Las Vegas, October 27 (Reuters) / PRNewswire / -- On Oracle AI World, Oracle and Google announced today that Google's Tensor Processing Unit ( TPU ) will be available through Oracle Cloud Infrastructure ( OCI ), marking the first time Google's TPU has been launched on a large-scale cloud platform outside of Google. By integrating Google's TPU , which is specifically designed for AI , into OCI 's AI computing product portfolio, customers will gain greater flexibility to match the most demanding workloads with infrastructure that offers near-linear scalability and highly optimized economics.
Mahesh Thiagarajan, President of Oracle Cloud Infrastructure, stated: "Different AI workloads have different computational power requirements. By integrating Google TPU into OCI, customers are able to utilize a variety of leading AI computing architectures on a single cloud platform. This option allows them to select the most suitable infrastructure for each AI workload in terms of performance, throughput, and cost-effectiveness."
Google TPU is a custom accelerator designed in collaboration with open-source software, used to support the entire lifecycle of AI – from cutting-edge model training and large-scale inference to the multi-step inference data requirements needed by agents AI. OCI will provide power, cooling, and data center networking to enable large-scale deployment of Google TPU within its infrastructure.
Google's Senior Vice President of Infrastructure and Chief Technology Officer Amin Vahdat stated: "The next generation of AI will require diverse computing power. TPU offers a powerful and specially tailored option for demanding training and inference workloads. Introducing Google TPU to OCI will enable more organizations to access this technology and provide greater flexibility when building and running the most complex AI applications."
Provide more AI computing power options for each workload.
As the scale of models increases and the demand for inference in production environments rises, the performance and cost-effectiveness of each training and inference request become increasingly important. Google's TPU architecture is designed to maximize the AI processing power per watt of electricity, as well as to enable fast access to model data and rapid communication between chips.
Google TPU supports both large model training and post-training, including dense models and hybrid expert models ( MoE ). It can also be used for inference, as well as in production environments for intelligent agents and long-context applications. Customers will be able to choose the Google TPU that is optimized for their workload, performance requirements, and cost objectives within OCI.
Leveraging over a decade of experience in developing custom AI chips, Google's TPU product portfolio offers specialized computing power optimized for large-scale machine learning workloads. Google's eighth-generation TPU introduces two dedicated architectures: the TPU 8t for training and the TPU 8i for inference. Google claims that the TPU 8t can reduce the development cycle for cutting-edge models from several months to just weeks, and that the computing performance of each pod is nearly three times that of the previous generation; the TPU 8i is designed to handle the iterative work of many specialized agents.
Google's services on OCI will be launched within the natural year of 2027 and will be gradually expanded over the coming years to meet customer needs.
More resources
- Understand Oracle Cloud Infrastructure
- Understand Oracle AI infrastructure
- Understand Google Cloud TPUs
About Oracle
Oracle provides integrated application suites as well as secure, autonomous cloud infrastructure services. For more information about Oracle (NYSE ticker: ORCL), please visit oracle.com.
About Google Cloud
Google Cloud provides a powerful and optimized AI technology stack – including AI infrastructure, leading models such as Gemini, data management capabilities, multi-cloud security solutions, developer tools and platforms, as well as agents and applications – to help organizations transform their businesses in the era of agents. Customers from over 200 countries and regions regard Google Cloud as a trusted technical partner.
About Oracle AI World
Oracle AI World is an event where customers and partners can learn about the latest products and technological innovations, see how AI can be applied in various industries, and engage in discussions with experts and peers. Attendees will receive practical advice and insights that can have an immediate impact within their organizations, and they will also explore how Oracle can help unleash the full potential of AI. Please visit oracle.com / ai-world to register, or follow the news and discussions through oracle.com / news and linkedin.com / company / oracle.
Forward-looking Statements Disclaimer
Statements in this article regarding Oracle's future plans, expectations, beliefs, and intentions are considered "forward-looking statements" and are subject to significant risks and uncertainties. Many factors could affect Oracle's current expectations and actual results, and there may be substantial differences between actual results and those anticipated. A discussion of these factors and other risks that affect Oracle's business is contained in the documents submitted by Oracle to the U.S. Securities and Exchange Commission (SEC), including the sections under the "Risk Factors" heading in Oracle's most recent 10-K and 10-Q reports. These documents can be found at the SEC website or on Oracle's websites at oracle.com / investor. All information in this article is as of October 27, 2026, and Oracle assumes no obligation to update any of these statements due to new information or future events.
Trademarks
Oracle, Java, MySQL, and NetSuite are registered trademarks of Oracle Corporation. NetSuite is the first cloud company that ushered in a new era of cloud computing.
