Types of Digital Wallets: Closed, Open, Crypto, and MPC

The types of digital wallets are usually presented as one list, and that's why no two lists agree. Wallets get sorted along four independent axes: the loop they can spend in, the form they take on a device, who holds the credential, and what asset sits inside. Alipay and a hardware crypto signer both qualify as digital wallets, and almost nothing about their architecture is shared.

A digital wallet is software that holds a payment credential and presents it for you, so what reaches the merchant is never the card number or the private key. Everything below follows from what that credential is and who controls it.

The short version

• Four axes classify the digital wallet types in use today: by loop (closed, semi-closed, open), by form (mobile and NFC, web or e-wallet, super-app, hardware), by custody (custodial or non-custodial), and by asset (fiat pass-through, stored value, crypto, tokenised deposits, CBDC).
• The loop axis is the one with legal teeth. UK and EU rules define electronic money as value accepted by someone other than the issuer, and value confined to a limited network is explicitly excluded, so a closed loop wallet sits outside the licensing perimeter and an open one does not.
• Digital wallets took 56% of global e-commerce transaction value and 33% of in-store value in 2025, per Worldpay's Global Payments Report 2026, a survey of more than 63,000 consumers across 42 markets.
• Card wallets hold no money. Apple Pay stores a Device Account Number in the Secure Element and Apple says it doesn't keep your original card number, so the wallet is a tokenisation layer over an account someone else maintains.
• Crypto wallets are the branch where the credential is a key rather than a token, which removes the issuer and therefore removes recovery. Hot and cold, custodial and non-custodial, hardware, MPC, multisig and smart-contract wallets all live there.
• Choosing a type is a custody and licensing decision before it is a product decision.

What is a digital wallet

A digital wallet is an application that stores a payment credential and uses it on the holder's behalf. That definition is deliberately narrow, and it's the one that keeps the taxonomy honest, because it separates three things that get muddled: the interface, the credential, and the money.

The interface is the app. The credential is what the app holds: a tokenised card reference, a balance the issuer owes you, or a private key. The money sits somewhere else entirely, in a bank account, in an issuer's safeguarded float, or on a blockchain ledger. Almost every argument about wallet types is really an argument about the middle layer.

Regulators define the money part precisely. Under the UK Electronic Money Regulations 2011, which transpose the EU E-Money Directive, electronic money means "electronically (including magnetically) stored monetary value as represented by a claim on the electronic money issuer" that is issued on receipt of funds and "is accepted by a person other than the electronic money issuer". Three clauses, and the third one draws the map for the whole category.

Notice what that definition excludes. A wallet that only presents a tokenised card isn't issuing e-money, because no monetary value is stored and no claim on the wallet provider exists. Apple Pay and Google Wallet aren't e-money issuers when they tokenise your bank's card. PayPal, holding a balance you topped up, is a different animal entirely, and it is regulated as one.

How many types of digital wallets are there

There is no single number, and anyone who gives you one has picked an axis without telling you. Run the four axes properly and you get three loop types, four form types, two custody types and five asset types. Flattening those into a single list produces the ten to fifteen types you see elsewhere, at the cost of double counting: a "mobile wallet" and a "crypto wallet" aren't alternatives, since a mobile crypto wallet is both.

Counts differ because the classifier differs.

• Regulators classify by what you owe the customer. India's central bank runs the most explicit scheme. Its Master Directions on Prepaid Payment Instruments, dated 27 August 2021 and updated 27 December 2024, define Closed System PPIs as instruments "issued by an entity for facilitating the purchase of goods and services from that entity only", then split everything else into Small PPIs and Full-KYC PPIs by identity verification rather than by loop. The older semi-closed and open naming was retired in favour of a KYC axis.
• Payment schemes classify by rail. Card networks care whether a wallet passes a token over their rails or takes funds first and pays out separately, because the two carry different fraud liability.
• Product teams classify by form factor, which is why marketing lists open with "mobile wallet" and "desktop wallet". That axis is the least useful of the four for architecture and the most visible to users.

