Cerebras Systems Inc. (CBRS) Earnings

Management highlights

- Market & Technology Leadership • AI has exponentially expanded the addressable compute market by enabling computers to process unstructured real-world data (text, images, video, robotics) that was previously intractable, and demand for AI compute will accelerate for decades. • Speed is the critical competitive advantage in AI inference: faster inference increases productivity, drives higher adoption, enables new use cases, and eliminates the trade-off between safety guardrails and user experience. • Cerebras' unique wafer-scale architecture delivers inference that is up to 13x faster than leading GPUs, avoids reliance on constrained HBM memory, and has durable long-term advantages for future innovations including memory stacking and optical integration. • Demonstrated 13x faster inference than leading GPUs on the 1 trillion parameter Kimi K2 open source model, completing the same prompt in 21 seconds versus 4 minutes 37 seconds for the competitor. - Key Customer Partnerships • Signed a definitive multi-year agreement with OpenAI for over $20 billion of Cerebras compute in December 2025, achieved production deployment in 35 days from signing, and GPT-5.4 is now running on Cerebras infrastructure for OpenAI engineers and select customers. Work is ongoing to deploy GPT-5.5. • Completed a definitive agreement with AWS to deploy Cerebras CS3 in AWS data centers for a disaggregated inference solution, where AWS Trainium 3 handles pre-fill and Cerebras handles sequential decode, targeting an order of magnitude faster performance. This partnership enables Cerebras to reach enterprise customers where they already operate. - Supply Chain & Capacity Expansion • Cerebras avoids the key industry supply chain constraints: it does not use HBM, does not rely on TSMC's CoWoS process, and manufactures at the less constrained 5nm node, resulting in more stable and lower-cost manufacturing. • Secured long-term wafer supply from TSMC, expanded U.S.-based manufacturing with two contract manufacturers (Flextronics and Sanmina), and added hundreds of thousands of square feet of new cleanroom and manufacturing space. • Has added new data center capacity across North America and Europe, and is in early discussions for additional capacity in the Middle East, Asia, and Australia to meet strong customer demand. • Is temporarily renting back systems from an existing customer to accelerate capacity deployment to meet contracted backlog while new owned data centers are built out.

Guidance

- Q2 FY2026 guidance: core revenue of ~$194 million (88% year-over-year growth); core gross margin of 36% to 38%; core operating margin of -30% to -32%. - Full year FY2026 guidance: core revenue of $855 million to $865 million (69% year-over-year growth at the midpoint); core gross margin of 38% to 41%; core operating margin of -28% to -32%. - Cloud and services revenue is expected to see increasing year-over-year growth rates each quarter in 2026, with revenue ramping significantly in the second half of the year as cloud capacity deployment accelerates. - Hardware revenue is expected to decrease over the next few quarters as production mix shifts toward cloud deployment for contracted customers, though this trend can change quickly based on customer deployment choices. - Temporary third-party capacity rental is expected to reduce core cloud and services gross margin by 10 to 15 percentage points in the near term, before margins ramp back to a target of 60%+ once owned data center capacity is deployed. - Core hardware gross margin is expected to normalize back to the low 30% range after the Q1 2026 elevated margin from one-time contract pricing incentives. - Long-term targets are 60% total gross margin and 40% operating margin, achieved via scale economies, improved manufacturing efficiency, higher cloud utilization, and performance-based pricing.

Segment performance

Core total revenue for Q1 FY2026 was $191.3 million, representing 92% year-over-year growth. Core hardware revenue reached $111.6 million, accounting for 58.3% of total core revenue, with 60% year-over-year growth and a 42% gross margin. Core cloud and services revenue was $79.8 million, accounting for 41.7% of total core revenue, with 167% year-over-year growth and a 52.9% gross margin. On a non-GAAP basis, total operating expenses were $92.6 million, with R&D expenses of $69.8 million, sales and marketing expenses of $12.9 million, and G&A expenses of $9.9 million. Core non-GAAP operating loss was $3.5 million (an operating margin of -2%), and core non-GAAP net loss was $2.5 million. The company ended the quarter with $3.3 billion in cash, cash equivalents, restricted cash, and marketable securities.

Risks & headwinds

- Global data center capacity and power availability are extremely constrained, which is the primary binding constraint on Cerebras' growth, and delays in securing or building new data center capacity could slow revenue ramp. - Variable customer deployment choices (for OpenAI, between cloud and on-premise hardware) create uncertainty in revenue and margin projections, as pass-through data center costs impact margin levels depending on deployment choice. - Temporary capacity rental to meet near-term demand will depress gross margins for the next several quarters until new owned capacity comes online. - Forward-looking results are subject to inherent uncertainties, and actual performance could differ materially from projections due to market, supply chain, and customer-related factors, as detailed in the company's SEC filings.

Analyst Q&A

• Q: What is the expected timeline for revenue impact from the definitive AWS agreement, and do multiple disaggregated partners mean there are other large customers beyond AWS? /

  A: Cerebras has secured sufficient TSMC supply to cover its 2026 plans and beyond, so AWS partnership revenue is expected to start impacting results in 2027. The opportunity to provide specialized decode hardware for other companies with existing GPU infrastructure is real, so Cerebras is exploring multiple partnerships in this space beyond AWS.

• Q: What is your total addressable market for AI inference, and how much of the market will pay a premium for fast inference? /

  A: Historical technology trends show slow offerings do not hold meaningful market share over time—no one chooses slow search or dial-up internet when faster options are available. Cerebras views the entire inference market as its addressable market, since fast inference is uniformly more valuable to users. Fast inference is already priced at a premium across the industry, as users consistently prefer faster performance even at higher cost.

• Q: What is the primary constraint on Cerebras' growth right now, and what explains the better-than-expected gross margin outlook? /

  A: Demand and component supply are not constraints; the only binding constraint is data center capacity, and Cerebras is pursuing capacity from dozens of operators globally to meet strong demand. Gross margins are better than projected due to higher market pricing amid strong demand, rising competitor pricing driven by higher HBM costs, and more favorable timing and pricing for new capacity deployment than previously forecast.

• Q: What determines if OpenAI takes its committed capacity in Cerebras cloud versus as on-premise hardware? /

  A: Only OpenAI controls this decision, and it will depend on OpenAI's portfolio balance of existing data center capacity versus the capacity Cerebras can provide in its own cloud data centers. The AWS agreement calls for deployment in AWS data centers by default.