Snowflake posts AI surge — plus lawsuits
Snowflake reported strong Q4 AI‑driven product revenue growth while also facing multiple investor notices and a newly filed securities class action (finance.yahoo.com) (globenewswire.com). The revenue momentum increases demand for embeddings and inference tied to Snowflake pipelines even as legal noise creates corporate distraction.
Product revenue in Q4 hit $1.23 billion (up 30% year‑over‑year) while total Q4 revenue was $1.28 billion, according to the company’s fiscal Q4 FY2026 release reported)). Snowflake said more than 9,100 accounts were using Snowflake AI features and that Snowflake Intelligence reached roughly 2,500 accounts within three months, while the company introduced 430+ new capabilities across FY2026 announced)). Management guided Q1 FY‑27 product revenue to $1.262–$1.267 billion (about 27% YoY) and full‑year FY‑27 product revenue to roughly $5.66 billion (about 27% YoY), with Observe expected to add ~1 percentage point to growth, per analyst summaries of the earnings guidance outlined)). Snowflake’s backlog and enterprise traction tightened the story: remaining performance obligations rose to $9.77 billion (up 42% YoY), the company added 740 net new customers (up 40% YoY), and 733 customers now have trailing‑12‑month product spend above $1 million reported)). Multiple plaintiff firms have filed notices and suits: Robbins Geller, Kaplan Fox, and Bernstein Liebhard have flagged a securities class action covering purchases from June 27, 2023 through Feb. 28, 2024, alleging undisclosed headwinds from product efficiency gains, Iceberg Tables adoption and tiered storage pricing and noting the stock fell ~18% after the Feb. 28, 2024 disclosure alleged)). Snowflake has also been building embedding and inference plumbing inside the platform — native vector/embedding functions appear in product docs, Snowflake hosts Arctic embedding models on Hugging Face, and the Feb. 4, 2026 acquisition of TensorStax was positioned to accelerate agentic AI and autonomous data‑engineering pipelines that support high‑throughput embedding/inference workloads documented)).
