Google TPUs gain traction with Anthropic, new architectures detailed · 4 sources tracked

By PulseAugur Editorial · [4 sources] · 2026-07-09 13:02

SemiAnalysis reports on the growing market share of Google's Tensor Processing Units (TPUs) for AI model training, highlighting their increasing adoption by external entities like Anthropic for their Claude model releases. The analysis details upcoming TPU architectures, including TPUv8i with its BroadFly topology designed for massive scaling and reduced latency, and the variants of TPUv7 Ironwood, which utilize different torus configurations for varying pod sizes. This expansion of TPU availability beyond Google's internal use presents a competitive alternative for AI labs in the training hardware market. AI

IMPACT Google's TPUs are becoming a more viable alternative for AI training, potentially impacting hardware choices for major AI labs.

RANK_REASON Analysis of hardware supply chain and adoption trends for AI accelerators.

Read on X — SemiAnalysis →

AI-generated summary · Google Gemini · from 4 sources. How we write summaries →

Google TPUs gain traction with Anthropic, new architectures detailed · 4 sources tracked

COVERAGE [4]

X — SemiAnalysis TIER_1 English(EN) · SemiAnalysis_ · 2026-07-09 13:02

Our AI Accelerator Supply Chain model contains the latest updates on the massive traction in external TPU sales and rentals, as well as the implications for the

Our AI Accelerator Supply Chain model contains the latest updates on the massive traction in external TPU sales and rentals, as well as the implications for the supply chain. 4/4🧵
X — SemiAnalysis TIER_1 English(EN) · SemiAnalysis_ · 2026-07-09 13:02

With the upcoming TPUv8i BroadFly topology, it can scale up to 1,024 chips within a 7-hop radius, compared with the traditional 3D torus 16-hop radius. This ena

With the upcoming TPUv8i BroadFly topology, it can scale up to 1,024 chips within a 7-hop radius, compared with the traditional 3D torus 16-hop radius. This enables lower latency, making it easier to overlap communication with compute without incurring the cost of a dedicated htt…
X — SemiAnalysis TIER_1 English(EN) · SemiAnalysis_ · 2026-07-09 13:02

TPUv7 Ironwood comes in two variants: TPUv7x, which uses a 3D torus for physical pod sizes of up to 9,216 chips, and TPUv7, which uses a 256-chip 2D torus. 2/4🧵

TPUv7 Ironwood comes in two variants: TPUv7x, which uses a 3D torus for physical pod sizes of up to 9,216 chips, and TPUv7, which uses a 256-chip 2D torus. 2/4🧵 https://t.co/BE2aSz8wg5
X — SemiAnalysis TIER_1 English(EN) · SemiAnalysis_ · 2026-07-09 13:02

Successfully training models on TPUs has been demonstrated by Anthropic through the past five-plus successful Claude releases. This is positive for the ML commu

Successfully training models on TPUs has been demonstrated by Anthropic through the past five-plus successful Claude releases. This is positive for the ML community, as Google TPUs continue to gain market share outside of internal Google workloads, giving frontier AI labs a https…

COVERAGE [4]

Our AI Accelerator Supply Chain model contains the latest updates on the massive traction in external TPU sales and rentals, as well as the implications for the

With the upcoming TPUv8i BroadFly topology, it can scale up to 1,024 chips within a 7-hop radius, compared with the traditional 3D torus 16-hop radius. This ena

TPUv7 Ironwood comes in two variants: TPUv7x, which uses a 3D torus for physical pod sizes of up to 9,216 chips, and TPUv7, which uses a 256-chip 2D torus. 2/4🧵

Successfully training models on TPUs has been demonstrated by Anthropic through the past five-plus successful Claude releases. This is positive for the ML commu

RELATED ENTITIES

RELATED TOPICS