AMD Buys Taalas To Etch AI Models Directly Into Silicon

AMD agreed to acquire Toronto's Taalas, a startup that hardwires AI model weights directly into CMOS logic, to bolster its Instinct-powered inference roadmap.

AMD Buys Taalas To Etch AI Models Directly Into Silicon

AMD (NASDAQ: AMD) said Wednesday it has agreed to acquire Toronto-based AI inference silicon startup Taalas, folding a hardware team that etches model weights, math and routing directly into CMOS logic into its Instinct accelerator roadmap.

Hardwired Models For AI Inference

Founded in 2023 by former Tenstorrent CEO Ljubisa Bajic, Taalas builds chips tailored to specific AI models rather than general-purpose processors, physically hardwiring inference dataflows to sidestep the compute and memory bottlenecks of traditional GPUs. AMD said the technology will complement Instinct GPUs, EPYC CPUs, the Helios rack-scale platform and ROCm software as inference becomes the fastest-growing slice of the AI market.

AMD Taalas AI inference acquisition

Nvidia Pressure Grows

The deal, whose terms were not disclosed, is expected to close in the fourth quarter of 2026 subject to regulatory approvals. It arrives days after AMD posted record Q2 revenue of $11.5 billion on the back of a doubling data-center business, and as Nvidia consolidates its grip on training silicon. Vamsi Boppana, senior vice president of AMD's AI Group, said Taalas' team strengthens AMD's inference portfolio with differentiated performance and efficiency.

A Canadian Talent Bet

Taalas' Toronto footprint expands AMD's long-standing Canadian R&D presence and lets AMD keep pace with a wave of specialized inference startups. Bajic said joining AMD gives Taalas the scale and reach to accelerate a roadmap that begins with model-specific silicon. The acquisition tracks broader industry moves this month toward inference-first hardware alongside d-Matrix's Wallaroo.ai buy and Nvidia's Safe Superintelligence backing.

Reporting based on coverage from AMD Newsroom, CNBC, Bloomberg and The Register.

Category: M&A

Tags: AI Startups Mergers & Acquisitions Semiconductors AI Infrastructure AI Chips

Related Articles