Author: Francesco Andreoli, Head of Developer Relations at Consensys & MetaMask
Compiled by: Jiahuan, ChainCatcher
Everyone is counting funding rounds, no one is counting funerals. So I counted both sides.
Six months ago, I began counting neobanks because I found that no one could tell me how many companies were in this industry. Analysts charging $4,000 for PDFs didn’t know, VCs investing in the space didn’t know, and founders competing against each other didn’t know either.
The answer is, as of July 2026: 368 verified, actively operating neobanks. I track each one on neobankbeat.com, and all data is open.
But the number that truly changed my perspective on this industry wasn't 368, but how many I had to remove to arrive at that number.
368 newly verified banks still in operation
The rise is real. Let’s start with this point.
By summing up the user numbers disclosed by all companies in the dataset, the new banks we track serve approximately 1.46 billion people. This is not a projection or a number from a TAM chart—it is the actual reported customer count from each company.
The geographic distribution will surprise anyone accustomed to Western fintech media:
Of these, 817 million users are in Asia. Just WeBank serves over 400 million people, more than all the neobanks in the United States and Europe combined.
Nubank has 131 million customers, more than the total number of neobanks in the United States.
Revolut, Europe’s flagship, has over 50 million users. This is an impressive achievement, but still a fraction compared to the numbers in Asia.
The marginal momentum of the industry is also shifting. Of the new banks founded and still operating in the 2020s, 30% are web3-native self-custody applications where user balances are not held by any company whatsoever. In the cohort from the 2010s, that figure was 4%. Regardless of your view on cryptocurrency, builders have already voted with their feet.
So yes, the rise is real: 368 companies, three entirely distinct waves (254 traditional challenger banks, 58 fiat-crypto hybrid apps, and 56 web3-native apps), underpinned by 106 infrastructure providers and backed by 219 investors—all mapped out.
Next is the part that no one will include in the BP.
Out of 368, only 127 hold full banking licenses.
Read it again. Two-thirds of the “banks” in your phone’s app store aren’t banks at all. Their right to exist is rented—from a sponsor bank, an electronic money license, or some card issuer you’ve never heard of. And their customers almost never know which side of that line they’re on.
Banking license
This is not a technical detail issue, but a core structural risk across the entire industry, with lives on the line:
WaveCrest, 2018: Visa revoked the qualification of a card issuer, causing dozens of crypto card projects to die overnight.
Wirecard, 2020: A payment processor exposed a €1.9 billion hole, freezing funds from a wave of European "banks" whose only mistake was being built on top of it.
Synapse, 2024: A BaaS intermediary collapsed, and ordinary Americans discovered that "FDIC insured" didn't mean what they thought—because the very ledger recording who owned what money was the one that failed.
Ready, 2026: The same movie, with a new cast.
When a real bank fails, deposit insurance pays out. When a new bank’s infrastructure fails, customers receive a queue number in the bankruptcy process.
Death in this industry is silent. That is the real scandal.
After maintaining this dataset, one thing I didn’t anticipate was that deletions never stopped.
Who is in control?
This month alone, five entities vanished from the list—either liquidated, absorbed through merger, or quietly transformed into something else. No press releases, no after-action reviews. New banks don’t die as dramatically as FTX did.
The app simply stopped updating, customer service ceased responding, and one day the domain redirected to a partner’s landing page—hundreds of thousands of customers either had to migrate or disappeared entirely.
No one writes obituaries for new banks. Fintech media covers product launches and funding rounds because that’s where the advertising dollars and interview opportunities lie. As a result, this graveyard remains invisible, and every new founder stumbles into the same five traps, thinking they’re the first to see through them.
This is why we track exits with the same care as entries. Failed data is more valuable than funding data. You can’t learn anything from press releases.
But will AI solve the profitability model issue? Really?
Now every new bank pitch deck includes AI, so we reviewed all 368 one by one, cross-referencing financial reports, regulatory disclosures, and evidence of actual deployment—not marketing pages.
67 have been approved. 18%. Over 300 others are either still in pilot phase, or "exploring," or claiming their partner’s model as their own.
AI new bank?
The counterintuitive part is that the AI models performing best are not from well-known companies, but from emerging market lenders in Nigeria, the Philippines, Mexico, and Bangladesh. In these regions, credit reporting systems are virtually nonexistent; a model that can extend credit to those with no credit history isn’t just a nice-to-have feature—it’s the very reason this business exists.
The West talks about AI banks; the Global South is actually building them, because it has no other choice.
What does this map actually show?
Look at the infrastructure segment on the map: 106 providers support 368 consumer-facing brands. Within this box, a handful of custodial banks, BaaS platforms, and card processors each carry dozens of logos above them. The concentration invisible to consumers is precisely the concentration that will give rise to the next Synapse.
This is what the industry looks like in 2026 at its most honest: a spectacular, truly world-changing rise, where 1.5 billion people gained access to banking services through a single app—many for the first time in their lives—built atop a foundational layer that most customers have never heard of, where two-thirds of companies won’t survive a bad quarter from their landlord.
Both things are true at the same time. That’s exactly what makes it interesting.
Three predictions I’m willing to admit might be wrong
The licensing gap will narrow from both ends. Strong unlicensed players will acquire or obtain licenses, while weaker ones will be eliminated by 2027. The middle ground will disappear.
The first AI credit default will occur within two credit cycles. Of the 67 models already in production, most have never experienced a true downturn in their current form. Some are about to learn what was missing from their training data.
The next wave of customers isn't human. Banking infrastructure for AI agents—including wallets for agents, cards issued by machines, and machine-to-machine payments—is currently being built by only seven companies. It looks just like web3-native in 2021: small, strange, and structural.