So the question splits in two. If you mean how many types, the honest answer is four axes rather than one number. If you mean how many wallet products exist worldwide, no public authority counts them, and every figure in circulation comes from vendor research rather than from the ECB, the Federal Reserve or the World Bank.

The types of digital wallets at a glance

Each axis answers a different design question, so read the table as four decisions you make in sequence rather than a menu of fifteen products. The digital wallet types people name in a list are combinations of these rows.

A real product is one cell from each row. Apple Pay is open loop, mobile, custodial in the sense that the issuing bank still owns the account, fiat pass-through, consumer. A Starbucks-style app is closed loop, mobile, stored value, consumer. Safe is open loop, web, non-custodial, crypto, and increasingly B2B. Nothing in the taxonomy stops those combining.

Types of digital wallets by loop: closed, semi-closed and open

Loop describes where the balance can be spent, and it's the axis that decides whether you are running a product or a regulated financial institution.

Closed loop wallets spend only with the issuer. Retail apps with a stored balance, transit cards, arcade credits, campus cards. The RBI definition quoted above is the crisp one: purchases from that entity only, and no cash withdrawal. Commercially these are prepayment and loyalty instruments dressed as wallets, and they're popular precisely because the Electronic Money Regulations exclude "monetary value stored on instruments that can be used to acquire goods or services only in or on the electronic money issuer's premises".

Semi-closed wallets spend across a defined set of merchants that have a contract with the issuer, but can't be cashed out. This is the shape of a marketplace wallet, a delivery platform's credit, or a fuel card. The same regulation carves out instruments usable "under a commercial agreement with the electronic money issuer, either within a limited network of service providers or for a limited range of goods or services", which is the limited network exclusion that most semi-closed wallets are built to sit inside.

That exclusion is narrower than teams assume, and it's the most common place a wallet roadmap collides with a licence. One merchant outside the network, or one cash-out path, can move the product across the line.

Open loop wallets spend anywhere the underlying scheme is accepted, because they ride card or bank rails rather than a private ledger. Apple Pay, Google Wallet, a bank's own wallet, any wallet issuing a scheme-branded virtual card. Here you either partner with a licensed issuer or hold the licence yourself.

In practice the difference is cost, not paperwork. Closed loop is a database column. Open loop is an issuing partner, safeguarding accounts, scheme sponsorship, reconciliation, dispute handling and an audit trail that survives inspection. Our guide to payment gateway integration covers the acceptance side of the same problem, and banking as a service covers the shortcut most teams take instead of applying for their own authorisation.

Types of digital wallets by form: mobile, web, super-app and hardware

Form is what the user meets. It decides how the credential reaches the terminal, which is a narrower question than it sounds.

Mobile and NFC wallets hold a tokenised card in the device's secure hardware and present it over near-field communication at a contactless terminal. Apple Pay and Google Wallet are the reference implementations. Apple's own documentation is unusually direct about the mechanics: the Device Account Number "can't be decrypted by Apple but is stored in the Secure Element", and "Apple doesn't store the original credit, debit, or prepaid card numbers that you use with Apple Pay". The wallet holds a reference, not an account.

Web wallets and e-wallets live in a browser or an app and usually carry a balance. PayPal is the archetype. The distinction between the types of e wallet that matter is whether the balance is real: a wallet that holds funds is issuing e-money and needs the licence, while a wallet that stores a card on file and charges it at checkout is a credential vault with a checkout button.

Super-app wallets put payments inside a product people already open for another reason. Alipay and WeChat Pay are the largest examples, and they invert the usual order: the wallet is a feature of a messaging or commerce app rather than a standalone product. This is why market-share questions have regional answers. Worldpay's Global Payments Report 2026 names Alipay alongside Apple Pay and PayPal as the wallets carrying global volume, and the mix shifts market by market across the 42 it covers.

Hardware wallets keep the credential on a dedicated device. In payments that means a card or a wearable with a secure element. In crypto it means a signing device, and that branch gets its own section below because the recovery model is completely different.

One form resists the category: the pass wallet. Apple Wallet and Google Wallet both store boarding passes, tickets, loyalty cards and, in some places, identity documents. Those are credentials rather than payment instruments, so they follow identity rules, and a team shipping passes shares almost no compliance work with a team shipping payments.

