The most revealing object in venture capital may be the spreadsheet nobody wants to open. In a factory, it tracks machine settings copied by hand. At a port, it records a shipment already late. Inside a procurement department, it sits between a buyer, a supplier and an email chain with 47 replies. These are not glamorous scenes. They are exactly where Construct Capital goes looking.
The Washington, D.C., venture firm was founded in 2020 by Dayna Grayson, an investor who had spent years backing industrial technology at NEA, and Rachel Holt, an operator who helped scale Uber's sprawling U.S. and Canadian rides business. Their premise was plain: the industries that make, move and power the physical world had missed much of the software revolution. That omission was not a sleepy backwater. It was a market.
Construct calls these markets “foundational industries.” The phrase covers manufacturing, logistics, mobility, defense, energy, food production and critical infrastructure. It is broad enough to include a robotic arm portioning meals and a generator resting on the ocean floor. What connects them is less a technology than a condition: enormous systems still run on old equipment, fragmented data and labor-intensive processes, even as customers expect digital speed.
A thesis forged in two different careers
Grayson and Holt did not arrive by the same route. Grayson had developed an industrial investing record that included Desktop Metal, cloud design platform Onshape and custom framing company Framebridge. Holt came from the daily friction of a physical network at software scale: drivers, riders, cities, vehicles, regulations and the countless exceptions that appear when an app meets a street. One knew how technical companies were financed. The other knew how quickly an elegant model became an operational puzzle.
That pairing helps explain Construct's range. The firm is interested in software, but not software in isolation. It looks for the connective tissue between code and consequence: industrial version control from Copia; precision aerospace production at Hadrian; parcel networks at Veho; wearable robotics from Verve; food automation at Chef; automotive service at Kinetic. The buyer is usually a business with real assets, real labor and a measurable penalty for delay.
“We don't shy away from hard work and always aim to be a founder's first call.”Construct Capital, on its operating posture
This also describes the product Construct sells. To founders, it offers capital and company-building help. To limited partners, it offers exposure to a focused portfolio of private technology companies. Construct typically leads or co-leads Seed and Series A rounds. It says it brings the whole team's investing and operating experience to a limited number of companies, rather than collecting hundreds of logos and rationing attention afterward.
Why concentration is part of the product
The firm's fundraising charts its ambition without abandoning that constraint. Construct announced a $140 million debut fund in 2021. In 2022 it added $300 million: a $225 million second early-stage fund and a $75 million vehicle for selected later-stage opportunities. Fund I backed 15 teams by that announcement, and the first two early-stage portfolios were designed around roughly 20 companies apiece.
Then came Fund III. In March 2025, Construct closed $300 million and said it had deliberately stopped at the hard cap. Bigger funds can support bigger checks and fees, but they also pressure managers to deploy more money or drift later. Construct's chosen size lets it make meaningful early investments while keeping enough attention for board work, hiring, financing and the untidy decisions between milestones. Its investors include endowments, foundations and family offices; the exact economics of the funds remain private.
The business model is conventional venture capital: limited partners commit money, Construct invests it for equity, and the manager earns fees plus a share of gains if the portfolio performs. The distinction is in selection and service. Industrial companies can need more capital, longer development cycles and deeper customer integration than lightweight software. A small portfolio does not remove those risks. It makes ignoring them harder.
AI leaves the browser
Construct's most useful framework reduces industrial AI to three verbs: sense, think, act. First, collect and normalize information from machines, sensors and legacy software. Next, reason over it. Finally, execute a decision through an agent, a machine or a robot. Consumer AI made the “think” layer visible. In foundational industries, the last layer is where theory encounters heat, weight, tolerances and safety rules.
That makes physical AI different from a clever assistant. A model can draft a plausible memo and survive an awkward phrase. A robot handling food or a machine cutting a flight-critical component cannot improvise past a bad result. Industrial customers require reliability, integration and a credible return on investment before deploying widely. Sales can be slow. Pilots can strand startups in custom work. Hardware can consume cash before the economics are proven.
Construct argues that these difficulties can become moats. A company that earns trust on the factory floor accumulates workflow knowledge, proprietary operating data and integration depth that a remote software competitor cannot easily copy. The strongest businesses may combine hardware, software and operations rather than selling a thin tool. Hadrian, for example, uses automation and proprietary software to manufacture precision parts, while Copia brings Git-like workflows to the programmable controllers inside industrial systems.
A map of bottlenecks, not a basket of sectors
A portfolio that runs from logistics software to energy in orbit can look opportunistic. Construct's connective logic is bottlenecks. Podium Automation, whose $18 million Series A the firm led in June 2026, attacks the work before an industrial control panel is assembled: quoting, component selection, compliance logic, layouts and instructions. Endurance Energy is developing modular subsea geothermal generators for places where conventional grids are strained or unavailable. Cowboy Space Corporation is pursuing solar-powered compute infrastructure in orbit.
The customers differ, but each company moves expensive work away from scarce labor, brittle coordination or constrained infrastructure. Nauta helps importers manage global trade beyond email and spreadsheets. Sphere Semi applies software automation to analog chip design. Shade, a newer investment, addresses the file sprawl created as AI makes media production abundant. The portfolio stretches, but the recurring question stays intact: what old system becomes newly programmable?
Construct competes for those deals with specialists such as Eclipse, DCVC, Lux, Congruent and Ironspring, as well as large firms whose defense, climate and “American Dynamism” practices have made the physical economy fashionable. Its difference is not exclusive access to industrial startups. It is the combination of early-stage focus, cross-industry pattern recognition and an operator-investor founding team. The firm's Washington base also places it near policy, defense and infrastructure conversations, although its portfolio is geographically broader.
Timing has helped. Construct launched just as the pandemic exposed fragile supply chains and labor shortages. Geopolitical tension, reindustrialization policy and surging energy demand from AI made the same thesis more urgent. By 2026, “physical AI” had become a crowded phrase. Construct can fairly say it was pursuing the underlying problems before the label became useful marketing.
The test is what gets built
Being early to a theme is not the same as producing returns. Venture portfolios are judged by exits and distributions, and many of Construct's bets remain private and young. Industrial startups face unforgiving execution: factories must open, machines must work, procurement cycles must close and gross margins must improve. The firm's valuation and revenue are not public, nor are the performance figures for its funds.
Still, the strategy offers founders a practical lens. Find a large physical workflow where demand is rising but capacity cannot respond. Look for data trapped in old systems, specialists nearing retirement, or handoffs coordinated by phone and spreadsheet. Build a wedge that solves an immediate operational pain. Then use the resulting data and trust to widen the product. It is less romantic than inventing a new social habit. It may be more difficult to dislodge once it works.
Construct Capital's wager is ultimately about where technology becomes visible. Not in a demo, but in a shorter parts queue, a safer warehouse shift, a parcel that arrives on time or a turbine producing power where the grid cannot. The next technology cycle may still begin with code. This firm expects its value to be measured in motion.