Start your day with intelligence. Get The OODA Daily Pulse.

Home > Briefs > Google Expands Custom Silicon Strategy with TPU 8tx and TPU 8i Max to Intensify Nvidia Rivalry

Google Expands Custom Silicon Strategy with TPU 8tx and TPU 8i Max to Intensify Nvidia Rivalry

Google introduced two new specialized variants of its proprietary Tensor Processing Units (TPUs), the TPU 8tx and the TPU 8i Max bringing its total TPU v8 configurations to four. The 8tx is engineered specifically for confidential computing, pairing a TPU 8t core with AMD processors to implement chip-level data encryption tailored for government agencies and tightly regulated industries like healthcare and finance. The 8i Max integrates two TPU 8i chips into a single package, dramatically expanding memory capacity and throughput to handle massive, multi-megabyte context windows required by modern frontier models, code generation, and complex analysis. With TPUs running roughly 30% cheaper than comparable Nvidia GPUs while delivering high data throughput, Google is aggressively scaling its merchant silicon footprint, backed by massive adoption commitments from cloud partners (like Oracle) and major AI labs (including Anthropic and OpenAI). Highlighting a strategic shift toward workload-specific vertical integration, Google’s hardware team is leveraging its close ties with Gemini model development to drastically shorten its product iteration cycles and capture surging market demand for inference infrastructure.

Full report : Google rolls out specialized eighth-generation TPU variants targeting confidential government workloads and massive long-context inference as merchant chip demand accelerates.

Tagged: AI Chips Google