Traversal Raises $120M in Series B Funding to Build AI Outage Engineers

Traversal raises $120M to fix outages with AI

New York startup Traversal raised $120 million in Series B funding at a $1 billion valuation, announced November 3, 2026, to build AI outage engineers: autonomous agents that investigate software outages, find their root cause, and help teams fix them before a bad night gets worse. Kleiner Perkins, Redpoint Ventures, and Khosla Ventures co-led the round, with Sequoia Capital, Battery Ventures, and Amex Ventures participating. The raise brings Traversal's total funding to $173 million, two and a half times what it raised when it emerged from stealth in June 2025.

The pitch lands because the problem is getting worse, not better. AI coding tools now generate software faster than ever, which means more code reaches production, which means more things can break at 2 a.m. Companies already spend enormous amounts on downtime: Traversal cites research putting the cost of unplanned outages at $600 billion a year for Global 2000 companies, up 50 percent in two years, with engineers spending roughly seven hours a week troubleshooting. When the pager goes off, the expensive part is not the fix. It is the hours spent figuring out what broke.

Traversal's customers include Nubank, Harvey, DigitalOcean, and PepsiCo, plus unnamed Fortune 100 financial institutions. That list matters because it shows the product is being tested where downtime costs real money, not just in demos. The company says its monthly investigations grew from 513 to nearly 73,000 this year, and it reports more than 90 percent root-cause accuracy on high-impact incidents plus an 85 percent average reduction in mean time to recovery on critical incidents.

What AI outage engineers do all day

Site reliability engineering is the discipline of keeping software running. When a service goes down, engineers sift through logs, metrics, and traces to find what changed and what failed. Traversal wants to hand much of that investigation to AI agents, combining agentic systems with causal machine learning so the software can tell a genuine cause apart from a mere correlation. Two ideas sit at the center: a Production World Model that turns telemetry, code, and deployment history into a continuously updated picture of a production system, and a Causal Search Engine that tests thousands of hypotheses in parallel.

The product lineup expanded in September 2026, when Incident Workers became generally available and Alert Workers entered public beta. Incident Workers investigate the moment an anomaly triggers; Alert Workers triage the flood of alerts that usually buries the important one. That division is sensible, because alert fatigue is a real failure mode: when everything pages, nothing gets attention, and the critical alert drowns in noise.

There is a benchmark behind the pitch, too. On ORCA-bench, a test Traversal built with Cornell Tech and Columbia University, the best of five frontier agents managed only 25.3 percent accuracy on medium-difficulty incidents, falling to 10 percent when reports were vague. Traversal's argument is that general-purpose agents are weak at this job precisely because they lack a model of the production system they are investigating. Whether its own numbers hold up under independent testing is the question investors are betting $120 million on.

The skeptical read on a $1 billion valuation

Every number in the funding announcement that sounds impressive comes from Traversal itself, and that deserves a plain warning. Accuracy and recovery figures have not been independently verified, and growth multiples look dramatic partly because the starting base was small. A jump from 513 to 73,000 monthly investigations is real usage growth, but it is not the same as revenue growth, and the company reports ninefold revenue growth year to date without disclosing the underlying figure.

There is also a safety question that no funding announcement answers. An AI agent that can diagnose an outage is one step away from an AI agent that applies a fix, and an unsuitable automatic fix can make a problem worse. The sensible deployment pattern is human approval for anything that changes production, with the agent doing the investigation and proposing the repair. Traversal's framing keeps humans in the loop for now, and buyers should insist on that: autonomous diagnosis is a gift, autonomous remediation without guardrails is a risk.

Even with those caveats, the round says something real about where startup money is going. Investors are shifting from AI that chats toward AI that does operational work: fixing outages, running infrastructure, and keeping systems alive. That is less glamorous than a chatbot, and far closer to where companies actually lose money. If Traversal can prove its accuracy claims outside its own marketing, a $1 billion valuation will look cheap. If it cannot, this will be remembered as another AI round priced on a story. The next year of customer results, not the next funding announcement, will settle it. For context on how AI agents are changing developer work, see our coverage of GitHub's agentic workflows and GitHub Octoverse 2026.