AMD acquires Toronto-based Taalas, which integrates model weights directly into silicon to boost inference performance, for an undisclosed sum (Tobias Mann/The Register)
AMD is buying a startup built around a hard bet: fixed-model silicon can make AI inference faster and cheaper.
Taalas says its chips are designed around the model itself, with weights integrated into silicon rather than treated as data to be moved around. Techmeme’s note cites early demos showing model-specific integrated circuits producing up to 17,000 tokens per second. AMD did not disclose the deal price. The acquisition adds a specialized inference approach to AMD’s AI compute push. Techmeme's note
Taalas says its chips are designed around the model itself, with weights integrated into silicon rather than treated as data to be moved around. Techmeme’s note cites early demos showing model-specific integrated circuits producing up to 17,000 tokens per second. AMD did not disclose the deal price. The acquisition adds a specialized inference approach to AMD’s AI compute push. Techmeme's note
score 6