I’ve been to nearly every Y Combinator Demo Day since my own in 2013, so people often ask me what each new batch feels like in terms of trends and quality. Since Rebel Fund invests in hundreds of expensive YC startups and maintains the world’s largest database of YC startups, founders, and outcomes outside of YC itself, I prefer to answer that question with data. So here we go for the S26 batch!
YC pivots to atoms
Industrials (robotics, manufacturing, hardware, energy, defense, logistics) make up a whopping 23% of the 245 companies launched thus far in this batch vs a five-year average of just 6.6%. That’s the highest industrials share of any batch we’ve tracked and roughly double the last two batches, which were already trending upwards.

The mirror image is SaaS, which fell to 12% of the batch vs a 28% five-year average. B2B software overall is down to 52%, its smallest share since S22.
In terms of themes, what I find most telling is that AI appears to be on a downward trend, with 67% of S26 companies carrying an AI tag, down from a peak of 83% in S25. Nobody tagged themselves as an “internet company” after 2005 because every company was one, and I think AI has quietly crossed the same threshold. It’s now the substrate rather than the pitch, and the more interesting question is what founders are pointing it at. In S26, the answer is often robots, factories, and defense.

YC has flirted with hard tech before. Sam Altman pushed it hard in the mid-2010s and produced a handful of real winners, but those companies were always a small minority of each batch. This wave is different in scale, and I believe in kind, because AI has collapsed the cost of the software layer that used to make hardware companies so slow and expensive to build.
Robotics founders in this batch are shipping control software in weeks that would have taken a team of ten over a year to ship pre-AI. I’ll also note, since it’s relevant to how Rebel invests, that Rebel Theorem has historically found industrials companies to be more likely to succeed than consumer companies and much of software, not less. YC funds the best founders it can find regardless of sector, so I’ve always read its industry mix as a barometer of where the strongest technical people think the opportunity is. Right now, more than ever think it’s in things you can touch.
The missing middle
If you read my post On the ‘new’ Y Combinator, you know YC founders have gotten much younger under Garry Tan’s leadership. S26 pushes that to a new extreme, but in a way we haven’t exactly seen before.
Nearly 60% of S26 founders are under 25, up from 38% in the prior batch — the highest share in five years. The average age, though, barely moved. That’s because the growth in under-25s came almost entirely out of the 25–29 band, which fell to a five-year low of 23%, while founders in their 30s and 40s held steady.

That should give my long-time readers pause, because I’ve written that Rebel Theorem has identified a founder age “sweet spot” of 26–30 years, where the odds of a breakout outcome peak. Founders under 26 have historically been much more likely to end up with dead startups. S26 hollowed out exactly the cohort our model likes best, but I’m not panicked about it for two reasons.
First, as I’ve written before, founder age is one of hundreds of features Rebel Theorem weighs, and the interactions between them matter far more than any single one. Second, these young founders are unusually well matched to what they’re building. Education fit and work experience fit, key founder-product fit dimensions we score at Rebel, rose significantly.
However, other founder-product fit dimensions have fallen. Target-geography fit, for example, which measures how well a founder knows the market they’re selling into, fell significantly, as did relevant go-to-market experience.

Put the founder fit moves together and you get the S26 founder in one line: knows the technology unusually well, knows the customer unusually little. For a batch full of first-time founders selling to factories, hospitals, and the Pentagon, that’s a risk worth considering.
Solo founders as a software factory
The finding that surprised me the most: 19% of S26 companies have a single founder vs a five-year average of 8.1%. That’s the highest solo share in five years, and the third elevated batch in a row, despite YC’s historical aversion to solo founders.

I assumed this was the hard-tech wave again, i.e. a lone roboticist with a thesis project, but it’s the opposite. Solo founders in S26 are disproportionately building software (64% of them vs 52% for the batch overall). The solos are young as well, with 55% under 25, the same share as founders on two-person teams.
In my W26 post I raised the question of whether Claude Code, Codex, and the like have changed the fundamentals of building a startup. I think this is the first hard evidence in our data that the answer is yes, at least for software. A 23-year-old first-time founder can now get a software company far enough alone to be worth YC’s money. Whether a solo founder can carry a hardware company is a separate question.
I’ve watched a lot of YC batches, and I can’t remember one this internally consistent: heavy on physical products, often built by very young technical founders out of university labs, many of which have never founded a company or raised a round. Whether they’ve picked markets that reward their technical expertise, or are about to learn how hard atoms and legacy industries really are, is the billion dollar question.

