Cerebras Raises 2026 Revenue Guidance to $940–960M and Signs Microsoft as Cloud Customer
Cerebras, which makes giant AI chips and rents them out through the cloud, raised its adjusted revenue forecast for 2026 to $940–960 million, up from $880–890 million in August. Its new customer is Microsoft: the contract was signed in October, and payments will start in the first quarter of 2027. Cloud revenue grew 265% to $141.9 million, while hardware sales fell. If you are short on GPUs, Cerebras is worth keeping in mind as a backup option.
AI-processed from TNW; edited by Hamidun News
Imagine you are launching an AI product, but fast GPUs are impossible to get in the cloud: there are queues, prices are rising, and delivery times are long. This is exactly the gap Cerebras is targeting. The company makes huge processors the size of a dinner plate and provides access to them through the cloud.
On Thursday, it reported results for the third quarter, and its shares rose about 17% in after-hours trading. There are three reasons: revenue beat analyst forecasts, the company raised its full-year plan to $940–960 million (in August it had guided to $880–890 million), and it gained a large new cloud customer. For those building products on fast inference, that is, on running ready-made models, this is a signal: the alternative to GPUs is gaining serious commercial weight.
What the Numbers Are and Why There Are Two Sets
Cerebras's main metric is called core revenue. It is an adjusted measure: it excludes employee stock compensation and part of data center costs, but adds the value of warrants (rights to buy shares) the company has issued to customers. There is also ordinary revenue under GAAP standards. The difference between the two matters, and it is clearly visible below.
For the quarter ended September 30, core revenue grew 70% to $230.9 million. Analysts had expected $214.7 million (data from LSEG, cited by CNBC). Under GAAP, revenue came to $210.8 million, up 55%. For the fourth quarter, the company forecasts $310–330 million, while analysts had expected $271 million.
The business mix has shifted. Cloud and services brought in $141.9 million, up 265%. Hardware sales fell to $68.9 million from $96.8 million a year earlier. Cerebras is selling less and less hardware and increasingly operating and renting it out itself, acting more as a provider than as a manufacturer.
Microsoft, OpenAI, and the Focus on Data Centers
In the report, the new customer is described only as "one of three largest hyperscalers," with details promised next year. CEO Andrew Feldman told CNBC that it is Microsoft: the contract was signed in October, and revenue will begin arriving in the first quarter of 2027. Partners already include Amazon and Oracle, which plans to offer inference on Cerebras.
A large part of the growth rests on OpenAI. The deal with OpenAI underpins $26.1 billion in remaining contract obligations, meaning contracted but not yet recognized revenue (data from Reuters). Feldman told analysts that OpenAI volumes at Cerebras have been growing since the second quarter, but capacity for a broad rollout was not yet sufficient.
That is why the money is going into infrastructure. The company has contracted 700 MW of data center capacity through the end of 2027, including new sites in Norway and Finland. Over nine months, operating activities consumed $388.9 million, and another $1.05 billion went to property and equipment. At the end of September, the company had $7.8 billion in cash and short-term investments. Feldman acknowledged that the company is considering financing options. According to him, data center availability is constraining the entire industry, and changes in US policy could complicate construction.
Profit Exists, But Only in Adjusted Terms
On an adjusted basis, Cerebras earned $29.8 million, although analysts had expected a $38.4 million loss (Reuters). Under GAAP, the net loss widened to $161.2 million from $22.2 million a year earlier, and $155.1 million of that came from stock-based compensation. The adjusted metric shows how the business itself is performing; GAAP shows what it really costs to attract and retain people.
Feldman expects revenue to triple in 2027. CNBC calculated this at roughly $2.85 billion, slightly below the $2.92 billion analysts had factored in. Even after the jump, the shares remain about 47% below their price at listing. The company plans to begin shipping its new CS-5 system in early summer 2027, and the Wafer Scale Engine 5 processor inside it is already undergoing system testing.
What to Do With This
If you are stuck in line for GPUs or struggling with inference costs, there are more cloud options, including through familiar providers. Feldman says plainly that long delivery times and expensive graphics processors open a niche for Cerebras hardware.
The limitations are also clear. The company itself acknowledges a shortage of data center capacity. Money from Microsoft will not arrive before 2027. Dependence on OpenAI remains high. It makes sense to view Cerebras as an additional option alongside GPUs rather than the only path, and to watch how quickly it expands its capacity.
- Nov 21, 2026Microsoft outage: nearly 83,000 Outlook complaints on Downdetector, Teams and Copilot down too
- Nov 21, 2026MLPerf Training Adds Nvidia Rubin and TPU Ironwood in First Appearance: 153 Systems
- Nov 21, 2026Microsoft Turns Copilot Into an "Operating System for Work": Word and Excel Become Agent Plugins
- Nov 21, 2026Google to split Gemini chat history from model training, so you can opt out without losing your history
Key Questions Answered
What the Numbers Are and Why There Are Two Sets?
Cerebras's main metric is called core revenue. It is an adjusted measure: it excludes employee stock compensation and part of data center costs, but adds the value of warrants (rights to buy shares) the company has issued to customers. There is also ordinary revenue under GAAP standards. The difference between the two matters, and it is clearly visible below. For the quarter ended September 30, core revenue grew 70% to $230.9 million. Analysts had expected $214.7 million (data from LSEG, cited by CNBC).
What are the key details on Microsoft, OpenAI, and the Focus on Data Centers?
In the report, the new customer is described only as "one of three largest hyperscalers," with details promised next year. CEO Andrew Feldman told CNBC that it is Microsoft: the contract was signed in October, and revenue will begin arriving in the first quarter of 2027. Partners already include Amazon and Oracle, which plans to offer inference on Cerebras. A large part of the growth rests on OpenAI.
What are the key details on Profit Exists, But Only in Adjusted Terms?
On an adjusted basis, Cerebras earned $29.8 million, although analysts had expected a $38.4 million loss (Reuters). Under GAAP, the net loss widened to $161.2 million from $22.2 million a year earlier, and $155.1 million of that came from stock-based compensation. The adjusted metric shows how the business itself is performing; GAAP shows what it really costs to attract and retain people. Feldman expects revenue to triple in 2027.
Ship AI Faster with Founders, Engineers & Weekly Architecture Reviews
Inside Vibe Club (hamidun.com/club) — weekly live masterminds with Zhemal Khamidun, Tech Radar, full LMS access, and 370+ production Claude Code skills.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.