Cerebras claims its technology can serve frontier models at approximately 14x speed with 750 output tokens per second. To achieve this, the company stores model weights in 44 GB of on chip SRAM, avoiding the heavy reliance on high-bandwidth memory that may face supply shortages in 2027 and 2028. This latency advantage is designed to support AI agents that trigger numerous consecutive model calls for autonomous tasks.
Wednesday, October 21, 2026
Story
Cerebras Shares Jump 6% After Sam Altman Confirms OpenAI Partnership
AI16 days ago
The chief executive of OpenAI has praised the hardware capabilities of Cerebras, describing the chipmaker as a 'close partner' in a shared effort to advance AI processing speed. Shares of Cerebras rose 6% after Sam Altman cited a deep engagement between the two firms, though he did not provide new commitments regarding model access or production capacity. The comments resolve previous market speculation about a formal tie-up between the AI research company and the chipmaker.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.