AMD announced on August 6, 2026 that it is acquiring Taalas, a Toronto-based startup founded in 2023 that builds "model-specific integrated circuits" — chips that etch an AI model's weights and dataflow directly into silicon instead of shuttling them in and out of high-bandwidth memory. Financial terms were not disclosed, and the deal is expected to close in Q4 2026 pending regulatory approval.
Taalas's first chip, HC1, was fabricated on TSMC's 6-nanometer process and served Meta's Llama 3.1 8B model at roughly 17,000 tokens per second, claimed to be tens of times faster than comparable Nvidia GPUs at a fraction of the power draw. The catch: a finished chip runs only the model it was designed for, since only a couple of its 100-plus silicon layers change between designs. A second chip, HC2, targeting models around 20 billion parameters, is due later in 2026. AMD plans to fold the technology into its Helios rack-scale systems alongside Instinct GPUs and EPYC CPUs.
"AMD is building a full-stack AI platform that gives customers the flexibility to deploy the right compute solutions for every AI workload," said Vamsi Boppana, AMD's senior vice president of its AI group. It's AMD's third AI acquisition in nine months, following MK1 and Mext, and comes seven months after Nvidia's $20 billion purchase of Groq — underscoring how the chip industry's race is shifting from training capacity toward faster, cheaper inference.