Cerebras Systems reported Q2 2026 core revenue of $210 million, more than doubling year-over-year, with its cloud business growing 287% from a year prior (on a core basis). The company, which went public in May 2026 at $185/share, beat guidance across all core business metrics. CEO Andrew Feldman attributed the acceleration to surging demand for "fast inference," reflecting a market shift toward speed-optimized AI inference architectures as companies deploy agents and require single-digit latency.
The core strength is Cerebras' Wafer-Scale Engine-3 (WSE-3), the largest commercially available AI processor at 46,225 mm² silicon with 900,000 cores and 21 petabytes-per-second memory bandwidth. Unlike GPU clusters, the WSE-3 avoids HBM memory, CoWoS packaging, and 3nm lithography bottlenecks—all currently supply-constrained. New partnerships with AMD and AWS are particularly significant: Cerebras and AMD stood up disaggregated inference solutions that increase throughput by 5x while maintaining Cerebras latency, entering production in Q4 2026.
Cerebras expanded its data center footprint to 600+ megawatts live and under contract by end of 2027, with manufacturing capacity increasing 10x in 2026. The company has $25.4 billion in remaining performance obligations as of June 30. OpenAI signed a 750 MW deal worth over $20 billion; AWS partnerships are expanding. New cloud customers include Cognition and Lovable (AI coding), CrowdStrike (security), Block, Figma, and AlphaSense.
For infrastructure investors, Cerebras signals that speed is now a defensible, monetizable differentiator in AI inference. The company's avoidance of HBM/CoWoS constraints gives it supply-chain resilience that pure GPU providers lack. With $25.4B in RPO and 287% cloud growth, Cerebras is capturing a distinct layer of the inference stack against both hyperscaler and neocloud competitors.