Cerebras expects to deploy 600MW of computing capacity by 2027 as it scales its AI hardware business beyond individual processors.
Cerebras Systems has introduced its latest AI computing platform, the CS-4, as it seeks to strengthen its position in the rapidly expanding AI inference infrastructure market.
The new server system is built around three Cerebras AI processors and is designed to accelerate inference workloads, where trained AI models process queries and generate responses. Cerebras competes with Nvidia in AI computing, with its architecture focused on reducing the data movement required between processors.
The CS-4 is based on Cerebras’ Nexus server architecture, which uses modular units to house its processors. The system is powered by the company’s WSE-3 Turbo chip, manufactured using TSMC’s 5nm process technology.
Cerebras has also developed new networking components aimed at improving data movement between the processors. Its large-chip architecture is designed to keep more processing within a single chip, potentially reducing the performance and energy overhead associated with transferring data across multiple processors.
The company has also focused on simplifying data-centre deployment. Cerebras said the CS-4 uses 50% fewer components, which could reduce infrastructure complexity and help operators accelerate the construction of AI computing facilities.
The CS-4 is expected to be available in the third quarter of 2026. Cerebras is also developing a new generation of its processor and server architecture for 2027.
The company expects to deploy around 600MW of computing capacity by the end of 2027, highlighting its plans to scale AI infrastructure to meet rising demand.
The launch comes after Cerebras reported $180.1 million in revenue and an adjusted loss of $6.9 million in its latest financial results. The company has raised its annual targets, citing strong demand for AI computing infrastructure.
CEO Andrew Feldman said Cerebras is targeting substantial gains in processing speed and throughput across its upcoming generations of AI systems as it continues to expand its AI hardware portfolio.
