AI Semiconductors
Google TPU
TPU is Google's custom AI training and inference chip family, with the latest Ironwood (v7) generation delivering 4,614 TFLOPS per chip — analysts describe it as on par with NVIDIA's Blackwell — offered externally via Google Cloud as a price-competitive alternative to NVIDIA GPUs. Anthropic confirmed a deal for access to up to 1 million TPUs, Google's largest disclosed external commitment to date.
Sources
Every claim traced to a primary source — evidence, recent activity, and insider filing behavior, all in one place.
Evidence
2 sources · 81% avg. source confidence
- Google, official blog — Google's Ironwood TPU delivers 2x performance-per-watt over its prior generation88% source confidence
- SemiAnalysis / WinBuzzer — Anthropic confirms deal for up to 1 million Google TPUs as Google courts NVIDIA-focused neoclouds74% source confidence
Dependency Map
What Google TPU depends on below, and who depends on Google TPU above — click any node to make it the new center, 5 levels deep.
Connection type
Click any node to make it the new center. Scroll to zoom, drag to pan.
The Story So Far
Auto-generated from this entity's dated milestones, relationship updates, and sourced evidence — not AI-written, just sorted.
Google TPU's Timeline
A sourced, dated history of Google TPU's key moments — founding to present.
Jan 2015 · First TPU deployed
Google deployed its first-generation TPU internally, beginning a custom-silicon strategy that let it partially avoid dependency on third-party AI accelerators.
Jan 2026 · Ironwood (v7) generation, Anthropic deal
Google's Ironwood (v7) TPU generation delivered 4,614 TFLOPS per chip, on par with NVIDIA's Blackwell by some analysts' accounts, and Anthropic confirmed a deal for access to up to 1 million TPUs.
How These Connections Evolved
How a relationship changed over time, not just its current state — sourced, dated, and traceable.
Google TPU ↔ Alphabet (Google)
Jan 2026
Google/Alphabet's Ironwood (v7) TPU generation became its most advanced custom AI accelerator as of 2026, reinforcing Google's internal-silicon alternative to third-party GPU dependency.
Now
Google TPU is developed internally by Google/Alphabet, with the Ironwood (v7) generation representing its most advanced iteration as of 2026.