As artificial intelligence models become more capable, the demand for computing power, energy and data-center capacity is accelerating. Cerebras Systems is pursuing a different path for this next phase: wafer-scale computing.
A different architecture for AI
Rather than dividing silicon wafers into many individual chips, Cerebras designs processors across an entire wafer. The approach is built to handle demanding AI workloads and reduce some of the bottlenecks associated with conventional chip architectures.
Founded in 2015 by CEO and co-founder Andrew Feldman, the company provides AI computing through on-premise systems and cloud infrastructure. Its latest platform, CS-4, represents the newest generation of its wafer-scale AI systems.
Compute requires a wider infrastructure vision
The conversation around AI scale now extends beyond processor performance. Larger models require reliable electricity, cooling, manufacturing capacity and data centers capable of operating advanced systems at significant scale.
Cerebras has stated that it expects more than 600 megawatts of data-center capacity to be operational or contracted for delivery by the end of 2027. The company is also expanding manufacturing capacity and planning additional European infrastructure, with a target of up to 200 megawatts in the region by the end of 2027.
At TechCrunch Disrupt 2026, Feldman is set to discuss whether AI can continue scaling at its current pace, how alternative hardware designs may support that growth, and which infrastructure decisions could shape the industry's next chapter.
Wafer-scale systems and expanded computing infrastructure could help make AI development more efficient, opening the way for faster scientific research, more capable digital services and new forms of technological creativity.
