Is this new financial infrastructure or a fake stitched-together platform: what is a Web3 Neobank doing?

Bank the Unbanked, the original promise of financial inclusion

“Bank the Unbanked” was once one of the most morally compelling slogans in the fintech industry. What it promised was not shrinking bank branches into a prettier mobile app, but giving people who were excluded by geography, income, identity documents, credit history, and cross-border costs—perhaps for the first time—an account to receive payments, make payments, and store value. At that stage, the account itself was scarce: without an account, wages, remittances, savings, and credit were all difficult to access through the formal financial system. What early Neobanks sought to lower wasn’t the number of financial products—it was the entry threshold for financial services.

If we compress the early Neobank story, they mainly did three things: moving account opening from physical branches to mobile phones, making confusing account fees more transparent, and putting cards, FX exchange, and cross-border remittances into a single interface. The industry’s typical companies—Revolut, Nubank, and Chime—did not follow the same path: some later obtained banking licenses, while others continued to rely on partner banks. But they all achieved a distribution revolution. By the end of 2025, this revolution had already produced real account relationships: Revolut’s annual report disclosed 68.3 million retail customers, though that’s not the active user metric; documents Nubank filed with the U.S. Securities and Exchange Commission (SEC) disclosed 131 million customers and a 83.4% monthly active rate under the company’s definition; Chime disclosed active members who moved funds during the most recent calendar month reached 9.5 million. Together, these three figures show that account experience, service speed, and fee transparency are no longer just product packaging—they can be converted into ongoing financial competitiveness.

Then Web3 raised the issue in a different way. Today’s Neobank accounts aren’t limited to fiat balances; they may also include stablecoins, self-custody wallets, crypto cards, fiat on/off-ramp channels, cross-chain swaps, decentralized finance (DeFi), and real-world asset (RWA) yield. Funds appear able to move around the clock among banks, cards, and blockchains, and financial entry points are more abundant than ever. But the more entry points there are, the harder it becomes to avoid another question: does more access mean more financial rights, or does it mean users must simultaneously assume more custody, contract, liquidity, and counterparty risks they may not even understand?

Does a Web3 Neobank continue the mission of Bank the Unbanked, or does it repackage banks, cards, wallets, and yield agreements into a one-stop interface? The judgment can’t stay at the feature list. Who gets services that didn’t exist before, and where do fees and yields come from? When an account is frozen, a card payment fails, a counterparty exits, or an on-chain agreement is paused, who can explain and take responsibility? Access inclusion is only the first step. The decision of whether it resembles financial infrastructure or a stitched-together interface is whether fragmented rights, rules, and remediation mechanisms can be reorganized.

From bank accounts to on-chain accounts: what has changed for Neobank

Understanding a bank account is inseparable from the balance sheet. Banks take deposits, extend loans, and participate in clearing—maintaining capital buffers amid credit creation, maturity mismatch, and liquidity management. Regulation, deposit insurance, and resolution mechanisms define what users can claim after risks materialize and what the bank must bear. In legal and accounting terms, an account balance isn’t a stack of cash locked away in a vault; it’s the user’s claim against the bank, and also a liability on the bank’s balance sheet. A mobile app may display a balance, but it can’t grant the capability to accept deposits, issue credit, or absorb losses just through a screen.

The first thing traditional Neobanks moved was the counter, not the vault. Licensed digital banks can operate loan and deposit businesses through digital channels that resemble those of traditional banks. BaaS (banking-as-a-service) or partner-bank Neobanks hand account, card, and payment experiences to users, while keeping the license, clearing, and some risk responsibilities with the partner bank, the card issuer, or the payment institution. Chime explicitly stated it is not a bank, and that banking services are provided by partner banks—this is the most direct example of that structure. Both product types look “lightweight,” but “lightweight” comes from entirely different sources. For users, branding, interfaces, and customer service entry points may belong to the same company, while custody of funds, card issuance, and dispute handling may correspond to three separate contracts. Neobank didn’t eliminate bank responsibility; it just hid that responsibility behind a smoother back end.

On top of this, Web3 Neobanks make accounts “wider.” Stablecoins bring some dollar liquidity on-chain; wallets let users or smart contracts control assets directly; fiat on/off-ramps connect fiat on-ramps and off-ramps; DeFi and RWA provide new paths for asset allocation; and cross-chain routing helps the same piece of money find settlement and liquidity across different networks. Its strength is composability—payments, swaps, and yields can be called in sequence. Its problem is also composability: the more layers of protocol, custody, or liquidity source you add, the more new control points and potential failure points you create. More assets connected to an on-chain account don’t automatically grow bank capital, deposit insurance, or unified consumer protection.

