Prime Highlights :
- Cerebras has launched CS-4, a three-chip AI server designed to speed up AI inference and chatbot responses.
- The company plans to deliver 600 megawatts of computing capacity by the end of 2027.
Key Facts :
- Cerebras develops AI chips and computing systems focused on inference workloads.
- CS-4 uses 5-nanometre chips and 50% fewer components than the company’s earlier system design.
Background :
Cerebras Systems has introduced a new AI server system designed to process chatbot queries faster as competition in the AI hardware market grows.
The new system, called CS-4, uses three large AI chips developed by Cerebras. The company said the design can improve performance by reducing the need to move data between separate chips. This helps lower delays and energy use during AI workloads.
Cerebras competes with major chipmakers such as Nvidia and focuses mainly on AI inference. Inference is the process through which an AI system generates responses to user requests.
CS-4 uses Cerebras’ Nexus server architecture, which allows chip modules to be connected to the system. The rack also includes the company’s WSE-3 Turbo chip and new networking technology aimed at moving data between chips more quickly.
The chips are built using Taiwan Semiconductor Manufacturing Co’s 5-nanometre process. Cerebras plans to make the new system available in the third quarter.
The company has also reduced the number of components needed to build the system by 50%. Cerebras Chief Technology Officer Sean Lie said this could help data centres install the systems faster.
Cerebras plans to introduce another generation of its chip and server technology in 2027. Chief Executive Officer Andrew Feldman said the company expects to deliver 600 megawatts of computing capacity by the end of 2027.
Feldman also said Cerebras aims to significantly increase the speed and processing capacity of its future systems. The company expects strong demand as businesses expand their use of AI applications.
Cerebras recently reported an adjusted loss of $6.9 million on sales of $180.1 million.
