A new industry review shared by Consensys / MetaMask developer relations lead Francesco Andreoli says the global neobank boom may be sitting on a weak structural foundation. Writing on X on July 27 Taipei time, Andreoli said 368 active neobanks now report a combined 1.46 billion users, but only 127 of them hold full banking licenses.
Asia leads the market by user count
Andreoli said he spent significant time tracking and verifying 368 active neobanks around the world. Based on self-reported figures from those firms, total users reached 1.46 billion.
Asia accounted for 817 million of that total, making it the largest regional market in the dataset. In China alone, WeBank was listed at more than 400 million users. By comparison, Revolut, one of Europe’s best-known neobanks, was cited at more than 50 million users.
Web3-native banking apps have gained share
The report said crypto and decentralized technology are playing a larger role in the sector’s make-up. Among neobanks founded in the 2020s and still operating, 30% were classified as Web3-native self-custodial applications, or 56 firms. In the 2010s, that share was 4%.
Andreoli also counted 58 applications built around a hybrid model that combines fiat and crypto services.
Most firms rely on sponsor banks, e-money licenses, or BaaS providers
The report’s central warning was a large licensing gap. Out of the 368 firms reviewed, only 127 held full banking licenses. That leaves about two-thirds relying on sponsor banks, e-money licenses, or banking-as-a-service, or BaaS, platforms to operate.
Andreoli described that setup as the sector’s core structural risk and said many consumers do not realize how dependent their provider is on third-party infrastructure and licensing arrangements.
He pointed to several past failures. In 2018, Visa revoked WaveCrest’s issuing rights, which led to the collapse of dozens of crypto card projects. In 2020, Wirecard’s 1.9 billion euro hole froze multiple European fintech and neobank operations. In 2024, the failure of U.S. BaaS intermediary Synapse left many customers discovering that their funds were not protected by the Federal Deposit Insurance Corporation, or FDIC, in the way they had assumed.
Andreoli argued that when a traditional bank fails, deposit insurance can still provide a backstop for customers. When the underlying infrastructure behind a neobank breaks down, customers can end up waiting through bankruptcy proceedings instead.
He added that five institutions had quietly liquidated or ceased operations this month alone, forcing hundreds of thousands of customers to migrate without a formal press release or face the disappearance of their assets. He called that pattern a “silent death.”
AI in production was found at only 67 firms
The report also pushed back on how AI adoption is being presented across fintech. After reviewing the 368 institutions, Andreoli said only 67 had actually deployed AI in production, or about 18% of the total.
Most of the remaining firms were described as still being at the pilot stage or using partner-built models as part of marketing claims rather than live core operations.
The lenders making real use of AI were not primarily large Western firms, according to the report. Andreoli said more practical deployment is showing up in emerging markets such as Nigeria, the Philippines, Mexico, and Bangladesh, where weak traditional credit bureau coverage makes AI-based underwriting more central to the business model.
As he put it: “The West is talking about AI banks, while the Global South is shipping.”
Three predictions for the sector
Andreoli also outlined three forward-looking calls based on the dataset.
First, he said the licensing gap is likely to narrow from both ends: stronger brands may acquire or secure their own licenses, while weaker brands may be pushed out by 2027, leaving less room for middle-tier players.
Second, he warned that the first major AI underwriting blow-up could arrive within the next two credit cycles because many of the 67 production models have not yet been tested through a real economic downturn.
Third, he said the next wave of customers may not be human at all, but banking rails built for AI agents, including agent-operated wallets and machine-to-machine payments. Only seven institutions are exploring that direction now, the report said, but Andreoli argued it could become a structural growth area in the same way Web3-native apps did in 2021.