Cerebras lifts 2026 outlook on AI demand

Cerebras Systems raised its full-year 2026 core revenue forecast to a range of $940 million to $960 million, citing strong demand for AI computing capacity and high-speed inference. The updated outlook reflects growing uptake of the company’s AI infrastructure as it competes in training and inference workloads. Following the forecast increase, shares of Cerebras rose 11% in extended trading on Nov. 12.

For the fourth quarter of 2026, Cerebras expects core revenue of approximately $310 million to $330 million. The company also projects core gross margin in the range of 47% to 49% and core operating margins between a loss of 1% and a profit of 1%, signaling a path toward breakeven on a core basis.

Cerebras reported record financial results for the third quarter of 2026, ended Sept. 30. GAAP total revenue reached $210.8 million, up 55% year over year, while core total revenue came in at $230.9 million, an increase of 70% year over year. Third-quarter adjusted sales of $230.9 million exceeded analysts’ average estimate of $214.70 million.

Core cloud and other services revenue was $144.9 million in the quarter, up 272% from a year earlier, reflecting rapid growth in AI inference workloads and contributions from large cloud agreements. GAAP cloud and other services revenue was $141.9 million, up 265% year over year. Cerebras said it won commercial business with two new hyperscalers, as well as neoclouds and large enterprises.

On a profitability basis, GAAP gross margin improved sequentially to 20%, while core gross margin reached 41%, up both sequentially and year over year. GAAP operating margin stood at negative 106%, and core operating margin at negative 14%, indicating that the company remains loss-making even as it scales revenue.

Cerebras said surging demand for high-speed inference from new and existing customers is driving growth across its business lines. The company’s core cloud and services revenue includes a multi-year agreement to provide computing power for OpenAI, and management noted that the pipeline for high-speed AI inference opportunities increased by more than 10 times during the quarter.

To support this demand, Cerebras is expanding its infrastructure footprint. It has secured 700 gross megawatts of data center capacity for delivery by the end of 2027, moving its pipeline of opportunities to gigawatt scale and positioning Cerebras Cloud among the largest non-hyperscaler AI clouds. The company signed contracts for additional capacity in Norway and Finland, broadening its European reach and supporting long-term cloud inference needs.

Cerebras’ technology portfolio includes a wafer-scale engine chip the size of a dinner plate containing trillions of transistors, designed as an alternative to connecting thousands of smaller graphics processors. The company is also deploying production-scale heterogeneous disaggregated inference, which Cerebras says delivers more high-speed tokens per megawatt, aiming to improve revenue potential for data centers and lower costs per token for AI users.

Cerebras highlighted several customer and partner developments that contributed to its third-quarter performance and outlook. It announced an important cloud deal with a top three hyperscaler, with additional details expected in 2027. The company also expanded its reach through neocloud providers, giving customers more options to access its high-speed inference capabilities.

Cerebras’ high-speed inference will power the Bilt Neighborhood OS platform and Bilt Concierge AI assistant, serving eight million consumers and 50,000 businesses. In addition, Cerebras was a day-zero launch partner for Moonshot AI’s Kimi K3.1, a 2.8 trillion-parameter frontier open-weight model, which runs on Cerebras hardware at speeds of up to 2,000 tokens per second. Management said it expects continued progress as it pushes the limits of AI infrastructure performance.