A three-person startup that emerged from Y Combinator's 2026 summer batch is betting it can underwrite the infrastructure powering the AI boom. PRINCEPS launched coverage for GPU residual value, service-level agreement breaches, and compute outages, entering what it describes as a trillion-dollar buildout at a moment when cloud spending shows no signs of slowing.
The London-founded firm, which appears to now operate primarily from San Francisco, positions itself as the first purpose-built insurance company and managing general agent dedicated to compute infrastructure. Whether that claim withstands scrutiny depends partly on how one defines the category, which has attracted a handful of new entrants in recent months.
Underwriting What Doesn't Yet Exist
PRINCEPS underwrites four types of compute risk, according to its public-facing materials. SLA coverage protects cloud providers when they fail to deliver compute as promised. GPU residual value insurance sets a floor price for hardware at some future date. Outage coverage addresses availability failures. The company also mentions model-output liability on job postings, though that remains largely theoretical across the industry.
The firm says it builds software to price risks with little or no loss history, then places them with rated carriers. "We underwrite risk for the compute economy from San Francisco," the company wrote on its careers page. The approach relies on what PRINCEPS calls a "pricing kernel" and exposure data pipelines, language suggesting it pulls telemetry from live infrastructure to model potential losses.
That's no small task. Insurance has historically struggled with novel risks, particularly in technology where historical data runs thin and the pace of change outstrips actuarial models built for slower-moving asset classes.
A Market Moving Fast
Data centers have become one of the fastest-growing segments in commercial real estate and infrastructure investment. Amwins projected $7 trillion in global capital expenditure on data centers by 2030 in a market insight published mid-year. Cloud infrastructure spending alone jumped 35 percent year-over-year in the first quarter, according to Synergy Research data cited by AIG.
The question isn't whether compute infrastructure represents economic value. It does, and increasingly so. The harder question involves how to price uncertainty when autonomous systems fail or when yesterday's cutting-edge hardware becomes tomorrow's stranded asset.
Insurance Journal examined the underwriting challenges in an August viewpoint, noting that compute utilization rates and residual hardware values have become central variables for carriers attempting to enter the space. "How much economic value have organizations entrusted to autonomous intelligence, and what happens when that agency fails?" the publication asked.
Who's Building It
Conrad Frøyland Moe, founder and CEO, and Anya Trofimova, founder and COO, both have ties to Oxford Union, according to the startup's Y Combinator profile. Trofimova previously worked on project finance and software for the UK Parliament. The company lists London as its location in Y Combinator's directory but describes operating from San Francisco in job postings, a setup that suggests either a recent relocation or a distributed team.
PRINCEPS was recruiting for four founding roles as of late summer: underwriter-actuary, engineer, and two go-to-market positions.
A Crowded Field
Several firms moved into compute insurance in recent months. American Compute plans to launch a managing general agent with a $5 million delegated limit for GPU residual value coverage, The Insurer reported in late August. ForwardCompute offers parametric outage coverage, SLA liability, and residual value products, recently partnering with Canopy Cloud and Point64, a Bermuda-regulated intermediary.
AIG embedded cloud outage protection from Parametrix into its cyber policies. ATA Insurance launched a $750 million facility in January backed by Arch, Munich Re Specialty, and SCOR. Native offers AI compute insurance through a London market lineslip with capacity reaching $600 million.
The competitive landscape matters because these firms are essentially building insurance products in real time while the underlying risks evolve. GPU prices fluctuate. Service-level agreements change. What counts as an outage depends on the contract.
Infrastructure Hooks
PRINCEPS says it is working with Ornn, a compute marketplace that raised $33 million in June and publishes pricing indices for GPU capacity. "We are working with Ornn — one of the largest marketplaces for compute — to bring greater transparency into the reliability of neocloud providers trading on their platform," the company wrote. Silicon Data launched monthly GPU residual value benchmarks in August, adding another pricing reference.
Access to real-time market data could prove valuable, perhaps more than the founders initially expected. Pricing insurance for assets whose value can swing based on the latest model release or chip architecture requires something closer to derivatives modeling than traditional property coverage.
PRINCEPS presented at Y Combinator's Demo Day in September, part of the accelerator's summer batch. Whether the firm can scale beyond its three-person team and turn underwriting software into a durable business remains an open question. But the market it's chasing keeps getting larger.
