New Financial Infrastructure or Pseudo-Sewing Platform: What Web3 Neobank Is Doing

Bank the Unbanked: The Initial Commitment to Financial Inclusion

"Bank the Unbanked" was once one of the most ethical slogans in the fintech industry. Its promise was not to condense bank branches into a more aesthetically pleasing mobile app, but to allow those excluded by geography, income, identity documents, credit history, and cross-border costs to finally have an account to receive, pay, and save value. At that stage, the account itself was a scarce commodity: without an account, wages, remittances, savings, and credit struggled to enter the formal financial system. Neobanks initially aimed to lower the barriers to access financial services, not to increase the number of financial products.

Let’s compress the story of early Neobanks: they primarily did three things: moved account opening from branches to mobile phones, made confusing account fees more transparent, and integrated cards, currency exchange, and cross-border remittances into a single interface. The paths taken by typical companies in the industry, such as Revolut, Nubank, and Chime, varied; some later obtained banking licenses, while others continued to rely on partner banks, but they collectively completed a distribution revolution. By the end of 2025, this revolution had solidified real account relationships: Revolut's annual report disclosed 68.3 million retail customers, but this is not based on active users; Nubank disclosed 131 million customers in documents submitted to the U.S. Securities and Exchange Commission (SEC), with a monthly active rate of 83.4% as defined by the company; Chime reported that the number of active members who had moved funds in the most recent calendar month reached 9.5 million. These three figures collectively indicate that account experience, service speed, and fee transparency have become not just product packaging but can be transformed into sustainable financial competitiveness.

Then Web3 posed this question in another way. Today’s Neobank accounts no longer contain only fiat currency balances; they may also include stablecoins, self-custody wallets, crypto cards, fiat on/off-ramps, cross-chain exchanges, decentralized finance (DeFi), and real-world asset (RWA) yields. Funds seem to be able to move between banks, cards, and blockchains around the clock, with more financial entry points than ever before. However, the more entry points there are, the harder it is to avoid another question: do more entry points mean more financial rights, or do they mean users need to bear more risks related to custody, contracts, liquidity, and counterparties that they may not fully understand?

Is Web3 Neobank continuing the mission of Bank the Unbanked, or is it repackaging banks, cards, wallets, and yield protocols into a one-stop interface? Judgments cannot stop at the feature list. Who gains access to services previously unavailable, where do the fees and yields come from; when accounts are frozen, card payments fail, partners withdraw, or on-chain protocols pause, who can explain and bear responsibility? Universal access is just the first step. Whether it can reorganize the dispersed rights, rules, and remedy mechanisms will determine whether it is closer to financial infrastructure or merely a patchwork interface.

From Bank Accounts to On-Chain Accounts: What Neobanks Have Changed

Understanding bank accounts cannot avoid the balance sheet. Banks absorb deposits, issue loans, and participate in clearing, maintaining capital buffers between credit creation, maturity mismatches, and liquidity management; regulation, deposit insurance, and disposal mechanisms define what users can claim after risks occur and what banks need to bear. From a legal and accounting perspective, account balances are not just a pile of cash locked in a vault; they represent users' claims against banks and are also liabilities on the bank's balance sheet. Mobile applications can display balances but cannot gain the ability to absorb deposits, allocate credit, and bear losses through a screen.

The first thing traditional Neobanks moved was the counter, not the vault. Licensed digital banks can operate deposit and loan businesses similar to banks through digital channels; Banking as a Service (BaaS) or partner bank-type Neobanks hand over account, card, and payment experiences to users while leaving licenses, clearing, and some risk responsibilities with partner banks, issuing institutions, or payment entities. Chime has explicitly stated that it is not a bank, and banking services are provided by partner banks, which is the most direct example of this structure. The two types of products may seem "light," but their sources of lightness are entirely different. For users, the brand, interface, and customer service entry may belong to the same company, but fund custody, card issuance, and dispute resolution may correspond to three different contracts. Neobanks have not eliminated bank responsibilities; they have merely hidden those responsibilities in a smoother experience.

Web3 Neobanks, on this basis, have made accounts "broader." Stablecoins bring some dollar liquidity on-chain, wallets allow users or smart contracts to directly control assets, fiat on/off-ramps connect fiat entry and exit points, DeFi and RWA provide new paths for fund allocation, and cross-chain routing helps the same amount of money find settlement and liquidity across different networks. Its advantage is composability; payments, exchanges, and yields can be continuously invoked; its problem is also composability; each additional layer of protocol, custody, or liquidity source adds a new control point and failure point. On-chain accounts connect to more assets but do not automatically generate bank capital, deposit insurance, and unified consumer protection.

