In brief
There's a statistic in Euclid Ventures' 2026 vertical report that says more about the next decade than most market forecasts.
There's a statistic in Euclid Ventures' 2026 vertical report that says more about the next decade than most market forecasts.
In Manufacturing & Industrial, 57% of founders started their companies after 2022.
Not existing companies pivoting into AI. Net-new formation, concentrated in the last three years, in an industry that had gone largely unbothered by the previous software cycle.
Something changed about who is able to build.
What actually got unlocked
The obvious reading is "AI made software cheaper to build." True but incomplete. Software has been getting cheaper to build for twenty years, and it didn't produce this.
What changed is more specific: the boring middle of industry workflows stopped requiring a team.
Every vertical product needs the same unglamorous machinery. Parse the document that arrives as a photo of a fax. Classify the incoming request. Extract fields from a supplier's spreadsheet that follows nobody's schema. Turn a voice note into structured text. Route the exception to a human.
That work used to be most of the engineering budget, and it's exactly the work that made vertical software uneconomic for small markets. You'd spend two years on plumbing before touching the thing that made your product different.
Now the plumbing is a few API calls, and the expensive part is knowing which workflow to build — which is domain knowledge, not engineering.
That reversal is the whole story. It shifts advantage from people who can build to people who know what to build. And in industrial verticals, those people have mostly been running businesses, not writing code.
The founder profile this favours
Startup pattern-matching has spent two decades optimising for a specific person: technical, from a scaled tech company, young enough to work for equity.
The vertical wave favours someone else entirely. Someone who has run a business in an unglamorous industry for years, accumulated a very specific understanding of where the money leaks, and — critically — has never thought of themselves as the kind of person who starts a software company.
The clearest recent example isn't in manufacturing. In 2015 Luke Hansen was working in his family's roofing business in Nebraska. He looked for software to manage jobs, couldn't find any that worked, and built it himself. CompanyCam was valued at $2 billion in November 2025 — the first unicorn in the state's history — on a Series C that PitchBook put at $415 million.
The product photographs job sites and makes the photo the record. That's it. Ten years and two billion dollars for a camera, a timestamp, and the discipline to treat that as the source of truth instead of somebody's recollection.
Hansen wasn't a technologist who found a market. He was a roofer with a problem he couldn't buy his way out of.
Why domain knowledge is the moat now
Euclid's read on what investors say they're buying: category leadership and founder-market fit. General Catalyst led 24 vertical deals in 2025; a16z led 22.
"Founder-market fit" is a phrase that means little until you see which side of a trade-off it decides. In these categories it means: having worked in the industry beats having worked at a scaled tech company.
That's not sentimentality. It's a read on what's actually scarce.
Features get copied in a quarter. Model access is a commodity, and the provider can absorb your product any time. What can't be copied on a quarterly cycle is knowing that a mechanic won't walk to a terminal, that a roofer's dispute is always about what the homeowner claims wasn't done, that the real bottleneck in a shop is the three hours a car sits waiting for a customer to answer his phone.
You don't get that from user interviews. You get it from years of losing money to it.
The report describes the trades opportunity as "fragmented end markets, low technology adoption, high workflow complexity." Every one of those is hostile to an outsider and neutral-to-favourable to someone already inside.
The part operators get wrong
Having been on that side, I'd add the counterweight, because the story is usually told too cleanly.
Operators have the knowledge and almost always underestimate everything else.
The software is the easy half. I own a repair shop network in Ukraine and built MECH Orbit inside it — the car enters the bay, a camera reads the plate, the work order opens itself; the mechanic photographs and dictates from his phone; the customer approves specific line items with the time and scope stored. Getting that to work took less effort than getting my own team to change how they work. Adoption is measured in months, and I had the advantage of owning the shops and being able to insist. A vendor selling from outside doesn't have that lever, which is a large part of why so much of this software gets bought and abandoned.
Knowing the industry doesn't mean knowing the market. I knew exactly what shops need. I knew nothing about how to reach thousands of them for less than they'd ever pay me, and I spent real money learning that paid acquisition isn't a distribution strategy, it's a tap you rent.
Domain knowledge has a shelf life. It's most valuable in year one and quietly rots after that. The longer you build, the further you drift from the person you were when you had the problem — unless you keep running the business, which most operator-founders stop doing.
What this means for the next few years
If the tools that made the boring middle cheap keep improving, the constraint on vertical software keeps shifting away from engineering and toward access.
Which suggests the interesting founders for the rest of this decade won't be found in the usual places. They're running eight-bay repair shops, twelve-person fabrication businesses, regional logistics yards, small clinics. They have the knowledge, they've been ignored by two software cycles, and the barrier that kept them out just fell.
57% of industrial founders starting after 2022 isn't a blip in the data. It's what it looks like when a group of people who always knew what to build finally can.
Sources
- Euclid Ventures — The Vertical Report 2026
- Silicon Prairie News — CompanyCam reaches $2 billion valuation
What to check in your workshop
- Verify the primary source and date behind every figure
- Separate reported facts from the author's interpretation
- Compare the conclusion with data from your own workshop