As The Chip Competition Intensifies, AMD Increases Its Bet on AI Inference with Taalas

In order to develop its technology and capitalize on the rapidly expanding market for AI inference workloads, Advanced Micro Devices (AMD.O) announced on Thursday that it will purchase semiconductor company Taalas for an undisclosed sum.

Taalas, a company based in Toronto, creates customized silicon to lessen memory and processing bottlenecks in AI inference, which is the process of using trained AI models to provide predictions or replies.

As AI moves from training to real-time, high-volume deployment and businesses compete to reduce processing costs, specialized inference chips have become a crucial focus for semiconductor manufacturers.

In March, the larger competitor Nvidia (NVDA.O) introduced a new core processor and AI system based on technologies from the inference-focused chip firm Groq.

AMD intends to use AMD Instinct graphics processing units (GPUs) to create system-level solutions and incorporate Taalas’ technology into its accelerator roadmap.

According to a statement from Vamsi Boppana, senior vise president of AMD’s Artificial Intelligence Group, “AMD is developing a full-stack AI platform that gives customers the flexibility to deploy the right compute solutions for every AI workload.”

By “delivering differentiated inference performance and ⁠efficiency,” Taalas’ engineering and technology team enhances AMD’s AI portfolio, Boppana continued.

In order to promote the development of semiconductors suited for AI models, Taalas, which was created in 2023, raised $169 million in February, increasing its total capital to about $219 million.

Thru a series of acquisitions, AMD has been strengthening its AI business. The business announced in November that it had acquired MK1, a high-speed inference-focused AI software startup.

The business added FastFlowLM to its Artificial Intelligence Group in July after acquiring MEXT in June to broaden its AI portfolio.