Traversal Raises $120 Million at $1 Billion Valuation for AI Site Reliability Engineers

Traversal Inc. has raised $120 million at a $1 billion valuation to build AI agents that diagnose, fix and prevent IT infrastructure failures without human intervention. The round, announced Tuesday, makes the two-year-old startup one of the newest entrants in the crowded but thinly proven market for autonomous site reliability engineering.

Kleiner Perkins, Redpoint Ventures and Khosla Ventures led the financing. Sequoia Capital, Battery Ventures and Amex Ventures also participated — a roster of six blue-chip investors that Traversal will need to justify a ten-figure price tag before it has disclosed absolute revenue figures.

Why are reliability engineers under pressure?

The backdrop is a widening gap between how fast software gets written and how fast it gets kept alive. Developers have increased code output by roughly 180% over the past two years, according to a report by Splunk Inc. and Oxford Economics, and the infrastructure teams maintaining that code have not kept pace. Enterprise downtime costs have risen more than 50% over the same period, draining up to $600 billion annually from businesses worldwide.

That is the problem Traversal claims to solve. Its AI agent does not merely monitor systems and send alerts when something breaks. It investigates issues as they arise and attempts to remediate them, leaving the human engineer to accept or reject its recommendation. The promise is dramatic workload reduction; the company has not published independent verification of how often the agent resolves incidents fully on its own.

How accurate are general AI models at this job?

Not very, by the company's own benchmark. Co-founder and Chief Executive Anish Agarwal argues that even the best general-purpose large language models cannot handle SRE work. Traversal co-developed the ORCA-bench benchmark with researchers from Cornell Tech and Columbia University, and the results are stark: frontier models barely achieve 25% accuracy on medium-difficulty SRE tasks, with hallucination rates often exceeding 40%.

"Site reliability engineering is one of the hardest problems in applied AI because every company's software environment is different, and with downtime costing companies as much as $600 billion a year, the stakes are clearly enormous," Agarwal said.

The caveat is obvious: ORCA-bench is Traversal's own yardstick, co-developed with academic partners who share an interest in the field it measures.

What makes Traversal's approach different?

The company's claimed edge lies in causal machine learning, Agarwal's own area of expertise. Rather than relying on statistical correlations, Traversal's models identify cause-and-effect relationships in IT infrastructure. Deployed inside an enterprise's environment, the system analyzes logs, metrics and traces to establish a baseline of normal behavior. When problems occur, it uses causal inference to isolate the root cause and simulate the likely effects of different interventions before recommending a fix.

In plain terms: the AI is built to reason about why a system broke, not just to notice that it broke.

Who is using it, and how fast is it growing?

Traversal's traction numbers are strong, though relative rather than absolute. In the last year, the company says revenue has grown more than nine-fold and monthly active users have multiplied by 20. Named customers include PepsiCo, DigitalOcean Holdings Inc., Nubank and Harvey AI Corp. — a mix that spans consumer goods, cloud infrastructure, banking and legal AI.

"After two years refining our technology, we're seeing customers expand how they use Traversal across increasingly complex production environments," Agarwal said.

What comes next?

The $120 million will fund accelerated hiring across engineering, go-to-market and research and development teams. Agarwal's stated goal is a "superintelligent AI SRE" that lets enterprises keep pace with AI-enabled developer teams. Kleiner Perkins' Aditya Naganath, for one, is sold: Traversal is the first company he has seen bring "real intelligence" into complex production environments.

"As AI dramatically accelerates the pace at which software is written and shipped, the complexity of managing production will only grow," Naganath said. "Traversal gives engineering teams the intelligence they need to understand their systems, resolve incidents faster and operate with confidence at scales that simply weren't possible before."

The open question is whether nine-fold growth from an undisclosed base can sustain a $1 billion valuation in a market where incumbents and other AI startups are chasing the same $600 billion downtime bill.