AMD acquires AI chip startup Taalas to boost inference performance by etching models into silicon

AMD Just Bought a Startup That Etches AI Models Directly Into Silicon — Here's Why That Matters

AMD acquired Taalas, embedding weights into silicon; chip served Llama 3.1 8B at 16,960 tokens/sec, 48x faster than H100. Model-specific silicon cuts inference 10x, fragmenting AI hardware tiers and making continuous AI agents economically viable.