the move makes it the first company worldwide to bring GaN power electronics into volume production.

ChipsThe RegisterOct 28

Big Red gets cozy with Google to offer TPUs to AI customers

Oracle plans to offer Google's Tensor Processing Units (TPUs) on Oracle Cloud Infrastructure (OCI) starting in 2027, with capacity growing over several years, alongside its Nvidia hardware. Oracle says it will be the first hyperscale cloud platform outside Google's own cloud to offer the chips. The announcements at an event in Las Vegas also included expanded OpenAI integrations and a Snowflake deal.

Chipsheise onlineOct 27

Bit Noise: TSMC's chip production hums, PC sales shrink

TSMC, the world's largest contract chipmaker, reported September 2026 revenue of about €14.35 billion, up 55 percent year over year, driven by the AI boom. Full third-quarter figures were published just after the column's editorial deadline, with preliminary data suggesting revenue will exceed the company's own forecasts. The article also notes TSMC is building several new fabs, supported by suppliers such as Da Yi Cheng Technology, whose order books reportedly extend into 2028.