Types of digital wallets by custody, asset and user

Three shorter axes, each of which changes the architecture more than the form factor does.

Custodial and non-custodial digital wallets

Custody answers one question: when access is lost, can anyone restore it? In a custodial wallet a third party holds the credential and acts on your instruction, so a support team can verify identity and put you back in. In a non-custodial wallet the holder controls the credential outright, which means nobody can freeze the funds and nobody can recover them either.

Card wallets are effectively custodial even when they feel otherwise, because the account behind the token belongs to a bank that can reissue it. Crypto wallets are where the choice is genuine, and where getting it wrong is permanent.

Digital wallets by the asset inside

• Fiat pass-through. No balance. The wallet presents a token for an account held elsewhere. Lowest regulatory weight, least product control.
• Stored value: a balance the issuer owes the customer. That is e-money, and it brings safeguarding, capital and reporting with it.
• Crypto: a key that authorises a transfer on a public ledger. There is no issuer to reverse anything and nobody to call.
• Tokenised deposits. Bank deposits represented on a programmable platform. The BIS, in chapter three of its 2023 Annual Economic Report, describes tokenisation as "the process of representing claims digitally on a programmable platform" and sets out a unified ledger combining central bank money, tokenised deposits and other tokenised claims. Still mostly pilots.
• Central bank digital currency. A direct claim on the central bank. The ECB's digital euro moved to its next phase in October 2025, with a pilot planned for 2027 and potential first issuance during 2029, assuming the legislation lands. Design your wallet so a new asset type is a new adapter, not a new product.

Consumer, merchant and B2B wallets

Most product teams skip the user axis, and it changes the data model underneath them. A consumer wallet optimises for a balance and a transaction list. A merchant wallet optimises for settlement timing, payouts, refunds and reconciliation against orders. A B2B wallet optimises for entitlements, approvals and multi-user access, which looks far more like a treasury tool than a payment app. Building the consumer version and bolting on merchant features later is one of the more expensive mistakes in this category, because the ledger assumptions differ from the first commit.

How many people use digital wallets: the measured numbers

Wallets stopped being a challenger category some time ago, and the measured share depends heavily on channel and geography.

Worldpay's Global Payments Report 2026, built on a survey of more than 63,000 consumers across 42 markets, puts digital wallets at 56% of global e-commerce transaction value and 33% of in-person spending in 2025. The same report projects payment apps to carry 46% of global point-of-sale value by 2030. The United States sits below the global in-store figure at 17%, against 40% online, which is a useful corrective to any roadmap built on American intuitions about contactless.

Europe reads differently again. The ECB's SPACE 2024 study, published on 19 December 2024, found mobile apps at 7% of euro-area point-of-sale transaction value, behind cards at 45% and cash at 39%. Online, e-payment solutions covering payment wallets and mobile apps reached 29% of transactions against 48% for cards. Cash still accounted for 52% of point-of-sale transactions by number.

In the United States, the 2025 Diary of Consumer Payment Choice, covering 2024 and published in May 2025, reports consumers making an average of 11 payments a month with a mobile phone, up from four in 2018. The generational split is sharper than the average suggests: adults aged 18 to 24 used a phone for 45% of all payments.

Those three sources measure different things, which is why they should never be averaged. Worldpay counts transaction value, the ECB counts euro-area behaviour by number and by value, the Federal Reserve counts payments per person. Read together: wallets have won online, are winning in store at very different speeds by region, and are moving fastest with people under 25.

How digital wallets work: tokenisation, NFC and settlement

A card wallet works by never handling the card. That single design decision explains most of what follows.

When a card is added, the wallet asks the network to issue a payment token in place of the real number. EMVCo, the standards body behind the specification, describes the token as "a unique alternative value" that replaces the primary account number and is "constrained in how it can be used", for instance to a specific merchant, device or payment scenario. The current specification is the EMV Payment Tokenisation Specification Technical Framework, version 2.4, published on 9 July 2026. A token stolen from one device is worthless on another.

