Exclusive: The Startup Using AI To Build Better Data Centers
Build has raised $8.5M from Index Ventures and others, including OpenAI’s CFO, to speed up big projects for customers like the U.K. government.
In September 2024, James Stirrat-Ellis was hanging around a drop zone outside Seville in Spain when he struck up a conversation with another skydiver in between jumps.
His new friend was a senior executive at a data center provider, and when the two met up back in London, the exec urged Stirrat-Ellis – an architect by training, and recently a manager at 11x, the controversial AI agents startup – to apply his skill set to challenges in the AI-fueled data center boom.
“He said, ‘I know that you can automate this stuff, there’s a ton here,’” remembers Stirrat-Ellis. But it wasn’t love at first sight for Stirrat-Ellis and his co-founder, Ben McClusky. “To be frank, we weren’t that excited.”
Building software for the pre-construction phase of such projects – seemingly a small piece of the puzzle – didn’t seem like a high-octane startup niche. But as the founders spoke to more potential customers, they realized that the services involved ran into the millions of dollars – and involved much more of a challenge than initially met the eye.
If they could automate some chunk of that using AI agents, they could build a highly profitable agency. More, and they might have a venture capital-scale winner. And if they could make large-scale building projects mostly autonomous – taking tech-style margins, not traditional services ones – they could build a generational company.
So the duo founded Build.inc, a startup that runs dozens of AI agents to automate everything from the PDFs and Excel documentation of a project to site recommendations and permitting process.
Build already works with the U.K. government advising its AI growth zones initiative, among more than 100 major infrastructure projects, with other customers in real estate and data center development, the startup says.
Build claims it can compress months-long pre-build processes into several weeks, at half the cost. Contracts start in the six figures; and Upstarts estimates Build to have bookings into the millions in future sales; the startup declined to confirm that, but said that it had “tripled its booked revenue since April.”
The bigger ambition is to handle more of a large-scale building process across other categories, too, from energy and industrials to residential.
“In five or 10 years, we’re going to make building autonomous,” claims Stirrat-Ellis. “You can email Build, and your building goes from A to B, from concept to completion, as autonomously as possible.”
Just nine people today in New York, San Francisco and London, Build is now publicly launching with $8.5 million in seed funding led by Index Ventures, with Pebblebed, Puzzle Ventures, Tiny.vc and a number of personal investors including OpenAI CFO Sarah Friar and Blackstone CTO John Stecher.
With the funding, Build – which has relied almost entirely on word-of-mouth so far – will need to grow its go-to market and sales, says new lead investor Martin Mignot at Index.
Build is very early in its journey. But the startup’s launch is already interesting to the wider ecosystem for two reasons that we’ll explore below:
The rise of the ‘neo firm’
The data center wedge
“We’re going through the world’s greatest infrastructure build-out ever, full stop,” says Stirrat-Ellis. “These data centers will happen regardless. Build exists to free up the headroom to actually say, ‘Well, how do we build this the right way?’”
🤝 Interested in partnering with Upstarts to sponsor exclusive startup launch stories like this? Drop us a line or fill out our interest form 🤝
‘Neo’ pilled
Build is the latest in a wave of startups to pursue some combination of AI-native services and agents. When Stirrat-Ellis and McClusky started building this craze hadn’t started yet, and Build’s CEO says it was “very, very unpopular.”
But the duo figured that if they could generate a massive technical due diligence document for a data center site that would typically take six weeks in just six hours, they could charge $25,000 for it and enjoy margins more like a tech company than a consulting shop.
They spun up dozens of agents to handle different tasks; in the initial workflow, one of the founders would email those agents to get PDFs back, and they’d grind long hours to turnaround a finished document for a client. “It looked kind of pathetic, but behind the scenes, a lot of stuff was going on,” Stirrat-Ellis says.
Since then, they’ve hired experts from giants like Starwood, Tishman Speyer and data center businesses to embed their domain expertise alongside the agents. It’s an approach similar to how Crosby, another New York startup we’ve covered in Upstarts, looks to deliver faster and cheaper legal work around contracts.
Such businesses don’t sell software subscriptions, crucially but focus on customer outcomes, which they believe can better align with customer needs while still creating a healthy margin.
Outcomes over software was an idea Stirrat-Ellis says he first developed while still at 11x, although he ultimately cut his time short at that company amidst its turmoil and accusations around how it represented revenue and customers. (Of that experience, Stirrat-Ellis says: “One of the biggest lessons is to do things the right way.” Adds Mignot, his backer: “You learn a lot about someone from what they take away from these types of experiences, and he was very transparent and open.”)
So if Build isn’t a software company, and it’s not just a services firm, what is it, exactly? Does it call itself AI-native services, or services-as-software, like in our recent coverage of Gainsight? Build’s CEO tosses out other terms: neo firm, or AI operating partner, but there isn’t an ideal phrase yet.
Its revenue model is different, too: retainer pricing that starts in the six figures and can run into the millions as customers hit usage caps. Build’s agents are expensive to run – as much as 100x the cost of other agents in simpler use cases, per its CEO – but that’s baked into its healthy margins for now.
“We’re figuring out pricing, like every AI native company today,” Stirrat-Ellis says.
Data center crunch
Build is currently playing into the urgency to build large-scale data center projects globally – and, somewhat counterintuitively, the blockers and backlash.
For customers like the U.K., Build helps to map out the local power grid and site the centers in optimal locations, factoring in other variables including political community sentiment.
Stirrat-Ellis says he and his team understand why locals near some of these proposed developments object – “living beside a data center with massive amounts of noise is not a good place to be,” he agrees – but he argues that Build can help bring those concerns and discussions earlier into a process, helping all parties to reach a happier outcome.
“Without Build, these data centers are going to be really cumbersome, and annoying, and difficult to live with,” he claims.
Speeding up the time process to get permits and reach such compromises will meaningfully lower the cost to build such projects, argues Index’s Mignot, increasing the pace of AI development, which can then further drive down costs.
There’s a sovereign AI angle, too: “Every country is trying to compete with their neighbors for compute capacity throughout the world,” Mignot says.
But for Build to grow into the ‘generational’ company that its CEO aspires to lead, data centers will eventually need to be a smaller part of its business. Similar tech could help build nuclear giga-factories like the one that Valar Atomics aspires to introduce, or for large-scale new housing projects.
For Stirrat-Ellis, who didn’t grow up dreaming of launching such a startup, and who says he was taken more by the market than the idea initially, there’s a lesson:
“I think it’s small markets, that appear really dull, that actually end up building really, really great companies,” he says. “Because you kind of have to break some rules to make something happen in an excellent way.”
It’s hopefully a turn-of-phrase – or at least all clearly ethical rule-breaking – this time around.