# Cerebras bets speed can triple sales _Cerebras posted record Q3 revenue on surging inference demand and guided higher, testing if speed converts to scale._ **Published:** 2026-11-12 **Source:** https://www.startuphub.ai/ai-news/public-companies/2026/cerebras-bets-speed-can-triple-sales --- [Cerebras](https://investors.cerebras.ai/news-releases/news-release-details/cerebras-systems-delivers-record-revenue-third-quarter-2026) posted record third quarter revenue of $210.8 million on a GAAP basis and $230.9 million on its core metric, with cloud sales up more than 260% from a year ago. Will buyers pay extra for faster tokens? For now, the answer is yes. Inference is the step where a trained model generates answers for a user, and [Cerebras](https://www.startuphub.ai/ai-news/startup-news/2026/cerebras-systems-eyes-4-8-billion-ipo-amid-ai-chip-demand) argues that faster inference lets developers build agents that act in real time. Core cloud revenue hit $144.9 million in the quarter, up 272% year over year, the company said, while its pipeline for high-speed inference grew more than 10x. Andrew Feldman, co-founder and CEO whose bet is wafer-scale chips, said [Cerebras](https://www.startuphub.ai/ai-news/startup-news/2026/cerebras-systems-eyes-4-8-billion-ipo-amid-ai-chip-demand) remains the "undisputed champion of speed" and expects to triple revenue in 2027. That growth is forcing cloud operators and enterprise teams to rethink deployment budgets. Oracle will integrate Cerebras inference into Oracle Cloud Infrastructure in 2027, the company said, and Cerebras announced a cloud deal with a top-three hyperscaler with details due next year. Bilt will use Cerebras to power its Neighborhood OS and Concierge assistant for eight million consumers and 50,000 businesses, the company said. Can Cerebras deliver the capacity behind those contracts? Disaggregated inference is its technique that splits model work across different processors to get more tokens from the same power, and Cerebras said it is the first to run it at production scale with up to 5x more tokens per data center. The company said it has 700 gross megawatts under contract for delivery by the end of 2027, added capacity in Norway and Finland, and is scaling manufacturing with Flex and Sanmina to exit 2026 with more than 10x the capacity it started with. The limit is profit. GAAP gross margin was 20% and GAAP operating margin was negative 106%, the company said, with a net loss of $161.2 million, while core gross margin was 41% and core operating margin was negative 14%. Bob Komin, CFO responsible for guiding the company toward breakeven, said Cerebras beat guidance for the third straight quarter. Fourth quarter core revenue is guided at $310 million to $330 million with margins near breakeven, and full-year core revenue at $940 million to $960 million, the company said. The market was cautious, with shares closing at $93.66 before the print and slipping about 3% in pre-market trading, per CNN data as of the earnings day. Watch CS-4 production, the Wafer Scale Engine 5 now in system testing for a CS-5 targeted for early summer 2027 at up to 10,000 tokens per second per user, and whether $26.1 billion in remaining performance obligations as of September 30 converts to cloud use. --- Original analysis from [startuphub.ai](https://www.startuphub.ai), the #1 AI startup directory. © StartupHub.ai. All rights reserved. You may not republish this article in full without a license. Search engines and AI research tools may crawl and summarize for reference. Bulk reproduction or model training on this content requires a license. See https://www.startuphub.ai/terms.