Here is a thing that is true and slightly funny: the most important input to the artificial-intelligence revolution is concrete. You can have all the chips in the world, but a data center is a building, and a building has to be poured, wired, roofed, and - crucially - insured. Shepherd figured this out early enough to look prescient.
Shepherd started in 2020 as a fairly specific idea: commercial construction insurance is enormous, boring, and run on fax machines, so someone should point modern software at it. The founders - Justin Levine, a software person who became CEO; Mohamed "Mo" El Mahallawy, the CTO; and Steve Buonpane, a career underwriter who supplied the part of the business you cannot get from a YC batch - went through Y Combinator's Winter 2021 class and raised a $6.2 million seed round to sell excess liability coverage to mid-sized contractors.
The pitch was unglamorous and correct. Traditional construction insurance takes weeks, involves a broker, and treats every contractor roughly the same. Shepherd wired itself into the software that builders already use - Procore, Autodesk, drone-survey tools like DroneDeploy, telematics from Samsara - and used that data to underwrite in hours. If your jobsite data showed you were careful, you paid less. The company calls this Shepherd Savings, which is a nicer name than "we read your drone footage."
Then the ground shifted, helpfully
For a while this was a good, contained construction-insurtech story. Then hyperscalers started announcing data-center budgets with a lot of zeros, and the "physical layer" of AI - server farms, semiconductor fabs, the power plants and substations feeding them - became the fastest-growing construction category on earth. All of it needs insurance. Shepherd, an insurer that already spoke fluent construction data, was standing exactly where the money was about to land.
So it repositioned. By 2026 the company describes itself not as construction insurance but as "the insurance behind the AI infrastructure boom." The revenue backed the reframing: roughly 7x growth over 24 months, more than $400 billion of project value insured, over 1,500 policies for 600-plus customers. In March 2026 it raised a $42 million Series B led by Intact Private Capital, with Spark Capital and Costanoa Ventures returning, bringing total funding to about $67 million.
The ambitious part
Shepherd's stated goal is fully autonomous underwriting - a system where AI handles the entire workflow from submission to bound policy, with humans moving up the stack. That is a genuinely hard claim in a regulated, capital-heavy industry where being wrong is expensive by design. But it is also the only version of the story where an insurtech is a technology company rather than a slower bank. Shepherd is betting the data it already collects is the thing that makes the bet safe.
