Account Abstraction in 2026: Who Pays for Gasless Crypto

Open a crypto wallet in 2026 and the friction that used to define it is quietly draining away. You can sign in with a fingerprint instead of a seed phrase, send a stablecoin without owning a cent of Ethereum for gas, fold an approval and a trade into a single tap, and hand a game or an AI agent a spending allowance it can draw against without pestering you to sign every move. Account abstraction is the umbrella term for that shift, and it has stopped being a whiteboard idea. Smart accounts built on the ERC-4337 standard have now processed more than 1.24 billion UserOperations across over 64 million active accounts, with paymasters covering north of $13.4 million in gas on their users’ behalf, according to BundleBear.

The catch, and it is a real one, is that none of this is free. Gas still gets paid to validators, keys still guard money, and every bit of convenience pushes a cost somewhere on the board. Someone funds the paymaster that makes a transaction feel gasless. Someone runs the infrastructure that packages your intent into a block. Someone eats the markup when you settle a network fee in USDC. This guide is a tour of account abstraction for the wallets-and-exchanges reader, built around a single question the marketing tends to skip: when the gas goes away, who actually pays?

What account abstraction actually changes

Ethereum has always had two kinds of accounts. Externally owned accounts, or EOAs, are the ones almost everyone uses: a single private key controls the money, that key signs every transaction, and the rules are fixed by the protocol. You pay gas in ETH, you approve one action at a time, and if you lose the key the funds are gone. The second kind, contract accounts, can carry arbitrary logic, but historically could not start a transaction on their own. Account abstraction is the long project of collapsing that distinction so that any account can hold programmable rules about who may spend, under what conditions, in which token, and with what recovery path.

In practice, that means your wallet becomes a small program rather than a bare key pair. A smart account can demand two signatures over a threshold, sponsor its own gas, expire a permission after an hour, rotate the key that controls it without moving funds, and accept a login from an Apple or Android passkey. None of that is possible for a plain EOA at the protocol level. The table below lays out how the three models that matter in 2026 compare.

The rest of this article is about the machinery that makes the middle and right-hand columns work, and about the bills that machinery quietly generates. For a capabilities-first tour of what smart accounts let you do day to day, HOGE Wire covered that separately; here the lens is cost and plumbing.

ERC-4337: the alternate mempool that made it real

The standard that turned account abstraction from theory into a billion transactions is ERC-4337, finalized in March 2023 and deliberately built to need no change to Ethereum’s core protocol. Rather than forcing a hard fork, it introduces a parallel system that sits above the chain. A user expresses what they want as a UserOperation, a richer object than a normal transaction. Specialized actors called bundlers collect these UserOperations from a separate mempool, wrap a batch of them into one on-chain transaction, and submit it to a single audited contract, the EntryPoint, which verifies and executes each one.

Two supporting roles make the system useful. Account factories deploy a user’s smart account on first use, so a wallet can hand out an address before anything touches the chain. Paymasters, the piece this article keeps returning to, can agree to cover a UserOperation’s gas so the user never needs to hold ETH. The EntryPoint has been revised carefully over time; the eth-infinitism team shipped version 0.6, then 0.7, and in 2025 version 0.8, which added native support for the EIP-7702 path discussed below, per the project’s release history.

The scale is no longer trivial. BundleBear’s tracker counts more than 1.24 billion UserOperations and over 64 million accounts with activity, and shows paymasters have sponsored in excess of $13.4 million of gas over the life of the standard. The design’s great virtue, that it needed no permission from Ethereum’s core, is also its central weakness: because UserOperations live in their own mempool and depend on bundlers to reach a block, the whole edifice leans on a layer of intermediaries that ordinary transactions do not require. Hold that thought, because it drives both the economics and the politics later in this piece.

