Chai AI’s 5,000+ GPU cluster
Chai AI deployed a dedicated 5,000+ GPU cluster for model alignment and safety research, framing the move as ‘Safe Scaling’ that couples large‑scale compute with regulatory compliance demands. The deployment signals that safety‑oriented infra scale is becoming a competitive differentiator. (prunderground.com)
Chai AI issued a company press release via PR Newswire dated March 25, 2026, filed from Palo Alto, California. (prnewswire.com) The announcement cites 3× annual revenue growth, roughly $70 million in annual recurring revenue, and an estimated $2.1 billion company valuation. (financialcontent.com) Coverage and the release name specific safety capabilities under development, including real‑time suicide and self‑harm classifiers and platform-wide age‑gating policies. (tmcnet.com) Chai previously publicized a major compute milestone on June 11, 2025, reporting cluster performance that it characterized at about 1.4 exaFLOPS. (prnewswire.com) Chai’s own cluster status page lists sustained throughput claims of roughly 1.2 trillion tokens served per day and about 51,000 unique LLM instances handled, alongside the >1.4 exaFLOPS compute figure. (chai-research.com) The March 25 release was syndicated across outlets including Yahoo Finance and FinancialContent, and the PR lists Tom Lu as press contact with phone +1 (626) 594‑8966. (finance.yahoo.com)