Showing as $1 is not the same as users holding the same kind of money. Revolut’s 2025 annual report disclosed a total customer balance of £50.18B. Bank deposits, dollars in partner accounts, stablecoins custodied by the platform, and USDC in users’ own wallets correspond to different issuers, redemption paths, freezing permissions, deposit protection, and default risks. One side offers stronger self-management, around-the-clock settlement, and composability; the other side offers more mature risk absorption and remediation mechanisms. They’re not necessarily inherently superior or inferior, but they can’t be lumped together under the same dollar symbol. An on-chain account isn’t the same as a bank account; similar balances may come with completely different claim holders and risk bearers.

Who is served, and what is served: the real users of Neobank

“Do you have a bank account?” isn’t a black-and-white question. Users not served by bank services (Unbanked) may not even have stable account access and identity entry points; underbanked users already have accounts but are still kept outside effective service by cross-border fees, currencies, account-opening regions, settlement times, and product availability. A person who nominally owns a bank account doesn’t mean they can receive overseas income at low cost; a business that holds dollars doesn’t mean it can smoothly pay global suppliers. Financial inclusion can’t only count how many accounts are opened—it must also examine whether these accounts help users continuously complete receiving payments, making payments, and saving value, and provide remediation when things fail.

When banks, cards, and blockchains all enter the same company, friction shifts from “can we transfer money” to “can we complete the business.” An independent developer serving global customers might use e-statements or payment links to receive stablecoins, then convert part of the funds into local fiat; a Web3 business might manage bank accounts, multi-signature wallets, business cards, and on-chain treasuries simultaneously, while also requiring employees, vendors, and the finance team to follow different limits and approval processes; a finance person who sees an on-chain transfer that contains only an address and a transaction hash still has to go back to spreadsheets, chat logs, and the bank back end to complete invoices, contracts, notes, and accounting classifications. What users lack isn’t another button, but a system that can put different money “tracks” back into the same business context.

A few representative paths currently on display just happen to show that Web3 Neobank is not a homogeneous track. AllScale started with e-statements; its official announcement says its consumer and enterprise products have formed over 1.5 million registered wallets. Infini emphasizes enterprise payments, fiat bridging, approvals, reconciliation, and finance-lead workflows; its website says it has served over 100k users and supports more than 180 countries. Bitget Wallet represents a consumer-side wallet entry; a July 2026 announcement from the company says cumulative users exceed 100 million, monthly active users (MAU) reach 40 million, and card issuance exceeds 150k. In the first half of 2026, card spending reached $31 million, up 191% from the second half of 2025. Reah tries to put banks, cards, wallets, Treasury, and Agents into a unified control layer; its website claims its service capabilities can cover 150 countries. So the four paths get split: some compete for access points, some build vertical workflows, and others attempt to control the rules across all tracks. Strategies differ, yet they are all growing.

However, financial management isn’t a belief test that must be migrated to Web3. Users with strong local banking services, simple fund routes, emphasis on deposit insurance, no need for stablecoins, and no willingness to take private key and smart contract risks have no reason to change their account structure for something “more advanced.” Web3 Neobank is valuable only when complexity already exists: it doesn’t actively add new tracks—it reduces the cost of operating across tracks. Stablecoins can become underlying liquidity, and Neobanks can shape user mental models; whether users stay ultimately depends on whether the platform can complete receiving, paying, approvals, reconciliation, and exit end-to-end.

Free is only the entry point—where did the risk and fees go

Free is a price, not a business model. Free accounts, low FX exchange fees, cashback, Gas (on-chain transaction fees) subsidies, and high APY (annual percentage yield) can lower the barrier to first-time use, but the platform still needs to generate revenue from card transaction revenue shares, payments and FX, fiat on/off-ramp flows, subscriptions, SaaS, APIs, Treasury services, or partner compensation. Even if users don’t pay when opening accounts, the system may be paid through spreads, capital float, transaction behavior, or later value-added services. After scaling, these revenue sources can indeed form profits: Revolut’s 2025 revenue was £100k and net profit was £150k; Nubank achieved revenue of $4.52B and net profit of $1.31B in the same year under IFRS (International Financial Reporting Standards). Charging itself isn’t suspicious; what needs investigation is whether pricing is transparent, whether revenue aligns with the value users receive, and whether users can exit without bearing abnormal losses.

No branches doesn’t mean no costs. Chime’s 2025 revenue reached $15.78B, yet GAAP (U.S. Generally Accepted Accounting Principles) net loss was about $1.01 billion; but the same SEC filing also shows equity incentive awards and related tax expenses of about $2.87B that year, and the company’s defined adjusted EBITDA (profit before interest, taxes, depreciation, and amortization) was positive at $126.6 million. Note: these three figures must be viewed together—net losses can’t be wiped out, adjusted metrics can’t be treated as audited profits, but reading only one of them can still misjudge changes in operations. Research has already reminded that digitization can reduce some offline operating costs, but it doesn’t eliminate spending on funds, technology, marketing, compliance, customer support, and risk management. In Web3, the back end is actually longer: bank partners, card issuers, payment processors, KYC/KYB, wallet security, smart contract audits, on-chain liquidity, and abnormal transaction handling all require ongoing payment. Normal payments can be automated; but an abnormal payment may simultaneously pull in compliance, customer support, partner teams, and engineering teams. The longer the cooperation chain, the more the platform needs to cover these costs—costs that aren’t easily visible to users—through scale, fee-based revenues, or workflow stickiness.

