AMD has agreed to buy Taalas, a Toronto startup that makes chips hardwired to run only one specific AI model, betting that the next phase of the AI boom will reward inflexible, specialized chips over Nvidia's multi-purpose GPUs. Taalas' chips, used for AI inference, run "thousands of times faster" than a conventional GPU on the exact model they target, but cannot easily switch to new models. CEO Ljubisa Bajic says the platform can transform any AI model into custom silicon, with a new model realized in hardware in "only two months." The current chip runs a compact version of Meta's Llama 3.1 and is built on an earlier TSMC process with fast memory attached. Taalas is designing chips for larger models. AMD's CEO Lisa Su said at a July product launch, "I'm a big believer that there's no one-size-fits-all as it comes to chips," though GPUs will keep the majority of AI chip sales. AMD recently began shipping Helios, its first rack-scale system, competing with Nvidia's integrated server racks; Meta and Microsoft are early customers. The first Helios system was delivered on June 24, 2026, to a data-center facility in Rockdale, Texas. AMD previously bought Silo AI in 2024 for $665 million.