Google unveils two specialized TPUs launching in 2027: the 8tx, to be paired with an AMD CPU, and the 8i Max, a dual-chip package for long-context window tasks

You can now use Google's TPUs on Oracle Cloud Infrastructure starting in 2027, in addition to Google's own cloud.

Google is launching two new specialized AI chips, the TPU 8tx and TPU 8i Max, with availability expected next year to compete with Nvidia. The TPU 8tx is designed for confidential computing, encrypting data at the chip level and pairing with AMD CPUs, targeting regulated industries and public sector clients. The TPU 8i Max is optimized for AI tasks requiring long context windows, such as drug discovery and coding, by utilizing a dual-chip package for enhanced processing power and memory. These chips are reconfigured versions of existing silicon, adapted to meet specific customer requests. Google previously released two TPU models this year, a faster cadence than its historical one new processor every two years. This specialization strategy aims to deliver more value than a single, all-encompassing chip. Customers for Google's TPUs include Apple, Meta, and Oracle, with Anthropic having the right to acquire significant quantities.

Google's strategic shift to specialized TPUs, moving away from a one-size-fits-all approach, signals a maturing AI hardware market where tailored solutions offer distinct advantages over general-purpose processors. This differentiation caters to emerging use cases like confidential computing and extended context windows, which require specific architectural features. The availability of these chips on Oracle Cloud Infrastructure marks a significant expansion beyond Google's own ecosystem, intensifying competition among cloud providers for AI workloads and providing customers with more choice and potentially specialized performance.

The introduction of these specialized TPUs, particularly the 8tx's confidential computing capabilities, addresses growing concerns around data privacy and regulatory compliance in AI development. For businesses in sensitive sectors, this offers a path to leverage AI without compromising data security. The 8i Max's focus on long-context windows directly impacts fields like scientific research and software development by enabling more comprehensive and nuanced AI analysis. This specialization indicates a broader industry trend toward bespoke AI hardware optimized for specific applications, pushing the boundaries of what AI models can achieve.