AMD signs AI inference deal with Cerebras after Anthropic, AMD stock falls 3%

Advanced Micro Devices AMD and Cerebras Systems (CBRS) have announced a technical partnership to develop a disaggregated artificial intelligence inference solution, combining AMD's Helios rackscale AI infrastructure with Cerebras' Wafer-Scale Engine technology.

The companies unveiled the offering at Advancing AI 2026, saying the joint solution is designed to meet the growing demand for AI infrastructure capable of handling different inference workloads while improving efficiency and reducing latency.

The system integrates AMD Helios rackscale solutions with Cerebras' Wafer-Scale Engine in a single inference workflow.

According to the companies, AMD Helios will process prompts and large context windows, while the Cerebras Wafer-Scale Engine will handle token generation for applications requiring faster response times.

Based on modeling conducted in July 2026, the companies said the combined system is expected to deliver up to five times higher tokens per second per watt compared with a Cerebras Wafer-Scale Engine-only configuration.

The deal marks the second successive deal AMD made within 2 days. AMD on the previous day signed a deal to invest $5 billion in Anthropic.

Despite the announcements, AMD shares fell 3% on Thursday, while Cerebras stock rose 3.9%.

Companies target diverse AI inference workloads

AMD and Cerebras said AI inference workloads are becoming increasingly diverse, creating demand for infrastructure optimized for different computing needs.

According to the companies, high-volume inference workloads prioritize maximizing token generation, while applications such as coding assistants, real-time copilots and live AI agents require significantly lower latency and faster response times.

"AI inference is becoming one of the largest infrastructure opportunities in AI, and its growing diversity requires a more flexible approach," AMD Chief Executive Lisa Su said.

"Together with Cerebras, we are extending that leadership into the most latency-sensitive applications."

Andrew Feldman, CEO and co-founder of Cerebras, said, "Partnering with AMD gives us an incredible opportunity to bring that performance to even more customers."

Cerebras to deploy AMD Helios systems

As part of the partnership, Cerebras plans to deploy AMD Helios systems across its data centers.

The companies said the joint inference solution is expected to become available initially through Cerebras Cloud in the second half of 2026.

AMD said the collaboration is intended to provide customers with infrastructure that matches compute technologies to specific workload requirements as AI deployment continues to expand across enterprise and cloud environments.

Partnership follows Anthropic agreement

The Cerebras announcement comes a day after AMD reached a reported agreement with Anthropic covering tens of billions of dollars' worth of AI servers, according to a Wall Street Journal report.

Under the reported agreement, Anthropic will purchase up to 2 gigawatts of AMD's next-generation Instinct MI450 chips beginning in the first half of 2027.

The report also said AMD will invest up to $5 billion in Anthropic, contingent on the AI startup meeting specific deployment milestones.

Cerebras has experienced sharp volatility since its May market debut.

After pricing its IPO at $185 per share, the stock surged to an intraday high of $386.34 on its first day of trading before retreating to below $161 in late June. Following Thursday's rally, the shares were trading at $217.78.

Earlier this year, Cerebras announced a more than $10 billion agreement with OpenAI to provide 750 megawatts of AI computing capacity through 2028, marking one of the company's largest commercial deals to date.