Cerebras Systems released its first earnings report as a public company on June 23, 2026, and the numbers are hard to ignore. Revenue hit $193.4 million in Q1 2026, up 92% from the same quarter a year ago. For a company that only went public weeks earlier in what became the largest semiconductor IPO in history at $6.4 billion, it was a statement quarter.

But the headline result was not just the revenue. Cerebras confirmed a multi-year agreement with OpenAI for 750 megawatts of compute capacity valued at more than $20 billion, and announced a new partnership with Amazon Web Services to bring its fast inference capabilities directly to AWS customers.

The stock fell roughly 10% in after-hours trading despite the strong results. Investors appeared focused on gross margin compression in the forward guidance, a signal that scaling up to meet demand from these landmark deals will require significant upfront investment before the economics normalise.

What Cerebras Actually Does

Cerebras builds wafer-scale AI chips that take a fundamentally different approach to AI compute. While Nvidia stitches together thousands of GPU cores across multiple chips and manages the interconnects between them, Cerebras built the entire processor on a single wafer the size of a dinner plate. The CS-3 chip contains 4 trillion transistors, 900,000 cores, and 44 gigabytes of on-chip memory.

The practical result is inference that runs dramatically faster for certain workloads, particularly large language model serving where bandwidth between memory and compute is the bottleneck. When an enterprise needs to run millions of AI queries per hour and latency matters, Cerebras’s architecture offers something Nvidia cannot easily replicate.

That is why OpenAI signed a deal measured in gigawatts rather than GPUs.

The OpenAI and AWS Deals Explain Everything

The $20 billion OpenAI deal alone is larger than most enterprise software companies generate in total lifetime revenue. It confirms that the hyperscalers and frontier AI labs are not simply buying commodity compute. They are signing long-term capacity commitments with alternative infrastructure providers, hedging against Nvidia’s dominance and locking in pricing before demand outstrips supply further.

The AWS partnership adds a distribution layer that will matter significantly for enterprise adoption. Most enterprises already have AWS accounts, procurement relationships, and security reviews in place. Putting Cerebras inference on Bedrock-adjacent infrastructure removes the friction of evaluating an entirely new vendor relationship for every AI workload.

This is the emerging pattern in AI infrastructure: labs build on top of it, hyperscalers distribute it, and enterprises access it through familiar channels.

Q1 2026 By the Numbers

• Core revenue: $191.3 million (record)
• Year-over-year growth: 92%
• GAAP net loss: $14.0 million (narrowed from $23.9 million a year earlier)
• Core net loss: $2.5 million (near break-even)
• Full-year 2026 guidance: $855 million to $865 million in core revenue, representing roughly 69% growth at the midpoint

The narrowing loss while nearly doubling revenue is a sign the business model is maturing. Cerebras is not burning cash to buy growth. The unit economics are moving in the right direction even as the company scales a capital-intensive hardware business.

The guidance for 69% growth at the midpoint for the full year suggests the company expects H2 2026 to be bigger than H1 given the Q1 base rate. That would imply Q2 through Q4 acceleration, likely tied to the OpenAI capacity ramp beginning.

What the Stock Reaction Tells You

A 10% after-hours decline on a 92% revenue quarter is a market signal worth reading carefully. Investors are not questioning the growth. They are questioning whether the gross margin trajectory holds as Cerebras scales to fill the OpenAI and AWS commitments.

Manufacturing at wafer scale is expensive. The CS-3 chip requires TSMC’s most advanced processes. As Cerebras ramps production to meet multi-gigawatt demand commitments, the short-term manufacturing costs will pressure margins even if the long-run economics are excellent.

This is the same tension every hardware company faces at inflection. The market is pricing in execution risk, not doubting the opportunity.

What This Means for Business

For enterprise data and AI teams, the Cerebras results reinforce three conclusions.

Inference speed is becoming a competitive variable. The fact that OpenAI signed a 750MW deal specifically with Cerebras for fast inference means that latency at scale is not a solved problem on commodity hardware. Enterprise applications where AI response time matters, such as customer-facing voice agents, real-time analytics, or automated decision workflows, will increasingly live or die on inference throughput.

AI infrastructure is consolidating into partnerships, not point products. The Cerebras-AWS partnership means enterprises will access alternative compute through existing cloud relationships. The days of evaluating every AI hardware vendor as a standalone decision are ending. Your cloud contracts are becoming your AI infrastructure strategy.

The economics of AI compute are shifting. Cerebras’s near-breakeven core profitability at $191 million quarterly revenue, and its revenue guidance suggesting the full-year number could be nearly five times a typical enterprise SaaS company’s ARR growth, suggests the AI chip market is generating real economics faster than most predicted. That has downstream effects on pricing for enterprise customers as competition with Nvidia intensifies.

For businesses running data operations or deploying AI agents across workflows, this is the backdrop: infrastructure costs are scaling, speeds are increasing, and the infrastructure layer is establishing itself around a handful of platforms. Picking your AI infrastructure stack in 2026 is not a technical decision. It is a strategic one.

---

Enterprise DNA helps businesses navigate AI adoption across data literacy, agent deployment, and strategic decision-making. Explore Omni Advisory if you need help building your organisation’s AI infrastructure strategy.