High APY is the most attractive and the easiest to mislead. Yield shown on a Treasury page may come from on-chain lending, liquidity allocation, tokenized real-world assets, protocol subsidies, or third-party credit. The specific yield rate, underlying strategy, fees, and available regions must be judged line-by-line from each platform’s product pages and terms; you can’t extrapolate from one company to the entire sector. If these products are described merely as “earning interest on idle cash,” then credit, maturity, liquidity, smart contract, and redemption risks get compressed into a single number. On-chain transparency may make some positions and transaction routes easier to observe, but it can’t guarantee that underlying assets won’t default, protocols won’t be attacked, or liquidity won’t disappear. High yield doesn’t appear out of thin air—it’s just split, packaged, transferred, and then re-presented through a simpler account interface.

So stitching isn’t the sin; it’s aggregation without orchestration that is. Modern finance is already completed together by banks, card organizations, clearing networks, payment institutions, and technology providers; the issue was never whether there’s a third party, but whether third parties can form a unified service. Aggregation just places multiple entry points together; orchestration still has to handle state synchronization, rule priority, permission passing, and failure rollback. The former reduces the cost for users to search for tools, while the latter reduces the cost of completing the business. If you only aggregate accounts, cards, wallets, and yield, but leave the ledger, policies, and responsibilities scattered, that’s a pseudo-stitch. If you can unify the ledger, policy, audit, partner disclosures, exception handling, and recovery paths, then you begin to resemble infrastructure.

****The best way to judge this boundary isn’t to review the feature list again, but to place the platform into moments of failure. When a card is declined, who can identify the reason? When an account is frozen, who can explain the status of the funds? When a partner bank exits, who is responsible for migration? When an on-chain protocol pauses redemption, who explains losses and the exit path? When an Agent pays out of authority, who can pause, track, and remediate? The platform may not need to absorb every kind of underlying loss, but it must make users understand where the risks are, who has control, and who to contact if something goes wrong. Stitching itself isn’t the problem. The real problem is: when one link fails, who can reconnect the entire responsibility chain.

When Agents start paying, will Neobank become more “Neo”?

Agentic Payment (intelligent agent payments) isn’t an upgrade to payment without requiring a password. A typical automatic debit follows fixed rules chosen by the user in advance; an Agent may instead search for products on the user’s behalf, compare conditions, select merchants, create orders, and complete the payment. The system needs to verify not only the account and password, but also who this Agent is, who it represents, what task it’s performing, and how much money and for how long—and with what decision margin—the user has actually authorized. Payment shifts from a trust layer built solely at the transaction endpoint to a trust layer composed of identity, intent, permissions, and responsibility.

Neobank and Agents seem like a better fit—not because AI has finally learned to swipe a card, but because accounts already have the potential to be read and constrained by programs. A Policy Engine can encode budgets, merchants, categories, time, tasks, and approval conditions into rules that machines can execute; an Audit Trail records who initiated, who approved, what policy was executed, and what ultimately happened. Traditional card authorization typically only answers “can it be paid,” while Agent authorization must also answer “why it’s paid,” “under what conditions it continues,” and “when it must stop to find a person.” Reah and Rain’s official materials treat limited-use cards and pre-transaction controls as product directions; Infini’s official website places the AI Agent into expense management, reconciliation, and enterprise finance operations. These materials may prove what the vendor is designing, but they can’t prove that Agentic Payment has been adopted at scale, because all three companies have not published comparable Agent payment volumes, failure rates, or human takeover data. The direction remains clear: what AI needs isn’t a new button, but a machine-readable account policy.

Authorization also won’t jump from manual clicks directly to fully autonomous execution. The first layer remains per-transaction confirmation, applicable to high amounts, unfamiliar merchants, and irreversible transactions. The second layer is rule-based automated payments, where lower amounts, high frequency, and predictable outcomes are completed within fixed time windows and limits. The third layer is the enterprise budget pool, where Agents continue to execute under roles, merchant allowlists, multi-level approvals, and real-time limits. The stronger the Agent’s execution capability, the more the rules need to be observable, pausable, and revocable. Automation boundaries should expand with remediation capability—not with marketing terminology.

