Breaking News
Advanced Micro Devices announced Thursday that it is acquiring chip startup Taalas for an undisclosed sum, a move aimed at bolstering its technology as the company pushes deeper into the fast-growing market for AI inference workloads.
Toronto-based Taalas builds specialized silicon designed to reduce computing and memory bottlenecks during AI inference - the process of running already-trained AI models to generate outputs like responses or predictions. Inference-focused chips have become an increasingly central battleground for semiconductor companies as the AI industry shifts from training massive models to deploying them at scale in real time, with chipmakers racing to bring down the computing costs involved.
The move also puts AMD in more direct competition with larger rival Nvidia, which in March unveiled its own new central processor and AI system built using technology acquired from Groq, another startup specializing in inference chips.
AMD said it plans to fold Taalas' technology into its accelerator roadmap and use it to build system-level solutions alongside its Instinct line of graphics processing units. Vamsi Boppana, senior vice president of AMD's Artificial Intelligence Group, said the company is building a full-stack AI platform intended to give customers flexibility in choosing the right compute setup for different AI workloads. He added that Taalas' technology and engineering team will strengthen AMD's AI portfolio by delivering differentiated performance and efficiency in inference tasks.
Founded in 2023, Taalas raised $169 million in a funding round this past February to support development of chips optimized for AI models, bringing its total funding raised to roughly $219 million.
The Taalas deal is the latest in a series of acquisitions AMD has made to expand its AI capabilities. The company acquired AI software startup MK1, which specializes in high-speed inference, in November. It followed that with the purchase of MEXT in June to broaden its AI portfolio, and added FastFlowLM to its Artificial Intelligence Group in July.
Toronto-based Taalas builds specialized silicon designed to reduce computing and memory bottlenecks during AI inference - the process of running already-trained AI models to generate outputs like responses or predictions. Inference-focused chips have become an increasingly central battleground for semiconductor companies as the AI industry shifts from training massive models to deploying them at scale in real time, with chipmakers racing to bring down the computing costs involved.
The move also puts AMD in more direct competition with larger rival Nvidia, which in March unveiled its own new central processor and AI system built using technology acquired from Groq, another startup specializing in inference chips.
AMD said it plans to fold Taalas' technology into its accelerator roadmap and use it to build system-level solutions alongside its Instinct line of graphics processing units. Vamsi Boppana, senior vice president of AMD's Artificial Intelligence Group, said the company is building a full-stack AI platform intended to give customers flexibility in choosing the right compute setup for different AI workloads. He added that Taalas' technology and engineering team will strengthen AMD's AI portfolio by delivering differentiated performance and efficiency in inference tasks.
Founded in 2023, Taalas raised $169 million in a funding round this past February to support development of chips optimized for AI models, bringing its total funding raised to roughly $219 million.
The Taalas deal is the latest in a series of acquisitions AMD has made to expand its AI capabilities. The company acquired AI software startup MK1, which specializes in high-speed inference, in November. It followed that with the purchase of MEXT in June to broaden its AI portfolio, and added FastFlowLM to its Artificial Intelligence Group in July.
See What’s Next in Tech With the Fast Forward Newsletter
SECURITY
Tweets From @varindiamag
Nothing to see here - yet
When they Tweet, their Tweets will show up here.
