
AMD acquires Taalas, a Toronto startup that hardwires AI models into custom silicon, claiming chips in two months and 1,000x efficiency gains.
Advanced Micro Devices has agreed to acquire Taalas, a Toronto-based startup that designs chips to run AI models directly in hardware, bypassing general-purpose graphics processing units. Financial terms were not disclosed. The deal is subject to closing conditions and regulatory approval.
Taalas emerged from stealth in 2024 with $50 million in funding from Quiet Capital and Pierre Lamond among others. The company was founded by three former AMD employees who also led Tenstorrent, a Toronto-born AI chipmaker now based in California. CEO Ljubisa Bajic, COO Lejla Bajic, and CTO Drago Ignjatovic are all ex-Tenstorrent. Bajic said in a release: "We founded Taalas to rethink AI inference from the ground up by building the hardware around the model. Joining AMD will give us the scale, engineering resources and global reach to accelerate our innovation."
The startup claims its approach can produce chips in as little as two months, far faster than the industry norm. Taalas says its hard-wired designs can be a thousand times more efficient than software-based inference, with a single chip potentially outperforming a small GPU data centre.
AMD already has a footprint in Canada, dating to its 2006 purchase of Markham-based graphics chipmaker ATI Technologies. The Taalas deal is AMD's second acquisition of a Canadian AI chip firm in just over a year. In 2025, AMD bought the team behind Untether AI, a Toronto startup developing inference chips marketed as faster and more energy-efficient than rivals.
Vamsi Boppana, senior vice president of AMD's AI group, said in the release: "Taalas' technology and world-class engineering team strengthen our AI portfolio by delivering differentiated inference performance and efficiency."
For AMD, the acquisition adds specialized silicon that could help it compete more directly with NVIDIA in the fast-growing market for AI inference hardware. Taalas' focus on hard-wired computation, converting AI models into custom silicon, targets the same inference workloads that Nvidia's GPUs currently dominate, with a claim of dramatically lower cost and power consumption.
The deal also raises questions about talent retention and integration. Taalas is a small, early-stage company. AMD will need to keep its engineers focused while folding the team into a much larger organization. The two-month chip design timeline, if real, would be a material advantage in a market where new GPU architectures take years to develop.
AMD did not say when it expects the transaction to close. The company's stock has gained roughly 12% this year through Tuesday's close, while Nvidia shares have risen about 140% over the same period.
Drafted by a large language model from the source reporting linked above, then screened by automated publishing checks. It is not read by a journalist before publication. Some articles cite our Alpha Score. Verify prices and figures against the original source. Educational coverage, not personalized advice.