AMD and Cerebras join forces on AI inference
The companies say their systems can accelerate separate stages of AI workloads, from prompt processing to token generation.
AMD and Cerebras have announced a partnership to divide AI inference workloads across their respective systems, enhancing efficiency for AI services. Cerebras will deploy AMD Helios systems, with AMD chips handling prompt processing and Cerebras systems accelerating token generation. This collaboration aligns with a trend of using specialized chips for different AI tasks and contributes to AMD’s projection of AI expanding the global computing market to $2 trillion by 2030.
- AMD and Cerebras partner to split AI inference workloads.
- This deal aims to improve the speed and cost of everyday AI services.
- Cerebras will use AMD Helios systems, with AMD chips for prompt processing and Cerebras systems for token generation.
- This follows AMD’s recent deal with Anthropic to supply computing power.
- AMD CEO Lisa Su sees this as a move towards workload disaggregation.
- AI is projected to grow the global computing market to $2 trillion by 2030.
Continue reading https://www.axios.com/2026/07/23/amd-cerebras-ai-chips
Write a comment