Google Unveils 8th-Gen TPU Chips Promising 3x Performance Boost for AI Supercomputing

Google unveils its 8th-generation TPU chips — the TPU 8t and TPU 8i — delivering nearly 3x the compute performance per pod, 80% better performance-per-dollar, and 2x improved energy efficiency over the previous generation, set for general availability later this year.

• Google is unveiling its eighth-generation Tensor Processing Units — the TPU 8t and TPU 8i — two purpose-built chips designed to power the next era of AI supercomputing for training and inference workloads.
• The TPU 8t delivers nearly 3x the compute performance per pod over the previous generation, scaling to 9,600 chips and 121 ExaFlops, while the TPU 8i offers 80% better performance-per-dollar and is optimized for low-latency, agentic inference tasks.
• Both chips feature up to 2x better performance-per-watt than the previous generation, support popular AI frameworks like JAX and PyTorch, and are set to become generally available later this year as part of Google's AI Hypercomputer platform.