The half that isn’t AI

In 2025, roughly half of every venture dollar in the world went to artificial intelligence. Crunchbase puts the figure at $211 billion — up 85% from $114 billion the year before — out of about $425 billion invested across more than 24,000 private companies. It is the largest share any single category has ever taken from the rest, and it arrived fast enough that the number reads less like a trend than a takeover.

For the founder who isn’t building AI, though, the interesting number isn’t the one about AI. It is the one underneath it. In North America, where the concentration is sharpest, total funding rose 46% in 2025 — and the number of deals fell about 16%, to just under 10,500 rounds. More money, changing hands fewer times. The pool got deeper, and the number of people allowed to swim in it got smaller.

It is worth saying what “AI” means in these counts, because the label is generous. Crunchbase and PitchBook both file a company under AI largely if it presents itself that way, which sweeps in everything from foundation-model labs raising tens of billions to ordinary software firms that added a chatbot and a line to the deck. The half is real; some of what fills it is marketing. Which is exactly the pressure a non-AI founder now feels — the incentive to relabel is doing real work.

The bar split in two

A year earlier, AI took 34% of global funding. In 2025 it took close to half. A category doesn’t gain that much share in twelve months by growing — it does it by pulling money out of everything else. What used to be a single bar to clear — is this a good company, at a fair price — has quietly become two, and which one you’re measured against depends on a single word in your pitch.

That is close to the whole market in eleven words. The venture business has split along two axes at once — the AI companies and everyone else, the very large funds and everyone else — and a founder can land on the wrong side of both without changing a thing about the company they run.

The snapshot bears him out. By early October 2025, PitchBook counted $192.7 billion flowing into AI startups out of $366.8 billion in total U.S. venture investment for the year. In the most recent quarter it measured, AI took 62.7 cents of every American venture dollar. The rest of the economy — payments, logistics, health, climate, the ordinary software that runs ordinary companies — divided what was left.

What the other half is facing

None of this means capital has left. It means the terms of attention have changed. A non-AI founder in 2025 wasn’t raising into a closed market so much as a distracted one, where the default question in the room stopped being “how fast does this grow” and became “where is the AI in this.” The consequences are less about rejection than about drift:

• The comparison set moved. A solid software company at a fair multiple now sits in the same meeting as a model company promising a new category, and gets read against it.
• Checks got bigger and rarer. When dollars rise while deal count falls, the money is concentrating into fewer, larger bets — a harder table to get a seat at, not an easier one.
• The story stopped being enough. Narrative and market size now compete with companies that have better narratives; what travels instead is traction, margin, and a credible path to not needing the next round.

Some of that is a healthier bar than the one it replaced. For a decade the reward went to the best story about growth. For a founder outside AI in 2026, it increasingly goes to the plainest evidence of a business — which is harder to fake, slower to build, and quietly favors the founder who was going to run a disciplined company anyway.

The half of the money that went to AI will keep the headlines, and mostly earns them. But the more useful fact for everyone else is the quieter one: the other half didn’t disappear. It got choosier, and started asking for proof instead of a plot. Founders who have the proof will find it a stranger market than the one they trained for — and, in the end, not a worse one to be building in.