Cerebras Systems raised its 2026 adjusted revenue forecast to $940 million to $960 million, from its previous range of $880 million to $890 million, according to Reuters. The Sunnyvale, California, chipmaker announced the revised outlook with its third-quarter results, against a backdrop of growing demand for AI computing.
The company reported record GAAP total revenue of $210.8 million for the quarter, up 55% from a year earlier, in a Cerebras Systems news release. Adjusted sales, a measure the company highlights for investors, rose 70% to $230.9 million, above the average analyst estimate of $214.70 million compiled by LSEG, according to Investing.com. The Raleigh News & Observer also reported rising demand for AI computing capacity.
Cloud Growth and Contracted Business
Cerebras reported $144.9 million in non-GAAP Core cloud and other services revenue for the quarter, up 272% year over year. The company has a $20 billion, multi-year agreement to provide computing power for OpenAI, according to its earnings release. Separately, it reported $26.1 billion in remaining performance obligations as of September 30, 2026, representing contracted future business.
Cerebras CFO Bob Komin said the company’s pipeline for high-speed AI inference increased more than tenfold during the quarter, according to Cerebras Systems. Inference is the computing that takes place when a user queries an AI model, rather than when the model is initially trained. Cerebras says its wafer-scale engine places processing on a single, large chip, which it says can compete with systems built from many smaller processors. Reuters reported that the company is positioning its technology against rivals including Nvidia and major cloud providers.
Widening Losses Alongside Growth
Cerebras’ loss per share widened to 67 cents from 41 cents in the same quarter a year earlier, the company reported. It also said it had secured 700 gross megawatts of data-center capacity for delivery by the end of 2027, according to its earnings release.
Data centers add to California’s power needs
Data centers accounted for about 1,000 megawatts, or 2% of California ISO peak electricity demand in early 2026, a share projected to reach 4,500 megawatts, or 9%, by 2040, according to the California Energy Commission. The projection points to a growing electricity requirement for the state’s data centers, but does not establish whether power supplies will be sufficient for expanding facilities. The commission says its data-center forecast reflects a snapshot of reported data as of December 2025 and will be updated in future cycles as conditions change, according to its forecasting methodology.
Regional forecasts point to rising grid demand
Beyond California, PJM’s 2025 report projects data-center-related electricity demand in its regional footprint could grow by up to about 30 gigawatts between 2025 and 2030. NOVEC forecasts data-center summer peak loads in its Northern Virginia service territory rising from roughly 1,400 megawatts in 2025 to more than 5,000 megawatts in summer 2030. These are regional projections for future demand, not measurements of electricity use already occurring.
The 67-cent loss per share compared with analysts’ expected loss of 14 cents, CNBC reported. Adjusted sales growth of 70% also slowed from 103% in the second quarter, according to CNBC. The outlet said Cerebras shares had fallen about 47% since the company’s Nasdaq debut in May. Cerebras forecast fourth-quarter revenue of $310 million to $330 million, above analysts’ $271 million estimate, and said it expects revenue to roughly triple in 2027. That outlook would be just short of the $2.92 billion projected by analysts surveyed by LSEG, CNBC reported.
Investors React to the Earnings Beat
Shares moved sharply in extended trading following the report, though accounts of the size of that move differ. CNBC reported the stock soared 18% in extended trading, while a separate Investing.com report said shares surged over 16% in after-hours trading Thursday. Either way, the market's reaction reflected investor enthusiasm for a company racing to keep pace with what Reuters described, in a report from Bengaluru, as an AI boom that has spurred demand for chips used to train models and power inference, per a report from Reuters.