It is worth being precise about where the money goes inside this system, because the bundler is not a charity. A bundler pays the actual gas to include a batch on-chain and needs to come out ahead, so it collects fees from the UserOperations it packages and captures whatever value it can from ordering them. When a paymaster sponsors a user, it is really pre-funding a deposit at the EntryPoint that the bundler later draws against. Every gasless tap therefore passes through at least two commercial parties before it reaches a validator, and each one prices in risk, overhead, and margin. That layering is the origin of the markups examined below, and the reason a transaction that looks free to you is never free to the chain.

EIP-7702: meeting ordinary wallets where they are

ERC-4337 asks a user to move to a new, contract-based account. EIP-7702 takes the opposite tack: it lets the wallet you already have behave like a smart account without abandoning its address. Shipped in Ethereum’s Pectra upgrade on 7 May 2025 and co-authored by Vitalik Buterin, it adds a new transaction type (0x04) that lets an EOA post a small pointer, a delegation designator, telling the network to run a chosen contract’s code when that account transacts. Sign once and your familiar address gains batching, gas sponsorship, and session keys; delegate to the null address and it reverts to a plain EOA.

Adoption came fast precisely because it required no migration. BundleBear’s 7702 tracker records more than 225 million cumulative authorizations, nearly 50 million live delegations, and over 94 million set-code transactions since Pectra, per its dashboard. The same power that lets a wallet upgrade itself in one signature also lets an attacker drain a compromised account in one signature, which is why the early months produced both genuine convenience and a wave of copy-paste sweeper contracts, near-identical bytecode that watches for any incoming funds and forwards them straight to an attacker. Researchers found that the vast majority of the earliest delegations pointed at this reused sweeper code, though most of it never turned a profit, because the wallets it targeted were usually already empty. HOGE Wire’s look at why stolen keys, not broken contracts, drive most modern losses covers that dynamic in depth.

For the wallets-exchanges reader, the practical point is that account abstraction now arrives by two roads at once: a full smart account on ERC-4337, or an upgraded EOA on EIP-7702. Increasingly the two converge, since the latest EntryPoint version speaks 7702 natively and wallet makers route users down whichever path their existing account allows.

Paymasters, or who actually pays for “free”

The single feature that sells account abstraction to normal people is the gasless transaction, and the piece of machinery that delivers it is the paymaster. A paymaster is a contract that agrees to cover a UserOperation’s gas, settling with the EntryPoint so the user never needs ETH. That is the magic trick. The economics live in who funds the paymaster and how they recoup the cost, and there are three models in production that answer the who-pays question very differently.

In a sponsorship model, the application pays. A game, an exchange, or a wallet maker funds the paymaster’s deposit at the EntryPoint and writes the gas off the way any business writes off customer acquisition, per a 2026 breakdown by Eco. The user gets a genuinely free transaction; the company is buying activation. In an ERC-20 model, the user still pays, just not in ETH: the paymaster fronts the ETH, runs the operation, and afterwards pulls the equivalent value in USDC, USDT, or DAI from the account, typically charging a 5 to 15 percent markup to cover price-oracle drift and conversion. The third model, a verifying paymaster, is less a payer than a gatekeeper: an off-chain signer controlled by the app decides, per request, whether to sponsor at all, which is how most production services such as Alchemy’s Gas Manager and Pimlico’s paymaster actually operate. A single real transaction often blends these models: an exchange might run a verifying paymaster that fully sponsors a new user’s first several swaps to drive activation, then switch that same user to an ERC-20 model afterward, so the cost quietly moves from the company’s marketing budget to the user’s USDC balance without either side ever touching ETH.

None of these makes gas disappear. They relocate it, and sometimes add a fee on top. That is the first thing to understand about the smoother wallet: gasless is a billing decision, not a law of physics.

The real price of a gasless transaction

So what does a sponsored transaction actually cost the entity footing the bill? It depends almost entirely on which chain it runs on, and the gap is enormous. Eco’s 2026 estimate puts a single sponsored operation on Ethereum mainnet at roughly $1.10 to $4.50 once bundler and paymaster markup are included, while the same operation on a Layer 2 such as Base, Arbitrum, or Polygon costs a few cents or less.