The same display of $1 does not mean users are holding the same kind of money. Revolut's 2025 annual report disclosed a total customer balance of £50.183 billion. Bank deposits, dollars in partner accounts, stablecoins held by the platform, and USDC in users' wallets correspond to different issuers, redemption paths, freezing permissions, deposit insurance, and default risks. On one side is stronger self-management, around-the-clock settlement, and composability; on the other side are more mature risk absorption and remedy mechanisms. The two are not necessarily clearly superior to each other but cannot be conflated under the same dollar symbol. On-chain accounts do not equal bank accounts; while balances may be similar, the rights to claim and risk bearers may be completely different.

Who Is Served, What Is Served: The Real Users of Neobanks

"Do you have a bank account?" is not a black-and-white question. Unbanked users may not even have stable accounts and identity entry; underbanked users already have accounts but are still blocked from effective services by cross-border fees, currencies, account opening regions, arrival times, and product availability. A person may nominally have a bank account, but that does not mean they can receive overseas income at low cost; a business holding dollars does not mean it can smoothly pay global suppliers. Financial inclusion cannot just count how many accounts have been opened; it must also examine whether these accounts can help users continuously complete receiving, paying, and value preservation, and provide remedies in case of failure.

When banks, cards, and blockchains enter the same company, friction shifts from "can I transfer money" to "can I complete business." An independent developer serving global customers may use electronic invoices or payment links to receive stablecoins and then convert some funds into local fiat; a Web3 enterprise may simultaneously manage bank accounts, multi-signature wallets, corporate cards, and on-chain treasuries, while requiring employees, suppliers, and finance teams to adhere to different limits and approvals; finance personnel seeing an on-chain transfer with only an address and transaction hash must return to spreadsheets, chat records, and bank backends to complete invoices, contracts, notes, and accounting classifications. What users lack is not another button, but a system that can put different funding tracks back into the same business context.

Currently, several representative paths illustrate that Web3 Neobanks are not a homogeneous track. AllScale has entered through electronic invoices, with the company officially announcing that its consumer and enterprise products have formed over 1.5 million registered wallets; Infini emphasizes enterprise payments, fiat bridging, approvals, reconciliation, and workflows for finance heads, claiming to have served over 100,000 users across 180 countries. Bitget Wallet represents the consumer wallet entry, with the company announcing in July 2026 that it has accumulated over 100 million users, with monthly active users (MAU) reaching 40 million and issuing over 150,000 cards; in the first half of 2026, card spending reached $31 million, a 191% increase from the second half of 2025. Reah attempts to integrate banks, cards, wallets, Treasury, and Agents into a unified control layer, claiming its service capabilities can cover 150 countries. Thus, the four paths are separated: some compete for entry, some focus on vertical workflows, and others attempt to control the rules between all tracks, with different strategies but all thriving.

However, financial management is not a faith test that must be migrated to Web3. Users with sufficient local bank services, simple funding paths, a focus on deposit protection, no need for stablecoins, and unwillingness to bear the risks of private keys and smart contracts have no reason to change their account structures for the sake of being "more advanced." Web3 Neobanks only have value when complexity already exists: they do not actively add tracks but reduce the costs of operating across tracks. Stablecoins can become underlying liquidity, and Neobanks can provide user mentality; what ultimately determines whether users stay is whether the platform can complete receiving, paying, approvals, reconciliation, and exits comprehensively.

Free is Just an Entry Point: Where Do Risks and Costs Go?

Free is a price, not a business model. Free accounts, low exchange fees, cashback, gas subsidies, and high APY can lower the threshold for first-time use, but platforms still need to generate revenue from card transaction splits, payments and FX, fiat on/off-ramps, subscriptions, SaaS, API, Treasury services, or partner compensation. Users may not pay when opening an account but could pay for the system through price differences, fund deposits, trading behavior, or subsequent value-added services. After scaling, these revenue sources may indeed form profits: Revolut reported £4.516 billion in revenue and £1.305 billion in net profit in 2025; Nubank achieved revenue of $15.775 billion and net profit of $2.872 billion in the same year according to IFRS standards. Charging itself is not suspicious; what needs to be studied is whether the pricing is transparent, whether the revenue aligns with the value received by users, and whether users can exit without bearing abnormal losses.

