Nobody is founding neobanks anymore. Something else is being founded instead.

September 24, 2026 · 6 min read · ← all posts · the dataset

Every one of the 381 neobanks we track carries a founding year. Line them up and the shape is not the one the industry talks about.

Peak was 2019, at 43 new neobanks in a single year. 2025 produced 11. Take three-year windows to smooth the noise and the drop is cleaner still: 103 founded across 2019–2021, 57 across 2023–2025. Roughly half, in four years.

The half that actually matters

A decline in founding is interesting. What replaced it is more interesting, and it doesn't show up in a count at all — only in the mix.

Split the dataset into two five-year cohorts and ask what kind of company got founded in each:

In the first cohort, four out of five new neobanks were what everybody pictures: an app, a card, a partner bank or a licence, deposits denominated in a national currency. That model was 81% of new entrants.

In the second, it is 40% — and self-custodial, on-chain products went from a rounding error to the single largest group of new entrants at 43%.

So the sentence "fewer neobanks are being founded" is true and misleading at the same time. Fewer of that kind are being founded. The count of traditional neobanks started in a five-year window fell from 139 to 35 — a 75% collapse — while web3-native founding more than tripled.

Why the traditional model stopped attracting founders

We can't read minds, but the dataset makes three constraints visible, and they all point the same way.

The licence is a wall, and everyone can now see it. Of the 381 neobanks here, 132 hold their own banking licence and the rest run on somebody else's rails or on a narrower permission. When Nubank entered the US this month — 139 million customers, three national banking licences, a billion-dollar quarter — it still went live on a partner bank's charter. If that company rents, the barrier is not a funding problem a seed round solves.

The partner-bank layer turned out to be a dependency, not a shortcut. Three of the six neobanks in our graveyard died when the institution underneath them withdrew. The model that made launching cheap in 2018 is the model that made surviving conditional in 2024.

The general-purpose seat is taken. 253 of 381 tracked neobanks describe themselves as serving a general audience. Founding the 254th general-purpose challenger in a market with Revolut, Nubank, Chime and WeBank in it is a different proposition from founding the 30th. What's left is niches — 33 for SMBs, 23 for the underbanked, 15 for freelancers, 10 for immigrants — and niches support fewer companies each.

Meanwhile the web3-native route requires no charter, no partner bank and no deposit insurance, because it takes no custody of anything. Whatever else is true of that trade, it removes every barrier in the paragraphs above. The founding numbers suggest people noticed.

What to watch instead of the count

If the founding rate keeps falling, the number that matters next is the failure rate, because a market with few entrants and any exits shrinks. Our graveyard has six entries; that is a small enough number that each one is still a story rather than a statistic, and we write them up individually for exactly that reason. The day it becomes a statistic is the day this stops being a maturity story.

The second number is whether the 43% keeps holding. Web3-native founding surged through a period of unusually favourable conditions for on-chain products. If that share falls back without traditional founding recovering, the industry isn't rotating — it's just contracting, and 2025's eleven will read very differently in hindsight.

Both are answerable from an open dataset, which is why this one exists. Every figure above is reproducible from data.json with about ten lines of code, and the founding year of every entity is a field you can check against the company's own filings.