Cerebras posts $193m Q1 revenue amid $20bn OpenAI deal and record IPO
Cerebras Systems, the wafer-scale AI chip company, reported GAAP revenue of $193.4 million for the first quarter of fiscal 2026, a 94% increase year-on-year, as a string of landmark deals and a record-breaking public listing transformed its capital position. The results cover the three months ended 31 March 2026; Cerebras completed its IPO in Q2, raising $6.4 billion in gross proceeds.
Core revenue, the company's non-GAAP measure that strips out data-centre pass-through costs and warrant amortisation, reached $191.3 million, up 92% from the same period a year earlier. Cloud and other services revenue nearly tripled year-on-year on a GAAP basis, reaching $82.8 million, while hardware revenue rose 59% to $110.6 million. The company posted a GAAP net loss of $14.0 million, a marked improvement on the $23.9 million loss recorded in Q1 2025, and generated positive operating cash flow of $12.3 million.
The deals driving growth
The quarter's most consequential announcement was a multi-year agreement with OpenAI covering 750 megawatts of high-speed inference compute, valued at more than $20 billion. Alongside that, Cerebras said it had begun a partnership with Amazon Web Services to bring its inference capability to AWS's global customer base. The two companies are pursuing what Cerebras describes as a disaggregated inference architecture, in which AWS's Trainium 3 chips handle the prefill stage of inference and Cerebras's CS-3 system manages the decode step.
The company also reported customer trials of Kimi K2.6, described as the first trillion-parameter open-weight model served on its hardware, achieving close to 1,000 tokens per second as measured by the independent benchmarking service Artificial Analysis. Google DeepMind's Gemma 4 31B model was also launched on the platform. Cerebras and OpenAI co-launched Codex-Spark, a coding-focused model targeting interactive, low-latency use cases and delivering more than 1,000 tokens per second.
Andrew Feldman, co-founder and chief executive, said: "The growing importance of AI in our economy requires AI infrastructure that can power the most advanced applications at unprecedented speed. This is the Cerebras mission."
Market context and competitive position
Cerebras occupies a distinctive position in the AI accelerator market. Where Nvidia's GPU clusters dominate model training and are the standard inference substrate for most cloud workloads, Cerebras bets that its monolithic wafer-scale engine, which integrates the entire chip onto a single silicon wafer, can deliver lower per-token latency for autoregressive decode tasks. That proposition is increasingly relevant as large language model deployments shift from batch processing toward real-time, agent-driven workflows where response speed is a product differentiator.
The OpenAI and AWS partnerships lend significant commercial credibility to that thesis, though the concentration risk is notable: the forward-looking disclosures in the release explicitly name OpenAI, AWS, Group 42 and the Mohamed bin Zayed University of Artificial Intelligence as significant customers. Customer concentration at this stage of growth is a standard scrutiny point for institutional investors evaluating the durability of the revenue base.
Cerebras guided for Q2 core revenue of approximately $194 million, representing 88% year-on-year growth, but projected core gross margins of 36 to 38%, a step down from Q1's 47%, likely reflecting the cost profile of the data-centre expansion underpinning the OpenAI capacity commitment. Full-year 2026 guidance calls for core revenue of $855 million to $865 million, implying roughly 69% growth at the midpoint, alongside operating losses in the range of 28 to 32% of revenue. Sustained profitability remains a medium-term target rather than a near-term one, a point that equity analysts will weigh alongside the headline growth rate as the post-IPO lockup period approaches.