Advanced Micro Devices, Inc. (NASDAQ:AMD) just announced an acquisition that could give major competition to NVIDIA Corporation (NASDAQ:NVDA) in AI-inference. On August 6, AMD said it has agreed to buy chip startup Taalas as specialized inference chips become a critical area of focus for semiconductor makers. The financial terms of the deal have not been disclosed.
The company plans to integrate Taalas’ technology into its accelerator roadmap and develop system-level solutions combining it with AMD Instinct graphics processing units (GPUs).
“AMD is building a full-stack AI platform that gives customers the flexibility to deploy the right compute solutions for every AI workload,” Vamsi Boppana, senior vice president of AMD’s Artificial Intelligence Group, said in a statement.
AMD Expands its AI Inference Arsenal
The acquisition strengthens AMD’s AI portfolio by offering it differentiated inference performance and efficiency. The move also highlights the rising importance for leading GPU makers to offer integrated systems with several different components and chips rather than standalone processors.
The acquisition itself follows a string of smaller inference-focused deals made by AMD. Back in November, the company acquired MK1, an AI software startup specializing in high-speed inference. It also acquired MEXT in June and added FastFlowLM to its artificial intelligence group in July. Together, these moves may strengthen AMD’s AI inference capabilities to compete with giants such as Nvidia.
Taalas’s technology could also offer AMD both speed and lower costs. The company says that the HC1 demonstrator reportedly delivers around 17,000 tokens per second per user when running Llama 3.1 8B. It also claims that its systems can cost 20 times less to build and use 10 times less power because they avoid HBM, advanced packaging, and liquid cooling.
AI Inference to Become the Next Nvidia Battleground
AMD’s latest buying spree follows merely months after NVIDIA Corporation (NASDAQ:NVDA) struck a $20 billion deal with Groq, a designer of high-performance AI chips, for its inference technology and talent. Nvidia chips have dominated the process of AI model training for quite some time, but its graphics processors face greater competition from central processing units and custom processors.
Inference computing, the process of answering queries, occurs every time AI models are used. As AI assistants, coding agents, and similar applications handle billions of queries, factors such as latency, throughput, and power efficiency become increasingly important.
