Slicast
Tuesday, September 22, 2026
DarkSubscribe
Slicast
AI Infrastructure · News & Analysis
Home › Chips & Hardware › Report
Chips & Hardware · Report

AMD acquires Taalas to build model-specific AI inference chips with model weights etched directly into silicon.

Competitive escalation in inference chip design; signals market shift toward model-optimized rather than general-purpose inference acceleration.
Trade pressSlicast · August 7, 2026 at 01:21 UTC · US · Source: digitimes
importance 92
Read the original
Companies in this story
AMDCapex history →
In-depth analysis
CommentaryAMD AI Infrastructure, September 2026: $14B Core Scientific Deal and $450M LUMI-AI Contract Expand the Challenge to NVIDIACommentaryAMD MI355X Benchmarks and Infrastructure Contracts, September 2026
Related reports
Chips & HardwareAMD acquires Taalas, AI chip startup specializing in model-specific inference chips that etch weights directly into silicon for up to 17,000 tokens/second throughput.2026-08-07Chips & HardwareAMD announced acquisition of Taalas, a Toronto-based startup that etches AI model weights directly into silicon for inference acceleration.2026-08-07Chips & HardwareAMD acquires Taalas, a Toronto-based AI inference chip startup with custom silicon that hardwires AI models into silicon, delivering up to 17,000 tokens/second in early demos.2026-08-07Chips & HardwareAMD acquired Canadian AI chip startup Taalas to deepen AI inference capabilities and challenge NVIDIA dominance.2026-08-07Chips & HardwareAMD is acquiring Taalas, an AI inference chip startup that hardwires AI models into silicon, achieving up to 17,000 tokens/second.2026-08-07
AMD acquires Taalas to build model-specific AI… · Slicast