Cerebras Q2 Revenue Misses, Cloud Growth Surges 281%
Cerebras Systems reported Q2 FY2026 revenue of $180.1 million, missing the $194 million consensus estimate but still growing 74% year-over-year. The miss was driven by a sharp decline in hardware revenue, which fell 23% to $54.1 million versus the expected $73.1 million. Cloud revenue, however, soared 281% to $126.0 million, beating estimates of $116.3 million.
Core Revenue Growth Masks Margin Compression
Core revenue, which excludes hardware sales, reached $209.9 million, up 103% year-over-year. This figure exceeded the total reported revenue, reflecting a strategic shift toward recurring cloud services. Core gross margin improved to 41%, up 940 basis points year-over-year, but overall gross margin landed at just 14%, far below the 26.3% consensus.
The stark gap between core and overall margins highlights the drag from hardware. As Cerebras pivots to cloud and AI inference, hardware sales are becoming a smaller, less profitable component. Operating expenses ballooned to $502.8 million, up from $89.3 million in the prior year, driven by R&D and sales expansion. Core operating margin improved to -16%, a 2,600 basis point year-over-year gain, but remains deeply negative.
Raising Guidance Despite Q2 Miss: What It Signals
Management raised full-year FY26 core revenue guidance to $880-$890 million, above the $867.6 million consensus. Q3 core revenue is guided to $214-$216 million, also ahead of the $212 million estimate. The raise suggests that the cloud segment’s momentum will continue to offset hardware weakness.
However, core gross margin guidance for Q3 of 38%-40% is below the 41% reported in Q2, indicating near-term margin pressure. Core operating margin guidance of -25% to -23% for Q3 is worse than Q2’s -16%, implying increased investment in capacity and partnerships. This is a deliberate trade-off: sacrificing near-term profitability for long-term market share in AI inference.
Why Cerebras Avoids HBM and CoWoS Bottlenecks
Cerebras differentiates itself by avoiding supply-constrained components like HBM memory, CoWoS packaging, and 3nm process nodes. This strategic choice allows the company to scale manufacturing more rapidly than competitors reliant on these scarce inputs. Manufacturing capacity is set to increase more than 10x in 2026, supporting the 600 MW of data center capacity already live or under contract.
The approach is validated by key partnerships: OpenAI selected Cerebras as a launch partner for GPT-5.6 Sol, achieving 750 tokens per second. AMD will deploy disaggregated inference in production in Q4 2026, promising up to 5x throughput, with AWS bringing the same to Amazon Bedrock in Q1 2027. These deals de-risk the revenue model and provide credibility in the competitive AI chip market.
Balance Sheet Strength Backs Ambitious Expansion
Liquidity stands at $8.6 billion, plus $850 million in debt capacity, giving Cerebras ample runway to fund its expansion. Remaining performance obligations reached $25.4 billion, a massive backlog that underscores future revenue visibility. GAAP EPS was -$2.98, reflecting heavy investment, but the cash position mitigates near-term solvency concerns.
Investors should watch whether core gross margin can hold above 40% as scale increases. The Q3 guidance dip to 38%-40% may be temporary, but if margins fall further, the path to profitability lengthens. The company’s success hinges on converting its backlog into high-margin cloud revenue while managing hardware declines.
What to Watch: Margin Trajectory and OpenAI Deal Ramp
The next catalyst is the Q4 earnings report, where investors will see if core gross margins recover to 41% or better. Also, monitor any updates on the OpenAI partnership, particularly token throughput metrics and revenue contribution. If Cerebras can demonstrate sustained cloud growth with improving margins, the stock could re-rate despite current losses.
Conversely, a continued hardware slump or margin erosion would raise doubts about the business model. The $25.4 billion backlog is the key number to track—if it grows, the thesis holds; if it stagnates, competition from Nvidia and AMD may be biting.
Comments are closed.