AIBRIEF
The infrastructure startup says the real test of an AI agent isn't the demo. It's whether the system keeps working days or weeks later when something breaks.
Temporal, the company that sells infrastructure for long-running software workflows, raised $550 million in a Series E that values it at $12.55 billion. Lightspeed co-led the round with Wellington Management, Growth Equity at Goldman Sachs Alternatives, and Tiger Global; T. Rowe Price and SV Angel added strong participation. Returning backers include Andreessen Horowitz, Sequoia, Index, GIC, Sapphire Ventures, and Amplify.
The pitch: customers now want AI agents that run for days, weeks, or months, not just for the length of a chat. Temporal's product, which the company calls "durable execution," is the plumbing that lets those workflows pause, survive a crash, and pick up where they left off without a developer rebuilding the job by hand. Named customers include OpenAI, Snap, NVIDIA, and JPMorgan Chase; case studies cover Replit's agent reliability work.
The growth numbers are self-reported: annualized revenue run rate up more than 200% year-over-year, net dollar retention above 200% since February, and 4,300 paying customers, up 139% year-over-year. The platform processed 1.9 trillion billable actions in August, and open-source installs passed 43 million.
The company itself raises the harder question: a working agent demo is cheap to build. The real test is whether people trust the agent enough to keep using it when something breaks weeks into a job. Community reaction on Hacker News is split on whether the metrics support the multiple.