Cerebras (CBRS.US) Cloud Business Soars, Masking Hardware Decline; OpenAI Supports Half of Cloud Revenue

According to Woofun AI, on November 12 Eastern Time, AI chip and cloud service provider Cerebras Systems (CBRS.US) released its third-quarter financial report, hitting a record high for quarterly revenue and raising its full-year guidance. The stock price rose by about 15% after-hours, but hardware revenue was nearly 30% lower year-on-year.

This structural contrast reveals a critical turning point in the company's transition from pure hardware sales to cloud services. While the market cheers the revenue breakthrough, there is also a need to be wary of the shrinking risk in the core hardware segment. The financial results indicate that the driving force behind total revenue growth is not the traditional core products, but the explosive growth of cloud computing services. This dramatic shift in revenue composition is reshaping the company's valuation logic and delivery cadence. Investors interpreting what looks like a strong report must look beyond surface-level revenue growth to examine the shifting relationship between shrinking hardware business and expanding cloud services.

A deeper breakdown of revenue structure shows that, under the GAAP standard, revenue increased by about $75.1 million year-on-year, but this was not a universal increase. The “Cloud and Other Services” segment added around $103 million compared to the same period last year—its incremental growth even outpaced the company’s overall net increase.

This means that the decline in hardware business offset about a quarter of the incremental growth from cloud services. Looking back at last year, about $3 of every $10 in revenue came from cloud and other services. In this quarter, cloud and other services contributed around two-thirds of total revenue, solidifying their dominant position.

Notably, Cerebras is expanding new channels by bringing inference capabilities onto Oracle (ORCL.US)’s cloud platform, which is expected to be available on Oracle Cloud Infrastructure by 2027, providing the company with a new entry point to reach enterprise customers.

Woofun AI’s compiled data shows that the OpenAI agreement confirmed $72.4 million in revenue this quarter, which accounted for roughly half of cloud and other services’ income, becoming the cornerstone of the high cloud business growth. The shift from “selling devices” to “selling computing power” not only changes the timing of revenue recognition, but also redefines the delivery chain: moving from tracking device shipments to tracking data center power-up, machine installation, and customer service usage. The expansion of new channels and deep integration with major customers together now form the two pillars of high revenue growth, but also mean the company’s dependence on a small number of major customers has risen markedly.

The income statement shows a distinct dual-track feature, with big differences between the GAAP and Core metrics, revealing the impact of accounting treatments and stock-based compensation. Under GAAP, gross profit dropped from approximately $55.1 million last year to around $42.1 million this quarter. Although total sales increased, the funds left after direct costs for R&D, sales, and management expenses actually decreased, with margins in hardware especially narrowing. Deeper losses appear at the operating level: GAAP operating loss was $224 million, while the Core metric was only $33.11 million. The main cause of this massive gap is stock-based compensation, which affects both cost of revenue and various operating expenses, accounting for about 80% of the losses difference between the two metrics. While stock grants do not correspond to current cash outflows, they significantly impact shareholder equity distribution.

Additionally, the Core metric excludes certain data center pass-through revenue and costs and adds back amortization linked to warrants granted to certain customers, leading to higher Core revenues than GAAP revenues. This accounting adjustment means that for every $100 of GAAP revenue, about $106 in operating loss is recorded, but the adjusted metric, though still a loss, is much less severe. The gap between the two ledgers essentially traces to customer contract terms and employee compensation structures, and investors should be wary of the potential long-term erosion of profitability due to such non-cash outflows.

Looking forward, the raised guidance masks significant delivery pressure and challenges in achieving break-even. In the summer, the company’s full-year Core revenue guidance was less than $900 million. The lower end of the new range now surpasses the upper limit of the old one, raising full-year revenue expectations by $65 million. The midpoint for fourth-quarter Core revenue guidance is $320 million, about 39% higher than the actual third-quarter revenue, setting extremely high delivery requirements for year-end.

At the same time, fourth-quarter Core operating margin is expected to be between negative 1% and positive 1%, approaching the operating break-even line. Another optimistic expectation comes from a huge order backlog: unfulfilled contract obligations allocate $2.61 billion in revenue, with a considerable portion from the OpenAI agreement; total contracted data center power capacity is 700 megawatts, expected to be delivered by the end of 2027.

Additionally, Core gross margin guidance is 47% to 49%, higher than the third quarter’s 41%. This means not only must the company deliver more revenue in the fourth quarter, every dollar of revenue must also generate more gross profit. How far from break-even the company remains depends on whether both these targets can be achieved simultaneously. Revenue from contracts must be recognized steadily with computing power deliveries; any delay or gross margin disappointment could make the break-even goal elusive. Following the recent slowdown in hardware sales, this is Cerebras’ second major transformation challenge, which will determine whether the company can truly cross the chasm from a chip manufacturer to a cloud service provider.