Cerebras Posts Record Q3 Revenue as Microsoft Signs On as Third Hyperscaler Partner

Cerebras Systems reported record third-quarter results on Wednesday, beating its own guidance by $16 million and announcing that Microsoft had signed on as its latest hyperscaler customer — a milestone that extends the AI inference company’s reach across all three of the world’s largest cloud providers.

Core revenue for the quarter reached $230.9 million, up 70% from $135.7 million a year earlier and ahead of the company’s $214–216 million guidance midpoint. GAAP revenue came in at $210.8 million, a 55% year-over-year gain, narrowly missing the $214.7 million analyst consensus.

Cloud revenue, the company’s highest-growth segment, surged 265% year over year to approximately $144 million. Hardware revenue fell from $96.8 million to $68.9 million over the same period, reflecting an ongoing mix shift toward inference-as-a-service over direct chip sales.

Microsoft signed a commercial agreement with Cerebras in October, joining Amazon Web Services and Oracle to give the company relationships with all three major US hyperscalers. CEO Andrew Feldman said revenue from the Microsoft deal is expected to begin showing up in Q1 2027. The company also said it is evaluating financing options to keep pace with rising demand.

The fourth-quarter outlook came in well above expectations: core revenue guidance of $310 million to $330 million — versus the $271 million analyst consensus — and core gross margin of 47% to 49%. For the full year 2026, the company raised its core revenue guidance to $940–960 million.

Looking further ahead, Cerebras said it expects revenue to roughly triple in 2027 to approximately $2.85 billion, in line with the $2.92 billion LSEG analyst estimate. Remaining performance obligations reached $26.1 billion, suggesting contracted demand is building faster than the company can currently recognize as revenue.

Shares surged 17% in after-hours trading following the announcement, though Cerebras stock remains down roughly 47% from its May 2026 Nasdaq IPO price. The company competes with Nvidia in AI inference, targeting customers who prioritize throughput and cost efficiency over training workloads. Landing all three major US hyperscalers within a single quarter signals that those customers are actively diversifying their AI compute supply chains beyond a single supplier.

As reported by CNBC.