Multiply that by usage and the strategy writes itself. A Base application with 10,000 monthly active users each doing a dozen sponsored operations faces something like $1,200 to $9,600 a month, per Eco’s figures, a marketing line item a growth team can defend. The same traffic on mainnet would be ruinous. This single fact explains why almost all consumer account abstraction has migrated to Layer 2s: the abstraction is only affordable where the underlying gas is cheap. For a product team, that turns gas into a growth lever: sponsor aggressively on a Layer 2 and you remove the single biggest drop-off point in crypto onboarding, the moment a newcomer is told to go buy ETH before they are allowed to do anything at all. It is also why exchange-grade sponsorship, discussed below, tends to launch on Base and Arbitrum first. Ethereum’s own fee market still sets the ceiling; with ETH trading around $1,917 in mid-August, per Fortune, mainnet gas remains a real cost even when the user never sees it.

Paying for gas in the stablecoin you already hold

The most consumer-legible version of the ERC-20 paymaster is Circle’s, which lets a user pay network fees in USDC on any ERC-4337 account, no ETH required. It launched on Arbitrum and Base and has been expanding outward, and after an introductory waiver Circle applies a surcharge of roughly 10 percent on the networks where it is live, per the company’s own announcement. For a user who thinks in dollars and holds stablecoins rather than gas tokens, that is a fair trade: a small premium to never think about ETH balances again.

The quieter development is standardization. ERC-7677, a paymaster web service capability, defines a common interface so a wallet can ask any compliant paymaster to sponsor a transaction using the same two calls, per its specification. In plain terms, it lets a wallet switch gas providers the way a website switches payment processors, without rewriting code. That matters for the economics: portable paymasters put downward pressure on markups, because an app that dislikes one provider’s pricing can point at another tomorrow. Stablecoins as a gas currency, plus a standard for comparison-shopping paymasters, is how gasless quietly turns into a competitive market rather than a favor.

The plumbing has a few landlords

Every gasless, batched, passkey-signed transaction on ERC-4337 passes through infrastructure that someone has to run, and that layer is more concentrated than the decentralization talk suggests. Roughly eight providers operate production-grade bundler and paymaster services in 2026, among them Alchemy, Pimlico, Biconomy, ZeroDev, Coinbase Developer Platform, Stackup, Candide, and Etherspot, per Eco’s survey of the stack. Two of them, Pimlico and Alchemy, carry the bulk of all UserOperations, and a similarly short list runs the verifying paymasters behind most gasless apps.

That concentration is exactly what the protocol’s designers worry about. Because a smart account depends on bundlers and relays to reach a block, those intermediaries become, in Buterin’s framing, a source of fragility with no guaranteed alternative path to inclusion. He has argued that intermediary minimization is “a core principle of non-ugly cypherpunk Ethereum,” per remarks reported by Cointelegraph, and the fixes now on Ethereum’s roadmap are meant to make the plumbing censorship-resistant even if the commercial relays misbehave. Whether or not you find the cypherpunk aesthetics persuasive, the business point is concrete: the more your gasless experience depends on two companies, the more their uptime and their pricing are your risk.

Ethereum’s answer is to give pending operations a guaranteed route onto the chain that does not pass through any single relay. A fork-choice enforced inclusion mechanism drafts a rotating set of ordinary validators as includers, each able to push transactions the commercial bundlers ignore into the next block, restoring the censorship resistance that a plain transaction has always had. It is a reminder that account abstraction’s convenience layer and Ethereum’s neutrality guarantees are being negotiated at the same time, and that the outcome decides whether the smoother wallet is also a trust-minimized one.

