We are living through a generational change. The developed world is running short of workers, these industries are being asked to build more, and AI and robotics can finally do the work. We back the companies at the center of that collision, at Seed and Series A.
For decades, technology skipped the industries that build, move, fix and power everything else. Not because they were small, but because they were hard to serve.
That is ending. Defense, energy and domestic production are being rebuilt while the workforce shrinks. At the same moment, AI made software cheap and flexible enough for a small team to serve a contractor, a hotel or a utility.
Any one of these would be a trend. Together they will produce the next generation of large companies.
Messy workflows, legacy systems, long sales cycles and buyers who never asked for software kept startups away from the industries everything else depends on.
Birth rates across the developed world are below replacement. The dirty, dangerous and tedious jobs go unfilled first: hotels cannot staff the front desk, repair shops have no one to do the paperwork. The customer will still pay to get the work done.
Building for these industries used to take large teams and years of customization. Not anymore.
Software and machines take on the jobs nobody can hire for.
A handful of people can ship what used to take hundreds.
Running AI costs about 280 times less than it did in late 2022.
Customers buy the work done, not another tool.
Their biggest costs are people, power and risk. Software was always a rounding error in their budgets.
The money is in the work, not the software. That is the budget our companies sell into.
Every company we back replaces a cost far larger than software.
The hire that never comes.
The utility bill and the grid connection.
The bandwidth, compute and networks.
The breach, the fraud, the missed booking.
European NATO defense spending since 2014. Energy and domestic manufacturing are being rebuilt too.
New construction workers the US needs each year. The dirty, dangerous and tedious jobs go unfilled first.
Cost per million tokens to run a GPT-3.5-level model. Software that does the job now costs less than the hire who is never coming.
Models are getting cheap and alike. Knowing why a hotel loses a booking at 2am is not.
Per booking, per inspection, per hull cleaned. Paid from the payroll budget, not the IT budget.
Every model runs on electricity, and demand is outrunning supply. Whoever adds supply or cuts demand wins.
No humanoids. Machines that do one dirty, dangerous or tedious task, like cleaning a ship hull.
As AI writes more content and code, proving what is real and safe becomes its own layer.
Track record, portfolio construction and fund terms. Accredited investors only.
Request it →