Feldera cuts AI warehouse compute by 95%

Feldera raised $21.5 million across a Series A and seed round, and says its incremental SQL engine has cut warehouse compute bills by 95% or more while shrinking freshness from hours or days to seconds. The company’s announcement says the engine keeps SQL views current by recomputing only changed data, and is already used to secure agents at scale, power real-time logistics views, and detect fraud. Builders running retrieval, fraud, or agent-context pipelines should benchmark incremental views against batch refresh before paying for fresher data; the agent search cost benchmark shows how quickly infrastructure plumbing can dominate an agent bill.