AMD and Cerebras Forge Partnership for Cutting-Edge AI Inference Solutions

In a significant development for the artificial intelligence (AI) industry, Advanced Micro Devices Inc. (AMD) and Cerebras Systems have announced a technical partnership to create an innovative AI inference solution. This collaboration combines AMD's Helios™ rack-scale solutions with Cerebras' Wafer-Scale Engine technology, unveiled at the Advancing AI 2026 conference. The joint solution is poised to deliver ultra-low latency and high throughput, meeting the increasing demands of advanced AI applications.

The newly announced AI inference solution is designed to deliver exceptional performance metrics that could redefine industry standards. By integrating AMD Helios with Cerebras’ Wafer-Scale Engine in a single inference workflow, the companies aim to achieve up to five times higher tokens per second per watt (T/s/W). This dramatic enhancement in efficiency is particularly crucial as AI workloads evolve, requiring different balances of latency, throughput, token capacity, and scalability.

AI inference workloads are becoming increasingly diverse, with varying requirements based on use cases. For instance, high-volume workloads focus on maximizing token generation, while real-time applications like coding assistants and agentic workflows necessitate reduced response times. The AMD and Cerebras solution effectively addresses these challenges through a disaggregated inference approach, optimizing the two primary stages of the workflow independently.

AMD Helios contributes ultra-high throughput, adeptly processing prompts and large context windows, while Cerebras’ Wafer-Scale Engine accelerates memory-bandwidth-intensive token generation with ultra-low latency. This innovative combination ensures a differentiated platform that does not compromise on speed or scalability.

Dr. Lisa Su, chair and CEO of AMD, articulated the strategic importance of this collaboration, stating, “AI inference is becoming one of the largest infrastructure opportunities in AI, and its growing diversity requires a more flexible approach.” She highlighted that the AMD Helios platform delivers leadership performance across a broad range of inference workloads, further solidified through the partnership with Cerebras.

Andrew Feldman, CEO and co-founder of Cerebras, echoed this sentiment, emphasizing the surging demand for ultra-fast inference solutions. He noted, “Partnering with AMD gives us an incredible opportunity to bring that performance to even more customers,” indicating a strong market potential for the joint offering.

The partnership comes at a critical time as fast token generation becomes essential in various domains, including software development, autonomous agents, robotics, and scientific discovery. The response time directly impacts user experience and system effectiveness, making the joint solution particularly relevant.

AMD Helios is positioned to provide the necessary high-throughput capabilities for processing large volumes of complex requests, while Cerebras’ Wafer-Scale Engine ensures the real-time performance needed for token generation. This combination specifically targets the ultra-low-latency segment of the inference market, with AMD Helios serving as the foundation for balanced inference workloads across data centers.

Cerebras plans to implement AMD Helios systems within its data centers, with the joint solution expected to be available through Cerebras Cloud in the latter half of 2026. This rollout is anticipated to extend the reach of both companies’ advanced technologies, positioning them favorably in the rapidly evolving AI landscape.

The partnership between AMD and Cerebras marks a pivotal moment in the AI infrastructure space, promising to deliver unprecedented performance and efficiency to AI inference workloads. As the demand for ultra-fast inference solutions continues to escalate, this collaboration stands to set new benchmarks for the industry, catering to the diverse needs of modern AI applications. With the backing of two technology leaders, the future of AI inference looks brighter than ever.