Cerebras (NSDQ:CBRS) just booked a US$20 billion OpenAI deal

Core revenue jumped 92%, but the 750MW OpenAI commitment quietly rewires the long-term model

Cerebras Systems (NSDQ:CBRS) has delivered its first quarterly result as a public company, and the numbers themselves are not the most interesting part of the filing. Core revenue of US$191.3 million was up 92% year on year, hardware grew 60%, and cloud and services jumped 167%. Solid, but in a sector where every AI infrastructure name is putting up triple-digit growth somewhere on the page, that alone does not reframe the story.

What does reframe it is the multi-year, 750 megawatt deployment agreement with OpenAI valued at more than US$20 billion. For context, that single contract is roughly 23 times the midpoint of the company’s full year 2026 core revenue guidance of US$855 to 865 million.

Add the AWS partnership bringing Cerebras inference into the world’s largest cloud, the US$6.4 billion IPO closed in Q2, and a US$1 billion working capital loan from OpenAI itself sitting on the balance sheet, and the picture shifts. This is no longer a wafer-scale curiosity racing Nvidia on benchmarks. It is now the AI infrastructure stock with a named, hyperscale demand pipeline locked in.

The market’s job from here is to work out whether Cerebras can actually build, ship and operate at that scale without the margin profile collapsing along the way.

The US$20 billion OpenAI deal changes how this business should be valued

Strip away the noise and the OpenAI agreement is the single most important line in this filing. A 750MW inference commitment over several years gives Cerebras a contracted revenue runway most listed AI infrastructure peers can only describe in vague pipeline language.

It also explains the strange shape of the balance sheet. A US$621 million current loan from a customer, plus a US$362 million non-current portion, is OpenAI essentially pre-funding the capacity Cerebras will build to serve it. Customers do not lend you a billion dollars unless they need what you are building.

The risk sitting underneath this is customer concentration. OpenAI, G42, MBZUAI and AWS now dominate the revenue story. The Master Relationship Agreement with OpenAI is the most important commercial document this company has, and any wobble there would do real damage to the thesis.

Q2 guidance flags the margin cost of growing this fast

The Q1 core gross margin of 47% looks healthy. The Q2 guide of 36 to 38% does not. Management is openly telling investors that ramping the OpenAI and AWS deployments will compress margins in the near term as new data centre capacity comes online and starts absorbing fixed cost.

Full year core gross margin guidance of 38 to 41% confirms this is not a one-quarter event. Core operating margin is guided to negative 28 to 32% for the year, which means losses widen in absolute terms even as revenue scales 69% at the midpoint.

Our concern is that the market may have priced in the OpenAI headline without fully digesting the J-curve. This is a heavy capex, capacity-led business now, and the next four quarters are about building, not earning.

The balance sheet finally matches the ambition

Cerebras ended Q1 with US$3.3 billion in cash, equivalents, restricted cash and short-term investments, before the IPO proceeds landed. With the US$6.4 billion raise added in Q2 and an US$850 million revolving credit facility closed in April, the company is funded to chase the deployment schedule without an obvious near-term capital hole.

Property and equipment jumped from US$437 million to US$572 million in a single quarter, and customer warrants on the asset side grew from US$152 million to US$516 million in total. That is the visible signature of a company turning contracts into infrastructure at pace.

What this also does is buy management the freedom to chase further data centre acquisitions. They have flagged this strategy explicitly, and the credit facility is sized for exactly that.

The Investors Takeaway for Cerebras Systems

The bull case here is straightforward. Cerebras has the fastest inference hardware in the world by independent benchmarks, a US$20 billion anchor customer, AWS distribution, and US$10 billion-plus of capital to deploy. If management executes, the FY2027 and FY2028 revenue numbers could look genuinely transformative versus the FY2026 guide.

The bear case is just as clear. Margins compress through 2026, losses widen, customer concentration is extreme, and any slippage in the OpenAI deployment timeline would force a painful re-rate. Investors should watch quarterly progression of contracted megawatts actually brought online, because that is the real KPI underneath everything else.

Investors can find more in-depth coverage of US-listed semiconductor and AI infrastructure names at stocksdownunder.