Passkeys, session keys, and permissions you can revoke

Two capabilities do the most to change how a wallet feels, and both are pure account abstraction. The first is passkeys. A smart account can verify signatures from the secp256r1 curve that phones and laptops already use for Face ID, Touch ID, and Windows Hello, which means a user can approve transactions with hardware-backed biometrics instead of a seed phrase. Doing that cheaply on-chain took a dedicated precompile, RIP-7212, which cut the cost of verifying one of these signatures from hundreds of thousands of gas to around 3,450, per Alchemy, and it has since been adopted across the major Layer 2s.

The second is session keys: temporary, tightly scoped keys that can sign a bounded set of actions for a limited time. A game can move your items for an afternoon; a trading interface can rebalance within limits; an AI agent can pay for services up to a cap. The emerging standard for requesting these permissions, ERC-7715, lets an app ask a wallet for a specific, revocable allowance rather than blanket approval, per its draft. These permissions can be shaped to fit the job: a fixed periodic allowance for a subscription, a streaming budget that refills over time, or a one-off cap for a single task, each of them revocable the moment you change your mind. This is the machinery behind the on-chain trading agents and automated strategies now common on perpetual-futures venues, which HOGE Wire examined in its account of the on-chain futures boom.

The convenience is obvious; the discipline is the point. A session key or spend permission is a small, revocable delegation, which is far safer than the unlimited token approvals that drained wallets for years, but only if the wallet makes limits and expiry legible and cancellation easy. Programmable spending is a security upgrade exactly to the degree that a normal person can see and undo it.

Social recovery, and the slow death of the seed phrase

The seed phrase has been crypto’s single worst piece of user experience for a decade: twelve or twenty-four words that grant total control, cannot be reset, and are gone for good if the paper burns or the phrase is phished. A smart account moves the phrase out of the center of the design. Because the account is a program, the key that signs can be rotated, and the right to authorize a rotation can be spread across the people or devices you trust, an approach Vitalik Buterin has championed for years under the name social recovery.

In practice a user nominates a set of guardians, which can be their own hardware wallet, a second phone, a passkey stored in a cloud keychain, or trusted friends and family. If the primary key is lost, a threshold of those guardians, say three of five, can jointly approve a new key, moving control of the account without moving a single coin and without any one guardian ever being able to act alone. Recovery becomes a policy the account enforces rather than a secret the user has to guard perfectly forever.

The trade-off is that recovery is only as trustworthy as the guardians and the rules around them. A cloud-synced passkey is convenient but hands part of your security to a platform; friend-and-family guardians can be socially engineered; a poorly chosen threshold can lock you out or let a colluding minority in. This is also where self-custody begins to blur, because a provider that offers hosted recovery or holds a guardian key is doing something that looks a little like custody, a distinction that matters for the regulatory questions further down this page. Done well, social recovery is the feature most likely to bring the next hundred million users on-chain; done carelessly, it just relocates the single point of failure.

How exchanges rebuilt themselves around smart accounts

For the exchange half of the wallets-exchanges beat, the most consequential account-abstraction story is the embedded wallet: a self-custodial account that lives inside a consumer app and hides crypto’s rough edges entirely. Coinbase moved this from beta to general availability with its CDP Embedded Wallets in October 2025, letting any developer drop in a self-custodial wallet that a user opens with an email, phone number, or social login, with the private keys held inside trusted execution environments that Coinbase says even it cannot access, per the company’s launch note.

Coinbase pairs this with Base Account and a spend-permissions system that lets a user grant an app or an agent the right to pull funds within preset guardrails, published as an open standard on GitHub. The combination previews where consumer crypto is heading: a user who never sees a seed phrase, never buys a gas token, and delegates narrow spending authority to the apps they trust, all on top of a self-custodial smart account. Exchanges also have to adapt on the back end. An EIP-7702 delegated EOA looks like an ordinary address but now carries contract code, so deposit and withdrawal systems have had to learn to read the delegation designator and screen it, a quiet operational shift at custodians and trading venues that most users will never notice.

