Google has launched two new versions of its eighth-generation Tensor Processing Unit chips to rival Nvidia hardware, as Bloomberg reported on November 9, 2026, helping lift Alphabet stock up to 3.2% in premarket trading.
TPU 8tx and 8i Max Architecture Focus on Security and Capacity
The newly announced chips include the 8tx variant, which focuses on confidential computing and secure workloads for government agencies and regulated industries. This system pairs with an AMD central processing unit to encrypt data directly at the chip level, addressing the memory decryption requirements typically seen during active compute cycles, as chip division head Amin Vahdat explained in an interview reported by Bloomberg.
The second version, the 8i Max, packs two TPU units into a single package to deliver higher compute capacity and expanded memory. This configuration targets complex mathematical calculations, long-context-window requirements for drug discovery, and intensive coding applications.
TPU Technology Grows from Internal Tools to Commercial Cloud Sales
Google developed the underlying TPU technology alongside Broadcom over a decade ago. Initially deployed for internal search tasks, the architecture eventually expanded into a commercial cloud offering utilized by external clients such as Anthropic, Meta Platforms, Apple, and Oracle.
Bloomberg projected that Google’s TPU revenue, which encompasses both hardware and software, will surge past $135 billion by 2028, climbing from an estimated $12 billion this year. This forecast excludes additional revenue generated by customer deployments running through Google Cloud.
Price-to-Performance Advantages Over Nvidia
Nvidia continues to see processing unit demand outpace available supply, even amid broader macroeconomic debates regarding data center capital expenditure. Nevertheless, competitive hardware alternatives are carving out distinct market share.
One of the primary advantages of the TPU lineup is its cost structure. Bloomberg Intelligence analyst Mandeep Singh noted that TPUs run roughly 30% cheaper than rival Nvidia graphics processing units while maintaining competitive throughput performance.