Where does the token live? In hardware. On Apple devices that's the Secure Element holding a Device Account Number; on Android the equivalent protection sits behind the platform's own key storage. At the terminal the device doesn't transmit the token alone: it produces a one-time cryptogram bound to that transaction, which is what makes a replayed capture useless. Contactless acceptance runs on the EMV kernels the terminal already implements, which is why wallet payments worked at existing terminals with no merchant upgrade. Settlement then follows the ordinary card path: the acquirer sees a token, the network de-tokenises it, the issuer authorises against the real account, and the wallet touched none of the money.

E-money wallets work differently, and that difference is where the compliance work lives. Funds arrive, the issuer records a claim, and the balance moves on the issuer's own ledger until it leaves for a bank account or a card scheme. That ledger is the regulated object. It needs safeguarding, reconciliation and an identity layer, which is where KYC software stops being a checkbox and starts being infrastructure, and where a digital banking platform and a wallet begin to look like the same build.

Crypto wallets skip all of it. No issuer authorises anything: the wallet signs a transaction with a private key and the network settles it. There is no token to de-tokenise, and no chargeback waiting at the end of it.

Types of crypto wallets inside the digital wallet family

The types of crypto wallets sit inside this same taxonomy, in the branch where the credential is a key rather than a reference to somebody else's account. A crypto wallet stores no coins. It stores, or controls access to, the private key that authorises spending a balance recorded on-chain. Every distinction below is a distinction about that key: how exposed it is, who may use it, and what happens when it is lost.

Two axes do most of the work. Connectivity asks whether the signing key is ever present on an internet-connected device, which is the hot and cold split. Control asks who holds the key, which is the custodial and non-custodial split. They're independent. A custodial wallet can keep its keys cold, and a non-custodial wallet can be extremely hot.

That ordering is architectural, and it is a ranking rather than a score on purpose. Nobody publishes an exposure rating for wallet forms, and a rating would mislead anyway, because the ladder covers one failure mode: whether an attacker can lift the key off the device. The other way funds leave, a transaction you sign without reading, ignores the ladder entirely.

Hot wallets keep the key on a networked device. A phone app, a browser extension such as MetaMask or Rabby, a Solana wallet like Phantom, or a key in a server's memory. Signing is instant, which is what trading, DeFi and any human-in-the-loop flow needs. The cost is that the key lives in an environment running other people's code.

Cold wallets keep the key on a device that never goes online. Transactions are built on a connected machine, signed offline, and only the signature crosses back. Correct for reserves, wrong for a withdrawal queue.

Hardware wallets are the consumer form of cold storage. Ledger's documentation describes a Secure Element that generates and stores private keys inside a chip certified to Common Criteria EAL5+ or EAL6+ depending on the model, the same class of chip used in passports and payment cards. The connected computer can be fully compromised and still can't extract that key. What it can do is present a malicious transaction for you to approve, which is why the on-device screen exists and why you verify the destination there rather than in the browser.

MPC wallets remove the single key entirely. Multi-party computation, usually shortened to MPC, is the cryptography behind them. The key material is split into shares held by different parties or devices, and signing is a joint computation across those shares, so a complete key is never assembled. The underlying cryptography is published and peer-reviewed: Canetti, Makriyannis and Peled's threshold ECDSA protocol realises a threshold signature functionality with a periodic refresh mechanism giving "full proactive security", which is what lets a compromised share be rotated without moving funds to a new address. Custodians and exchange operators reach for MPC because it gives shared control without on-chain gas or contract risk, and because policies over thresholds and approvers are enforced cryptographically rather than by an admin panel.

Multisig wallets solve the same shared-control problem on-chain. Safe is the reference implementation: a smart account where you "define a list of owner accounts and a threshold number of accounts required to confirm a transaction", enforced by the contract itself. Approvals are public and auditable, which suits DAO and shared treasuries, and they cost gas, which suits nothing about high-frequency operations. Many teams run both, multisig for governance and MPC for throughput.