The broader pattern is the invisible wallet. When a consumer app lets someone hold and move a dollar-denominated stablecoin balance with nothing more than a login and a face scan, the account underneath is very often a smart account with a sponsored paymaster and a passkey signer, even though the word wallet never appears on screen. That is the endgame the exchanges are building toward: crypto rails with the ergonomics of a neobank, where account abstraction does the work and the user is never asked to understand any of it. The prize is onboarding people who would never tolerate a seed phrase; the risk is that the same convenience concentrates a great deal of trust in whoever runs the embedded-wallet stack.

Safe, and the money that moves in size

Consumer wallets get the headlines, but the largest pools of on-chain value sit in a different kind of smart account: the multisignature vault. Safe, the standard for organizational custody, reported roughly 130 million transactions in a single quarter of 2026 and more than $27 billion in self-custodied assets across over 63 million accounts, per figures in its Q2 report covered by The Block. DAOs, foundations, and corporate treasuries run on this infrastructure precisely because a smart account can enforce rules a single key cannot: require several signers, scope what each may approve, and add a timelock before anything moves. On top of that base, tooling such as the Zodiac module system turns a Safe into a programmable governance engine, wiring the outcome of a token vote directly to what the treasury is permitted to execute.

That activity has proven sticky through market weakness. “This activity held up even as crypto prices fell sharply in June,” said Lukas Schor, president of the Safe Ecosystem Foundation, noting that active accounts kept rising through the drawdown, per The Block. For an institution weighing self-custody against a bank or an exchange, the multisig smart account is the workhorse that made programmable custody boring, which in this context is the highest compliment you can pay a piece of financial infrastructure.

The security ledger: blind signing and bad paymasters

Every one of these conveniences is also an attack surface, and the honest version of the account-abstraction story keeps a running security ledger. The dominant failure mode is not broken contracts but blind signing: approving a transaction whose real effect you cannot see. The $1.5 billion Bybit theft in February 2025, the largest crypto hack on record, worked exactly this way, with signers approving what looked like a routine transfer on a compromised interface while the payload swapped out the wallet’s underlying logic, per analysis gathered by The Block. No smart-account feature protects a user who signs a lie.

The trendline is not all grim. Wallet-drainer losses fell 83 percent in 2025 to about $83.85 million, as better warnings and clear-signing standards took hold, per data from Scam Sniffer reported by Cointelegraph. Account abstraction does add its own line items to watch: a malicious paymaster can grief users, an over-broad session key is a standing liability, and the same batching that bundles an approval with a swap can bundle an approval with a theft. The industry’s structural fix is clear signing, a push behind standards such as ERC-7730 to give wallets human-readable descriptions of exactly what a transaction will do, so a hardware device can show a plain sentence instead of a hash. The rest of the defenses are the unglamorous ones.

• Readable transactions, so the wallet shows what you are actually signing rather than a hash.
• Scoped, expiring permissions in place of unlimited token approvals.
• Revocation that a non-technical user can actually find and use.
• A clear view of which paymaster is sponsoring you, and on what terms.

What US regulators and the taxman make of it

Because a smart account is still self-custody, most of it sits outside the parts of US law aimed at intermediaries. In April 2026, staff at the Securities and Exchange Commission indicated that software letting a user transact from a self-hosted wallet does not, by itself, make the provider a broker, per CoinDesk, a meaningful signal for wallet makers who feared being regulated like exchanges. Market-structure legislation still moving through Congress could firm that boundary up, but until something passes, the line between neutral software and a regulated intermediary is drawn case by case. The activities that touch custody, gas sponsorship, or recovery on a user’s behalf are where the harder questions still live, and the broader rulebook is a work in progress; HOGE Wire tracks the slipping legislative timeline in its regulatory countdown.

