Cerebras Aims for AI Supremacy with Wafer-Scale Engines
Cerebras Systems is betting on its unique dinner-plate-sized AI chips to take on industry leader Nvidia. The company has raised its revenue and margin forecasts for 2026, citing strong demand for its chips and cloud services.
The company's wafer-scale engines, essentially giant chips built from an entire silicon wafer, are designed to run inference workloads faster by keeping more data and memory close to the processors. This approach is starting to show results, with second-quarter sales rising 74.3% to $180.11 million.
Cerebras has also seen a significant increase in cloud and services revenue, nearly quadrupling to $127.73 million as OpenAI demand picks up. The company now expects 2026 adjusted revenue of $880 million-$890 million and adjusted gross margin of 41%-43%, ahead of analysts' estimates.
The real test for Cerebras will be scaling manufacturing and data-center delivery quickly enough to support its multi-year deal with OpenAI, which has a remaining performance obligation of $25.4 billion. If the company can execute on this plan, margins are likely to improve as production and cloud-delivery costs don't rise one-for-one with output.