Smart-contract wallets turn the account into programmable code. On Ethereum, ERC-4337 reached Final status and standardised account abstraction through an alternative mempool without any consensus-layer change, which is what made it deployable. It enables social recovery through guardians, spending limits, gas paid in tokens or sponsored by an app, batched approvals, and scoped session keys. EIP-7702, also Final and this time a core change, lets an ordinary externally owned account set code via a persistent delegation and gain batching, sponsorship and privilege de-escalation without migrating to a new address, which takes the address migration out of the upgrade path.

None of this is theoretical. Chainalysis attributed $713 million of stolen crypto value in 2025 to compromised personal wallets across roughly 158,000 incidents and at least 80,000 identified victims, 20% of everything stolen that year, in a report published on 18 December 2025. Down from 44% in 2024, and still the reason recovery-capable wallet types keep gaining ground.

Read down a column and the choice makes itself. If losing access has to be survivable, the first column is irrelevant and the second decides everything. If no single person may move funds, the third rules out software and hardware wallets on their own. And no column in that table stops a user approving a malicious transaction, which is why clear signing and transaction simulation belong in the design alongside the custody model. We go deeper on the build in crypto wallet development and on the threat model in blockchain security.

A wallet's category is two answers, not one: how exposed is the key, and who is allowed to sign with it.

Which type of digital wallet fits which need

Start from the constraint that cannot move, which is almost never the feature list.

For an individual:

For a business:

On the safest wallet question, the honest answer is that safety is a property of the type. A tokenised mobile wallet is safer than the plastic card it replaces because the token is constrained to a device and each transaction carries a fresh cryptogram. A hardware wallet is the safest consumer crypto option because the key is generated in certified silicon and never leaves. An organisation is safest under shared control, MPC or multisig, because no single compromise is sufficient. None of them protects a user who approves a transaction without reading it, and treating a brand name as an answer to this question is how people get robbed politely.

Building a digital wallet: licensing, cost drivers and build paths

The commercial decision isn't which wallet type to build. It's which regulatory perimeter you are willing to stand inside.

Three paths, in ascending order of weight:

• Stay outside. A closed loop wallet spendable only with you, or a pass-through wallet that tokenises cards without holding funds. No e-money licence, fastest path to market, least control over economics.
• Partner. Use a licensed e-money institution or a banking-as-a-service provider as the regulated entity while you own the product. Faster than authorisation, and you inherit their limits, their pricing and their risk appetite. Most wallets launched in the last five years took this route. Fair warning: founders consistently underestimate how much of the roadmap the partner ends up owning, because every new corridor or product needs their sign-off before it needs your engineers.
• Hold the licence. Apply for EMI authorisation yourself, or for crypto-asset service provider status where you custody digital assets. In the EU, MiCA entered into force in June 2023 and applied fully from December 2024, with national grandfathering running to 1 July 2026, so the transitional window most crypto businesses relied on has closed.

Cost tracks a short list of drivers, with the loop and custody choices at the top. Licensing and safeguarding is the largest single line once you hold funds. Ledger correctness comes next, because a wallet ledger that can't be reconciled is a liability rather than a feature. Then identity and screening, key management, rails integration per corridor, and the operational surface of disputes and support. Feature count barely registers.

We build this layer. Kanso is a multi-currency crypto wallet for web and mobile built design-first, with real-time portfolio valuation as the home screen and multi-factor authentication plus encryption underneath. Swissy is a non-custodial wallet on iOS and Android that generates the private key inside the phone's hardware secure enclave and replaces the seed phrase with social recovery. It reached 89% download-to-active-wallet onboarding completion, with 67% of users configuring guardians within thirty days. BitHolder folds buying, storage and transfers into one dashboard for first-time buyers. Three wallets, three different answers on the custody axis.

For the wider stack, web3 development sets the context, crypto exchange features and revenue models covers the venue side of custody, and DeFi as a trend covers what non-custodial wallets connect to. The engineering sits with our blockchain development practice.

The types of digital wallets aren't competing products. They're positions on four axes, and a real wallet occupies one cell on each. Decide the loop first, because it sets your licence. Decide custody second, because it sets your recovery story and your liability. Form factor and feature list are downstream of both, and they're the parts you can change later.