Taxes are the more immediate concern for most users, and here account abstraction can be a quiet trap. Batching several actions into one transaction does not change what each action is for tax purposes; a swap is still a disposal even when it shares a single transaction with an approval and a bridge. As US brokers move onto Form 1099-DA reporting, the mismatch between what a smart account does in one convenient tap and what the tax code counts as several taxable events becomes real money, a gap HOGE Wire’s 1099-DA filing guide walks through in detail.

The native chapter: when the protocol starts paying attention

The next phase aims to fold all of this into Ethereum itself, removing the dependence on a separate mempool and third-party bundlers. Three proposals compete. The Ethereum Foundation’s camp backs EIP-8141, which Buterin describes as “an omnibus that wraps up and solves every remaining problem that AA was intended to address,” and which he thinks could ship “within a year,” per remarks reported by Cointelegraph. Coinbase and Base back a leaner design, EIP-8130, which the team plans to ship in Base’s Cobalt upgrade and which it says makes a native USDC transfer more than twice as cheap in gas, per Base’s engineering blog. A third, minimal approach from the Paradigm camp keeps the primitives small on purpose. The three differ mostly on how much they let an account do: the Foundation’s design allows almost any validation logic, including the post-quantum signatures a future Ethereum may need; Base’s fixes a menu of supported key types in exchange for lower gas; and the minimal camp deliberately ships the least surface area it can. Whichever wins, the near-term catalyst is Base putting a native design into production, the first large-scale test of account abstraction with no separate mempool underneath it.

None of this is settled, and the people building it say so plainly. “It’s still a very early proposal, so we need to evaluate all the rough edges,” Ethereum core developer Marius van der Wijden said of the native direction, per DL News. The prize, if it lands, is an account-abstraction stack with no external bundlers to depend on and no side mempool to trust, which would take the plumbing landlords out of the critical path and, not incidentally, put competitive pressure on the paymaster markups this article has been tracking. For now the pragmatic reality is the one users already have: a smoother wallet, a small economy paying for the smoothness, and a set of choices about how much of that cost, and that trust, you are comfortable handing off.

Frequently Asked Questions

Does account abstraction make crypto transactions actually free?

No. A gasless transaction still costs gas; a paymaster pays it and recovers the cost. In a sponsorship model the app absorbs the fee as a customer-acquisition cost, and in an ERC-20 model you pay in a token such as USDC, usually with a markup of roughly 5 to 15 percent. Gas does not disappear, it moves and is sometimes marked up.

What is the difference between ERC-4337 and EIP-7702?

ERC-4337 is a full smart-account standard that runs above Ethereum’s core protocol using a separate mempool, bundlers, and an EntryPoint contract. EIP-7702, live since the Pectra upgrade in May 2025, instead lets your existing account temporarily run smart-account code without changing its address. In 2026 they increasingly work together, because the latest EntryPoint version supports the 7702 path natively.

Can I pay Ethereum gas fees with USDC or another stablecoin?

Yes, on any wallet that supports an ERC-20 paymaster. Circle’s Paymaster, for example, lets you pay fees in USDC on networks such as Arbitrum and Base without holding ETH, adding a surcharge of about 10 percent where it is live. The paymaster fronts the ETH and deducts the equivalent value from your stablecoin balance.

Are smart-account wallets safe?

They are as safe as how you use them. Account abstraction enables social recovery, spending limits, and revocable session keys that lower risk, but the same power lets an attacker act in a single signature if you approve something you cannot read. The largest 2026 losses came from blind signing rather than broken code, so readable transactions and easy revocation matter more than any feature list.

Does using a smart account change my crypto taxes?

Not in the way many users hope. Batching several actions into one transaction does not merge them for tax purposes; each swap, sale, or bridge is still assessed on its own. In the United States, as brokers move to Form 1099-DA reporting, a single convenient tap can still create several taxable events, so keeping records at the action level remains essential.

By Yuki Tanaka, senior wallets and exchanges correspondent at HOGE Wire.