Cerebras Cloud Revenue Up 265%, Company Lifts Full-Year Outlook

Cerebras Systems raised its full-year revenue forecast on November 12, 2026, after cloud revenue jumped 265% to $141.9 million in the third quarter. Total GAAP revenue climbed 55% year over year to a record $210.8 million. The GAAP net loss widened to $161.2 million, up from $22.2 million a year earlier.

The company now targets $940 million to $960 million in full-year core revenue. Its previous guidance, issued on August 12, 2026, ranged from $880 million to $890 million. For the fourth quarter, Cerebras projects a core operating margin between negative 1% and positive 1%. That adjusted measure sits near breakeven because core results exclude stock-based compensation and data-center pass-through items, and add back noncash customer-warrant amortization recorded against revenue.

Stock-based compensation alone represented $155.1 million of the adjustments in the third-quarter operating reconciliation. That gap explains why the core result looks far healthier than the GAAP one.

What is driving the revenue mix?

Cloud and other services supplied most of the quarter's GAAP revenue. Hardware revenue declined from a year earlier. The shift moves more of Cerebras's growth toward selling inference capacity as a service, alongside direct system sales and access through third-party clouds. The company attributed the 265% cloud growth to demand for high-speed AI inference.

Two new deals extend that model. Oracle is integrating Cerebras's fast inference into Oracle Cloud Infrastructure, with availability planned for 2027. Cerebras also signed a cloud agreement with an unnamed customer it described as a top-three hyperscaler, with more detail expected in 2027.

The demand backlog is large. Remaining performance obligations stood at $26.1 billion as of September 30, 2026, up from $25.4 billion at June 30, 2026. At that scale, deployment schedules and available capacity determine how much of the committed demand converts into reported revenue.

How much cash is the buildout consuming?

Expansion is cash-intensive. In the first nine months of 2026, Cerebras used $388.9 million in operating activities and purchased $1.046 billion of property and equipment. The company ended September with $7.8 billion in cash, cash equivalents, restricted cash and short-term investments. That war chest gives it room to finance the data-center buildout without immediate pressure.

Can Cerebras deliver the capacity it promised?

The company has secured 700 gross megawatts of data-center capacity for delivery by the end of 2027, including new contracts in Norway and Finland. Its August release described more than 600 megawatts live or under contract for the same deadline, a figure that also counted capacity already operating.

Manufacturing is scaling in parallel. Cerebras works with Flex and Sanmina and expects to exit 2026 with more than 10 times the manufacturing capacity it had at the start of the year. It plans additional factories and capacity in 2027.

The company says its disaggregated inference architecture can increase the number of tokens a data center delivers by up to five times, lifting potential revenue per megawatt. That claim matters because it targets high-speed inference without a proportional increase in powered infrastructure — the difference between buying less iron and buying more of it.

What happens next?

The raised outlook ties Cerebras's next phase to execution as much as to demand. The test ahead is converting $26.1 billion in commitments into deployed inference capacity while managing cash use and GAAP losses. Rapid cloud growth will only produce sustainable compute economics if the megawatts arrive on schedule.