Not long ago, generalist investors needed “construction tech” explained to them. Now contech has conferences, dedicated funds, and tier-one VCs who no longer need the pitch. At Leonard’s Demo Day, Darren Bechtel, founder and managing director of Brick & Mortar Ventures pointed to three shifts that brought it there. Multi-billion-dollar valuations like now disappeared Katerra’s first drew attention. Then, the Covid pandemic recast construction as essential infrastructure rather than a discretionary industry. And now as AI has begun to look like a threat to knowledge work, capital has been hunting for sectors it couldn’t easily automate away. Construction, still turning raw materials into physical assets in the real world, fits that description.
That interest has not moved in a straight line: contech funding ran hot into 2021, then corrected sharply. Through the first three quarters of 2025, though, the sector drew more venture capital than in any of the three preceding full years, with robotics and AI leading the way, according to Nymbl Ventures. As AI unsettles other sectors, a physical, hard-to-automate industry looks like a safer bet. For Bechtel, even now, this is “early innings.”
But drawing capital is only the first of AI’s effects on construction. It is also reshaping the industry directly: driving demand, reviving automation, and, with time, sharpening how it manages risk.
Data center demand is industrializing construction
US data center construction hit a record $41 billion in 2025, and for the first time the country is spending more to build data centers than offices, according to Census Bureau data. Bechtel puts it another way: in five years, he says, the industry will build what the United States took eighty years of railroads to lay down. And the build-out is increasingly global, expanding across Europe, Asia, and the Gulf.
That pace is driving many builders to industrialize. While industrialization of construction is often discussed, the urgent and often cost-insensitive demand for data centers is demanding it. The skilled workforce can’t scale to match it, and staffing up permanently for a boom that will eventually cool makes little sense. So builders are standardizing designs, working in repeatable modules, and pushing as much construction as possible off-site, turning out a data center much like the last one more like a product than a bespoke project.
The centers themselves are only the first wave of the demand surge. The power they draw is already straining grids, and the IEA expects data centers to account for close to half of US electricity-demand growth through 2030. This drives the need for additional infrastructure to facilitate increased power generation (including new technologies such as small modular reactors) as well as water delivery for cooling.
Construction robotics is finally attracting real investment
Building at this scale runs into a familiar limit: people. The labor shortage in the trades is well known, and with the workforce unable to grow to meet demand, automation becomes the way to add capacity and complete more work without more hands.
That is leading construction robotics to draw new capital and attention, as labor scarcity meets the “physical AI” narrative now sweeping the sector. While the label is largely a rebrand for a renewed enthusiasm for robotics merged with the hype of the AI trend, there is something real beneath it. AI, computer vision and better sensors are beginning to let machines adapt to changing site conditions in real time, the very thing that always made construction harder to automate than factory production. US-based Bedrock Robotics, founded by former Waymo engineers, retrofits excavators and bulldozers to run autonomously, and its bet on construction autonomy has drawn major backing.
Bechtel’s own bet is on the narrow over the general. “I’m not a believer, at least anytime soon, in general-purpose robotics in construction,” he says. The provable applications are specific: floor layout, drilling patterns, drywall finishing. That is the kind of bounded job companies like US firm Dusty Robotics does when it prints a building’s plan straight onto the slab. More fundamentally, the shift is toward off-site, prefabricated work, where factory robotics operate in the controlled conditions manufacturing already handles well. On the unpredictable job site, the more realistic model is having a human-in-the-loop with robotic assistance.
Low-carbon materials get a second act
If demand and labor are reshaping how the industry builds, decarbonization is changing what it builds with. Low-carbon materials failed to break through a decade ago, weighed down by a “green premium” that asked buyers to pay more for the same performance. A newer approach is proving more successful: drop-in substitutes that match cost and performance with no change in behavior. Bechtel likens this to swapping gluten-free flour into an unchanged recipe. Examples include American startup Sublime Systems, which makes a cement that meets the ordinary Portland standard without a fossil-fired kiln, and Dutch startup Paebbl turns captured CO2 into a mineral that stands in for part of the cement in concrete.
However, scaling a materials business is difficult. Physical goods must cross global supply chains and clear certification market by market. Unknown brands struggle to sell unproven materials, which keeps incumbents ahead and pushes most materials startups toward acquisition as the route to scale. The constraint is commercial rather than technical. As across much of construction’s decarbonization, the missing piece is a viable business model, not the technology.
Risk, not hype, sets the pace
All three shifts run into the same constraint. Construction operates on thin margins, and a single failed project can sink a firm. That calculus shapes how the sector is meeting this moment. It is why contractors industrialize rather than overstaff for a boom that will pass, and why incumbents tend to buy innovation rather than build it, waiting until a startup has proven a technology before acquiring it. Autodesk built its construction cloud out of PlanGrid and BuildingConnected; Liechtenstein’s Hilti bought the field-app maker Fieldwire; and AECOM paid $390 million for the Norwegian AI startup Consigli. Each let someone else take the early risk, while then leveraging their market penetration to distribute the new technologies at scale.
The AI technology boom driving these changes may eventually be turned on the industry’s risk model itself. Beyond adding capacity through robotics or reworking materials, AI is beginning to make the job site legible in real time: the Israeli firm Buildots tracks progress through computer vision, and the British firm nPlan predicts delays from a vast base of historical project data, both early attempts to catch problems while they are still cheap to fix.
Bechtel pointed to improved real-time visibility into the jobsite and supply chain as the largest opportunity for competitive advantage in the near future. It’s a promising direction. Most job sites still run on paperwork and memory, and in an industry with so little room to be wrong, this kind of visibility might arrive the way change often does: proven one project at a time.
