NVIDIA Corporation rival Cerebras (CBRS.US) Q3 revenue hits record high, cloud business revenue surges 265% year-over-year.

AI chipmaker Cerebras Systems (CBRS.US) reported its fiscal 2026 third-quarter results on Thursday. Driven by strong growth in demand for high-speed AI inference, the company's quarterly revenue hit a record high and it raised its full-year guidance. At the same time, its fourth-quarter revenue outlook significantly exceeded Wall Street expectations, indicating further breakthroughs in the AI chip market dominated by NVIDIA Corporation (NVDA.US). Following the earnings release, Cerebras shares rose more than 15% in after-hours trading. As of Thursday's close, the stock had fallen 47% year to date. The results showed that Cerebras' third-quarter revenue under U.S. generally accepted accounting principles (GAAP) reached $210.8 million, up 55% year-over-year, a record high. On a non-GAAP basis, third-quarter core revenue reached $230.9 million, up 70% year-over-year, also a record high. On profitability, the company's third-quarter GAAP net loss was $161.2 million, significantly wider than $22.2 million in the prior-year period; diluted loss per share was $0.67, compared with $0.41 in the prior-year period. However, on a non-GAAP basis, third-quarter core net income was $29.82 million, compared with a net loss of $10.86 million in the prior-year period, achieving a year-over-year turnaround to profit. The company's third-quarter GAAP gross margin was 20%, down from 40.6% in the prior-year period but improved from the previous quarter. Core gross margin was 41%, up from 40.7% in the prior-year period. Third-quarter GAAP operating loss was $224.1 million, compared with $26.47 million in the prior-year period; core operating loss was $33.11 million, compared with $15.12 million in the prior-year period. Core operating margin was negative 14.3%. It is worth noting that stock-based compensation expenses and amortization of customer warrants were important reasons for the large difference between GAAP and core results. The company's third-quarter stock-based compensation expense reached $155.1 million, and amortization of customer warrants was $36.42 million. Q4 revenue guidance significantly exceeds expectations; 2027 revenue expected to grow to three times this year's level Looking ahead to the fourth quarter, Cerebras expects non-GAAP core revenue to be between $310 million and $330 million, with a midpoint of $320 million, about 18% above the analyst average estimate of $271.2 million compiled by media. Based on the midpoint of the guidance, the company's fourth-quarter core revenue is expected to grow about 39% from the third quarter. Cerebras also expects fourth-quarter core gross margin to be between 47% and 49%, up from 41% in the third quarter; core operating margin is expected to be between negative 1% and positive 1%, meaning the company's core business is expected to approach break-even in the fourth quarter. For full-year 2026, Cerebras raised its outlook and expects core revenue of $940 million to $960 million, core gross margin of 42% to 44%, and core operating margin of negative 10% to negative 8%. Cerebras co-founder and CEO Andrew Feldman said the company is providing customers with more channels to use its technology through direct sales of hardware systems, its own cloud computing platform, and third-party cloud service providers, and he expects revenue in 2027 to grow to three times the 2026 level. Chief Financial Officer Bob Komin said the company has exceeded its own guidance across the board for three consecutive quarters, with all core financial metrics improving sequentially in the third quarter. The company also won important customers including large cloud service providers, emerging AI cloud companies, and large enterprises, and the scale of potential business opportunities in its high-speed AI inference business expanded to more than 10 times the previous level during the quarter. Cloud computing revenue surges 265% year-over-year to become the main growth engine Cerebras currently has two main businesses: selling computer systems equipped with its self-developed AI chips, and providing cloud computing services based on its own chip technology. By segment, the company's third-quarter GAAP cloud computing and other services revenue reached $141.9 million, up 265% year-over-year, a record high, accounting for about 67% of total revenue. On a non-GAAP basis, core cloud computing and other services revenue reached $144.9 million, up 272% year-over-year. Core gross margin for this business rose sharply to 46.9% from 16.5% in the prior-year period. By contrast, the company's hardware business came under pressure. Third-quarter GAAP hardware revenue was $68.88 million, down about 29% from $96.79 million in the prior-year period; core hardware revenue was $85.96 million, down about 11% year-over-year. GAAP gross margin for the hardware business fell to 11% from 50.4% in the prior-year period, while core gross margin fell to 31% from 50.5%. The rapid expansion of the cloud computing business shows that Cerebras is gradually shifting from mainly relying on selling AI hardware systems to a business model in which hardware sales and cloud computing services run in parallel. Wins orders from two large cloud service providers; remaining performance obligations reach $26.1 billion On customer expansion, Cerebras disclosed that it secured commercial partnerships with two new hyperscale cloud service providers in the third quarter. Among them, Oracle Corporation (ORCL.US) is integrating Cerebras' high-speed AI inference technology into its cloud infrastructure platform (OCI), with related services expected to launch in 2027, providing high-speed inference capabilities to AI labs, enterprises, and developers. In addition, Cerebras announced an important cloud computing partnership with one of the world's top three large cloud service providers, with more details expected to be disclosed next year. The company is also expanding the reach of its technology by partnering with emerging AI cloud service providers and reducing dependence on a single sales channel. As of the end of September, Cerebras' total remaining performance obligations reached $26.1 billion, reflecting the scale of future revenue corresponding to signed contracts. To meet growing AI computing demand, Cerebras has contracted 700 megawatts of data center capacity, expected to be delivered by the end of 2027. The company has also signed additional data center capacity contracts in Norway and Finland to expand its European footprint. On manufacturing, Cerebras is working with Flex and Sanmina to expand capacity, with its manufacturing capability expected to reach more than 10 times the level at the start of the year by the end of 2026, and it plans to add factories in 2027 to further increase capacity. Accelerating its challenge to NVIDIA Corporation; AI chip competition enters a new phase Cerebras, which completed its initial public offering (IPO) in May this year, is trying to expand its share of the AI chip market dominated by NVIDIA Corporation with its unique wafer-scale chip design and to benefit from the global data center infrastructure investment boom. Unlike traditional GPU designs, Cerebras uses a wafer-scale processor architecture that integrates large-scale computing units onto a single wafer to provide high-performance computing capability for AI models. In August this year, Cerebras launched its next-generation AI computing system, CS-4. The company said the system can deliver inference speed improvements of up to 30 times and will soon enter full production. At the same time, the company also showcased its next-generation computing system, CS-5, planned for launch in early summer 2027, with a goal of achieving generation speeds of up to 10,000 AI text processing units (tokens) per second per user. The new-generation wafer-scale processor WSE-5 used in the system has completed initial debugging and is currently undergoing system testing. Cerebras has also launched a heterogeneous disaggregated inference solution for production environments, which improves the inference capability per unit of power in data centers and reduces costs for AI users by optimizing the allocation of computing resources across different inference stages. Cerebras' strong fourth-quarter outlook provides a positive signal for its strategy of promoting its new chip architecture and challenging NVIDIA Corporation's market dominance. At the same time, NVIDIA Corporation is facing uncertainty from management changes. The company announced on Thursday that it had removed Jensen Huang from the role of chief executive officer, an unexpected development that has drawn more attention to the competitive landscape of the AI chip industry.