Traversal, a New York startup building an artificial-intelligence site reliability engineer, has raised $120 million in Series B funding at a $1 billion valuation, becoming the latest AI agent company to cross into unicorn territory. The round, announced on 3 November, was co-led by Kleiner Perkins, Redpoint Ventures and Khosla Ventures, with Sequoia Capital, Battery Ventures and Amex Ventures participating. Traversal says the money will fund hiring and its push toward what it calls a superintelligent AI SRE, a system that diagnoses, fixes and prevents production incidents.
The round and who backed it
The financing lifts Traversal’s total funding to $173 million, according to the company. It emerged from stealth in June 2025 with $48 million led by Sequoia and Kleiner Perkins, followed by an Amex Ventures investment. Kleiner Perkins, which led the Series A, increased its stake, while Sequoia, which led the seed round, used its growth fund to increase its position.
Traversal names Nubank, Harvey, DigitalOcean and PepsiCo among its customers, and American Express’s chief technology officer praised its work on production resilience. The company reports that annual recurring revenue has grown ninefold year to date, product usage is up nearly a hundredfold, and monthly active users are up nearly twentyfold. Those growth figures are company-reported, not audited.
What the AI SRE actually does
Traversal’s system sits inside a company’s production environment, analysing logs, metrics and traces to build a picture of normal behaviour. When something breaks, it uses causal machine learning to identify the root cause and proposes a fix, which a human engineer accepts before service is restored. Its Production World Model and Causal Search Engine track production systems and their dependencies.
The pitch resembles other always-on enterprise agents, such as Snowflake’s night-shift agent coworkers, but Traversal aims its agents at production incidents rather than data workflows. Where a traditional monitoring tool fires an alert and waits for a human, Traversal tries to close the loop itself, investigating the incident, tracing its cause and recommending the repair.
Why investors paid a unicorn price
The economics of downtime help explain the valuation. A 2026 Splunk and Oxford Economics report found downtime costs have risen by half in two years, with engineers losing seven hours a week to troubleshooting. Gartner projects 85 percent of enterprises will adopt AI SRE tooling by 2029, up from under 5 percent in 2025, although that is a forecast.
Traversal also argues general-purpose models are poor at this job. On ORCA-bench, co-developed with Cornell Tech and Columbia University, the best of five frontier agents scored 25.3 percent on medium-difficulty incidents and 10 percent when reports were vague. Hallucination rates reached 40 percent, but Traversal helped build the benchmark, so that yardstick is a company claim.
Our read
This round prices what Traversal calls the outer loop of software development. Coding agents write code faster than humans can maintain it, so the bottleneck moves to keeping it running. A $120 million bet at a $1 billion valuation says investors believe reliability is where agent value will concentrate, and Kleiner Perkins and Sequoia doubling down gives that thesis weight.
The claim still needs independent proof. Traversal’s growth numbers are self-reported, its benchmark is self-built, and Nubank describes its rollout as expanding wave by wave rather than complete. The human-approval step is telling: for now, the AI SRE recommends and the engineer decides, a guardrail for systems touching live production.
Why It Matters
Traversal lands in the same funding wave as Fal’s $16 billion inference round, but on the other side of the agent economy. Fal sells the compute that runs AI media; Traversal sells the agent that keeps enterprise software alive once AI-written code reaches production. Together, they show capital moving from frontier models into the agents that operate them.
If AI SRE systems work as promised, on-call engineering changes fundamentally: fewer night-time pages, faster incident resolution and less time spent reconstructing what a system already did. If they overreach, automated fixes in live systems carry their own risk, which is why the human-acceptance step and a public incident record will matter more than any benchmark score.
Background
Traversal was founded by causal machine learning researchers from MIT, Berkeley, Cornell and Columbia. Co-founders Anish Agarwal and Raj Agrawal earned doctorates at MIT, Raaz Dwivedi earned his at UC Berkeley, and Ahmed Lone came from quantitative trading at Citadel Securities. Following the raise, the company says it is hiring across research, engineering and go-to-market roles.
Source: Traversal funding announcement and SiliconANGLE report
0 Comments