Every wallet you have ever used on Ethereum, every dApp you have ever interacted with, every gas estimate your MetaMask has ever shown you — they all rest on a single number: 21,000. That is the minimum amount of gas required to send a basic ETH transfer on Ethereum. It is not a guideline. It is not a soft default. It is hardcoded into the protocol, and for over a decade it has acted as the atomic unit of cost on the network. Ethereum's next upgrade is going to change that, and understanding why requires understanding why that number existed in the first place and what kind of accounts are about to replace the ones you use today.
Gas is Ethereum's pricing mechanism for computation. Every operation the network performs — reading storage, writing data, running arithmetic — costs a certain number of gas units. The user pays for those units in ETH. The 21,000 floor exists because even the simplest transaction — move ETH from address A to address B — requires the network to do real work: verify a signature, check a nonce, debit one account, credit another. That work costs computation, and 21,000 was the protocol designers' estimate of the minimum cost of that work for a standard externally owned account, or EOA. An EOA is the kind of account you have right now: a private key controls it, one signature type is accepted, and the rules are rigid. Sign the transaction, broadcast it, pay at least 21,000 gas. That is the deal.
Smart contract accounts — sometimes called smart accounts — break that model. Instead of a private key with fixed validation logic baked into the protocol, a smart account is itself a piece of code. It can define its own rules for what counts as a valid transaction. It can accept different signature schemes, enable social recovery, allow a third party to pay the gas on your behalf, batch multiple operations into a single transaction, or enforce spending limits. ERC-4337, introduced a couple of years ago, was a first attempt to bring smart accounts to Ethereum without changing the protocol itself. It worked by routing transactions through a separate mempool and a smart contract called an EntryPoint that simulated the role of the protocol. Functional, but clunky. The overhead was real, and the 21,000 floor remained because underneath every ERC-4337 transaction there was still an EOA kicking things off.
What the Glamsterdam upgrade introduces — building on the EIP-7702 foundation that arrived earlier — is native protocol support for smart accounts. EIP-7702 allowed an EOA to temporarily delegate its behavior to a smart contract for the duration of a transaction. The next step goes further: the protocol itself gains the ability to initiate and validate transactions from accounts that have no fixed validation logic, which means the 21,000 minimum no longer applies universally. A smart account transaction can be priced according to what it actually costs to compute, not according to a floor designed for a different account type. For simple smart account operations that are cheaper than an EOA transfer, the cost can be lower. For complex batched operations, the cost reflects the real computation. The number 21,000 stops being a universal floor and becomes one data point among many.
Why does this matter beyond academic protocol design? Because 21,000 gas as a universal minimum has shaped every wallet interface, every dApp integration and every gas estimation library in the ecosystem. The moment that floor is variable, every one of those systems needs to rethink how it prices and estimates transactions. More importantly, smart accounts at the protocol level enable user experience improvements that Ethereum has been promising for years: seedless onboarding, gasless transactions subsidized by applications, one-click multi-step DeFi interactions, and account recovery that does not require a 12-word phrase written on a piece of pa. These are not small quality-of-life tweaks. They are the difference between a technology that remains difficult to use and one that can reach people who will never read a hardware wallet manual.
ETH, the native token, is how all of this gets paid for. Gas fees are denominated in ETH. Validators who secure the network stake ETH and earn ETH rewards for proposing valid blocks. Since EIP-1559 introduced fee burning, a portion of every transaction's base fee is permanently destroyed. That deflationary mechanism interacts with issuance: when network activity is high enough, the ETH burned exceeds the ETH issued to validators, making the supply contractionary. As of this month, 34.4 percent of all ETH supply is staked, a record. That concentration creates genuine liquidity questions — a significant portion of supply is locked and earning yield, which reduces float and raises real concern if validators ever rush for the exit simultaneously.
The risks here are real and layered. Ethereum has no hard supply cap, which is a legitimate philosophical objection for those who want absolute scarcity. The transition to a smart account world introduces a large surface area for bugs: the more expressive accounts become, the more ways there are for validation logic to go wrong, and the consequences of a wallet-layer exploit at scale are severe. Ethereum also faces sustained competitive pressure. Solana has captured meaningful developer mindshare and market share for applications that need high throughput. The L2 ecosystem, while growing, fragments liquidity and user experience in ways the Ethereum Foundation is still working to address. And the market has not rewarded ETH holders the way Bitcoin holders have been rewarded over the past two years — ETH at $1,911 today reflects real uncertainty about where value accrues in a world where most activity lives on rollups.
What Ethereum is doing with this upgrade is not adding a feature. It is changing the primitive. The account model is the foundation on which wallets, dApps and user experience are built. Breaking the 21,000 rule is the protocol's way of saying that the assumptions the ecosystem has operated under for a decade are no longer good enough — and that the infrastructure is ready for accounts that are programmable all the way down. Whether the market catches up to that thesis is a separate question. Whether the thesis is technically sound is not.