There are no physical outlets, but that doesn't mean there are no costs. Chime's revenue reached $2.187 billion in 2025, yet its GAAP net loss was approximately $1.01 billion; however, the same SEC filing also indicated that there were about $1.093 billion in equity incentives and related taxes that year, with the company's defined adjusted EBITDA (Earnings Before Interest, Taxes, Depreciation, and Amortization) being a positive $126.6 million. It is important to consider these three figures together: net losses cannot be erased, and adjusted metrics cannot be treated as audited profits, but focusing on just one will lead to misjudgments about operational changes. Research has already indicated that digitization can reduce some offline operational costs, but it will not eliminate expenses related to funding, technology, marketing, compliance, customer support, and risk management. In the Web3 era, the backend is even longer: banking partners, card issuers, payment processors, KYC/KYB, wallet security, smart contract audits, on-chain liquidity, and anomaly transaction handling all require ongoing payments. Normal payments can be automated, but an anomalous payment may simultaneously involve compliance, customer service, partners, and engineering teams. The longer the cooperation chain, the more platforms need to cover these less visible costs with scale, fee-based income, or workflow stickiness.

High APY is the most attractive but can easily mislead. The yields on the Treasury page may come from on-chain lending, liquidity provisioning, tokenized real assets, protocol subsidies, or third-party credit; specific yield rates, underlying strategies, fees, and available regions must be judged against each platform's product page and terms at that time, and cannot be extrapolated from one company to the entire sector. If these products are merely described as "idle money earning interest," credit, duration, liquidity, smart contracts, and redemption risks will be compressed into a single number. On-chain transparency can make some positions and transaction paths easier to observe, but it cannot guarantee that underlying assets will not default, protocols will not be attacked, or liquidity will not disappear. High returns do not arise out of thin air; they are merely split, packaged, transferred, and then presented again through a simpler account interface.

Therefore, stitching is not the original sin; only aggregation without orchestration is. Modern finance is inherently completed by banks, card organizations, clearing networks, payment institutions, and technology providers; the issue has never been whether there are third parties, but whether those third parties can form a unified service. Aggregation merely places multiple entry points together, while orchestration must handle state synchronization, rule prioritization, permission transfer, and failure rollback. The former reduces the cost for users to find tools, while the latter reduces the cost of completing business. Simply aggregating accounts, cards, wallets, and yields while allowing ledgers, policies, and responsibilities to remain dispersed is pseudo-stitching; being able to unify ledgers, policies, audits, partner disclosures, anomaly handling, and recovery paths is what begins to approach infrastructure.

The best way to judge this boundary is not to look at the feature list again, but to place the platform in a moment of failure. When a card is declined, who can pinpoint 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 suspends redemptions, who explains the losses and exit paths? When an Agent makes an unauthorized payment, who can pause, track, and remedy? A platform may not bear every type of underlying loss, but it must make users aware of where the risks lie, who has control, and who to turn to when problems arise. Stitching itself is not the issue. The real problem is, when one link fails, who can reconnect the entire chain of responsibility.

When Agents start to pay, will Neobanks become more "Neo"?

Agentic Payment is not an upgraded version of password-free payments. Ordinary automatic deductions execute fixed rules chosen by the user in advance, while Agents may represent users in searching for products, comparing conditions, selecting merchants, creating orders, and completing payments. What the system needs to verify is no longer just the account and password, but who this Agent is, who they represent, what task they are completing, and how much money, for how long, and how much judgment space the user has authorized. Payments are shifting from a confirmation action at the transaction endpoint to a trust layer composed of identity, intent, permissions, and responsibilities.

Neobanks and Agents seem to fit better, not because AI has finally learned to swipe cards, but because accounts are now capable of being read and constrained by programs. The Policy Engine can write budgets, merchants, categories, times, tasks, and approval conditions into rules that machines can execute, while the Audit Trail records who initiated, who approved, under what policy it was executed, and what happened in the end. Traditional card authorization usually only answers "Can I pay?"; Agent authorization must also answer "Why pay, under what conditions continue, and when must I stop to consult someone?" Official materials from Reah and Rain view limited-use cards and pre-transaction controls as product directions; Infini's website places AI Agents into expense management, reconciliation, and corporate financial operations. These materials can prove what vendors are designing, but they cannot prove that Agentic Payment has been widely adopted, as none of the three companies have publicly disclosed comparable Agent payment volumes, failure rates, or manual takeover data. The direction remains clear: AI does not need a new button, but a machine-readable account policy.

Authorization will not jump directly from manual clicks to complete autonomy. The first layer remains per-transaction confirmation, suitable for high amounts, unfamiliar merchants, and irreversible transactions; the second layer is rule-based automatic payments, allowing low amounts, high frequency, and predictable outcomes to be completed within fixed times and limits; the third layer is the corporate budget pool, where Agents continuously execute under roles, merchant whitelists, 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. The boundaries of automation should expand with remedial capabilities, not with the promotional rhetoric of models.

