Sail Research has launched from stealth with $80 million in combined seed and Series A funding to build what it calls max-efficiency infrastructure for long-horizon AI agents. Announced on June 25, 2026, the round was led by Kleiner Perkins at a roughly $450 million valuation, with Sequoia, Redpoint, Theory Ventures, Vine Ventures and CRV participating.
Infrastructure built for agents, not chat
Sail argues today's AI infrastructure was designed for brief, turn-by-turn chats, not agents that work autonomously over hours and days. The company notes enterprise AI bills have tripled even as per-token prices fell, because agentic workflows consume tokens 50 to 500 times faster than simple chat. Its answer is an inference stack rebuilt for throughput that promises up to 10x lower cost per token.
Sailboxes for day-long runs
The second pillar is Sailboxes, a stateful sandbox environment designed to keep agents running for days rather than seconds. Cofounder Neil Movva, an engineer who has worked across NVIDIA, Apple and Together AI, is now betting the next bottleneck is agent-grade serving. The launch lands amid heavy investment in AI compute efficiency, from Baseten to Positron AI.
Cutting the cost of autonomy
By targeting the economics of long-running agents, Sail is wagering that efficient inference will decide which agentic products reach scale, a race that also includes purpose-built silicon like d-Matrix's Corsair.
Reporting based on coverage from Fortune, PR Newswire and CTOL Digital.