New12,480+ smartphones indexed — free sandbox tier on every plan.Edge-cached JSON responses, typically under 100ms.Official SDKs available on NuGet (C#) and NPM (TypeScript).New12,480+ smartphones indexed — free sandbox tier on every plan.Edge-cached JSON responses, typically under 100ms.Official SDKs available on NuGet (C#) and NPM (TypeScript).
Get Free API Key
Hardware•3 min read•2026-08-22

Anthropic Hires Google TPU Founder Amir Salek to Head In-House Custom AI Chip Development

Anthropic recruits Amir Salek, founder and former head of Google's TPU program, to lead its custom AI silicon team and accelerate proprietary inference hardware.

Tech News Bot
DeviceSpecs Intelligence
Original Source Reference
#Anthropic#Google TPU#Amir Salek#AI Chips#Custom Silicon#Semiconductors#Claude AI

On August 22, 2026, reports confirmed that Claude developer Anthropic has hired Amir Salek—the foundational engineer who created and led Google's custom Tensor Processing Unit (TPU) program for nine years—to head its internal compute and custom semiconductor engineering team. The high-profile recruitment signals Anthropic's aggressive transition toward designing proprietary AI inference silicon, reducing long-term dependence on commercial merchant GPU foundries.

A Watershed Milestone for Anthropic's Custom Silicon Strategy

Amir Salek brings unmatched semiconductor engineering pedigree to Anthropic, having built Google's custom TPU program from its inception into one of the world's most successful domain-specific accelerator architectures. Salek will direct Anthropic's custom silicon roadmap, designing tightly co-optimized processor microarchitectures tailored specifically to Claude's transformer and reasoning model topologies.

Key Strategic Pillars of the Compute Initiative

  • Proprietary Silicon Architecture: Designing custom ASICs optimized specifically for Claude's attention mechanisms, KV cache retrieval, and multi-step reasoning models.
  • Nine-Year TPU Track Record: Leveraging Salek's foundational experience building six generations of Google TPU hardware for hyperscale cloud clusters.
  • Tri-Pronged Compute Buildout: Complements Anthropic's $5B AMD MI450 partnership and $15B Texas TPU data center campus with proprietary silicon IP.
  • Lowering Inference Latency & Cost: Targets substantial per-token energy and cost reductions across enterprise Claude API deployments.

The Frontier AI Lab Custom Silicon Race

By bringing in the architect behind Google's TPU empire, Anthropic joins OpenAI, Google, and Meta in building dedicated in-house chip design teams to control the full stack from silicon wafer to foundational AI software.

Leadership Profile

FeatureDetails
ExecutiveAmir Salek
New RoleHead of Compute & Custom Silicon, Anthropic
Previous ExperienceFounder & Lead Architect, Google TPU Program (9 Years)
Primary MissionDesign and Deploy Custom AI Silicon for Claude AI Models