Otherwise, AI will amplify both efficiency and errors. Prompt injection can change the task objective; malicious plugins can forge merchants; wrong context can cause duplicate payments or spend beyond authority. Refunds, chargebacks, and on-chain irreversible transactions also correspond to different responsibilities and remediation paths. A mature system needs emergency freezing, permission revocation, human takeover, and full traceability, with stricter human thresholds for high-risk actions. What truly matches AI isn’t “AI + a card,” but “Agent + policy-controlled accounts.”

Looking ahead: from yield entry points to a robust financial system

To understand whether a Web3 Neobank is mature, metrics like registered users, cards issued, number of transactions, TVL, and the highest APY are not enough. These numbers can prove the platform is being accessed, but they can’t prove it has assumed the functions of a financial system. Subsidies can drive card issuance; market conditions can boost TVL; a single hot event can generate transaction volume. Only repeated real business can turn entry points into account relationships. More meaningful indicators are whether real business users continue to stay, the payment success rate and exception rate, how long it takes to recover after failures, whether customer support and compliance costs are controllable, and how much funds movement truly passes through the platform’s policies, approvals, and audits. Fund flows controlled by policies aren’t an industry standard yet, but it re-centers evaluation on one question: what exactly the platform controls, and what it’s willing to be responsible for.

“Robust” doesn’t mean rebuilding Web3 as a closed bank. First, it means rights and responsibilities are clear: fund locations, custody methods, and yield sources must be transparent; different money tracks must operate under consistent policies; and in case of failures there must be customer support, recovery, and business continuity. Business continuity isn’t just servers being online—it also includes replacing counterparties, key recovery, account migration, dispute handling, and manual degradation. When automated links aren’t available, the enterprise still needs to know how to retrieve funds, resume payments, and complete audits. Users can choose higher-risk or more self-directed account structures, but risk sources, control methods, and exit conditions must be visible.

Yield and an operating system aren’t simply an either-or choice. Enterprises can use stablecoins, RWA, or DeFi to improve the efficiency of some idle capital; but for Treasury, the first priority is solving cash visibility, liquidity, permissions, approvals, and auditability—then comes yield enhancement. A healthy financial system can accommodate yield, but it can’t prove its value by the highest APY alone, nor can it use yield to cover underlying responsibility and liquidity gaps. Low yield isn’t inherently safer, and RWA and DeFi aren’t the same category of risk. True maturity is enabling every form of allocation to fall into a framework that is understandable, limitable, and detachable.

As you can see, capital uses money to vote, but the terminal platforms and the underlying infrastructure don’t get priced the same way. AllScale disclosed $1.5 million in funding in June 2025 and $5 million for its seed round in December. Rain, which provides stablecoin payment infrastructure, completed a $250 million C round in 2026, with a valuation of $1.95 billion and cumulative funding exceeding $338 million. Rain’s official announcement also states its annualized transaction volume exceeds $3 billion and it has more than 200 partners. This gap at least indicates that capital currently prefers to pay a premium for compliant payment rails that can be called by multiple terminals, while terminal Web3 Neobanks are still proving whether user scale can translate into stable business. From this, the “end state” of Web3 Neobank may be divided into three scenarios: in the optimistic case, a small number of platforms integrate banks, cards, stablecoins, and on-chain protocols into a trusted cross-rail control layer; in the neutral case, most platforms remain specific to certain regions and users, or stick to vertical workflows such as e-statements, payments, card controls, and Treasury; in the pessimistic case, platforms that rely on subsidies, ambiguous responsibility, and a single partner will gradually exit after channel interruptions, risk events, or tighter regulation. All three scenarios map to the same maturity standard: it’s not about exposing users to more financial products, but about putting those products into a financial order that can run continuously.

Stitching isn’t the endpoint—assurance is the boundary

Back to Bank the Unbanked: financial inclusion has never been just about letting more people see an account, get a card, or enter a yield product for the first time. True inclusion means users receive services that are understandable, usable long-term, and can be exited when necessary—and that, if something fails, they can get explanations and remediation. Traditional banks also suffer from high fees, slow service, and exclusion of certain users, and expanding access through Web3 still has value. But only when a trusted responsibility relationship is built behind those access points does that value solidify from product experience into institutional capability.

Back to the question: whether a Web3 Neobank is new financial infrastructure or a pseudo-stitching platform doesn’t depend on whether it relies on third parties. If you only aggregate accounts, cards, wallets, and yield products, then leave rule conflicts and failure costs to partners and users, that’s pseudo-stitching. If you can unify control, recording, explanation, recovery, and remediation, then it may become new financial infrastructure. Most platforms today are still in the stitching stage, but stitching may also be a necessary starting point toward a unified system across financial rails. The end state of a Web3 Neobank is hard to be only a better wallet or a cheaper card. The true boundary is whether the market is willing to hand over funds, rules, and responsibilities to it long-term.