Otherwise, AI will amplify both efficiency and errors. Prompt injection may change task objectives, malicious plugins may forge merchants, and erroneous contexts may lead to duplicate payments or unauthorized expenditures; refunds, chargebacks, and on-chain irreversible transactions correspond to different responsibilities and remedial paths. A mature system requires emergency freezing, permission revocation, manual takeover, and complete traceability, retaining stricter human thresholds for high-risk actions. What truly matches AI is not "AI + Card," but "Agent + Policy Controlled Account."

Looking ahead, from yield entry to a robust financial system

Understanding whether a Web3 Neobank is mature requires more than just registered users, card issuance numbers, transaction counts, TVL, and maximum APY. These numbers can prove that the platform is being accessed, but they cannot prove that it has assumed the functions of a financial system. Subsidies can create card issuance, market conditions can inflate TVL, and a hot event can generate transaction volume; only real business that occurs repeatedly can transform product entry into account relationships. More meaningful metrics are: whether real business users continue to stay, how payment success and anomaly rates are, how long it takes to recover after failures, whether customer service and compliance costs are controllable, and how much fund movement truly goes through the platform's policies, approvals, and audits. Funds flowing under policy control are not yet an industry standard, but they pull the evaluation back to one thing: what the platform actually controls and what it is willing to be responsible for.

Robustness does not mean turning Web3 back into a closed bank. It first means clear rights and responsibilities, transparency in fund locations, custody methods, and sources of income, consistent policies across different funding tracks, and having customer service, recovery, and business continuity in the event of failures. Business continuity is not just about servers being online; it also includes partner changes, key recovery, account migration, dispute resolution, and manual downgrades. When automated links are unavailable, enterprises still know how to retrieve funds, restore payments, and complete audits. Users can choose higher-risk or more autonomous account structures, but the sources of risk, control methods, and exit conditions must be visible.

Yield and operating systems are not a simple binary choice. Enterprises can improve the efficiency of some idle funds through stablecoins, RWA, or DeFi, but the Treasury must first address cash visibility, liquidity, permissions, approvals, and audits before enhancing yields. A healthy financial system can accommodate yields but cannot rely on the highest APY to prove its value, nor can it use yields to obscure underlying responsibilities and liquidity gaps. Low yields are not inherently safe, and RWA and DeFi are not the same type of risk. Truly mature products allow every type of fund allocation to enter a framework that is understandable, limited, and exitable.

It can be seen that capital is also voting with amounts, but the pricing received by terminal platforms and underlying infrastructure is not the same. AllScale disclosed $1.5 million in financing in June 2025 and $5 million in seed round in December. Rain, which provides stablecoin payment infrastructure, completed a $250 million Series C round in 2026, with a valuation of $1.95 billion and total financing exceeding $338 million. Rain's official announcement also stated that its annual transaction volume exceeds $3 billion, with over 200 partners. This disparity at least indicates that capital is currently more willing to pay a premium for compliant payment tracks that can be called upon by multiple terminal parties, while terminal Web3 Neobanks are still proving whether user scale can be converted into stable business. Therefore, the endpoint of Web3 Neobanks can be divided into three scenarios: in an optimistic case, a few platforms integrate banks, cards, stablecoins, and on-chain protocols into a trusted cross-track control layer; in a neutral case, most platforms remain in specific regions, specific users, or vertical workflows such as electronic billing, payments, card control, and Treasury; in a pessimistic case, platforms relying on subsidies, vague responsibilities, and single partners will gradually exit after channel interruptions, risk events, or regulatory tightening. The three scenarios correspond to the same maturity standard: it is not about letting users access more financial products, but about allowing these products to enter a sustainable financial order.

Stitching is not the end; assurance is the boundary.

Returning to Bank the Unbanked, financial inclusion has never been just about getting more people to see an account, obtain a card, or enter a yield product for the first time. True inclusion means that the services users receive can be understood, can be used long-term, can be exited when necessary, and can provide explanations and remedies in case of failure. Traditional banks also face issues of high fees, slow service, and exclusion of specific users, making the expanded entry of Web3 still valuable; but this value only transforms from product experience into institutional capability when a trustworthy relationship of responsibility is established behind the entry.

Returning to the question, whether Web3 Neobank is a new financial infrastructure or a pseudo-suture platform does not depend on whether it relies on third parties. Simply aggregating accounts, cards, wallets, and yield products while leaving rule conflicts and failure costs to partners and users is pseudo-suturing; only when it can unify control, record, explain, restore, and remedy can it potentially become new financial infrastructure. Most platforms today are still in the stitching phase, but stitching may also be a necessary starting point for moving towards a unified system across financial tracks. The endpoint of Web3 Neobank is unlikely to be just a more user-friendly wallet or a cheaper card. The real boundary lies in whether the market is willing to entrust funds, rules, and responsibilities to it in the long term.