AMD buys chip startup that hardwires AI models into its silicon
AMD just acquired the startup betting on baking AI models directly into silicon — a bet that on-device inference is the next frontier.
Why it matters
AMD is doubling down on specialized silicon for embedded AI inference, acquiring Taalas to own the IP and roadmap for hardwired model execution. This signals a shift in the compute buildout: away from centralized GPU farms toward distributed, model-in-silicon edge inference — and AMD's willingness to acquire rather than build it internally.
The key facts
5 to knowAMD acquires Taalas (startup hardwiring AI models into silicon)
Taalas' current chip runs Meta's Llama 3.1 (small version)
Taalas roadmap includes larger, more advanced models
Acquisition date: August 6, 2026
Strategic focus: on-device/edge AI inference vs. cloud-centric compute
Go to the source
CNBC Technologycnbc.com
Publisher excerpt: Taalas' current chip runs a small version of Meta's Llama 3.1, though the company is working on chips for bigger and more advanced models.