Ethereum has no 21-million cap. It has a story.
That story is a sequence of monetary-policy decisions: a 42-day token sale in 2014, a decade of proof-of-work block rewards, a protocol change in 2021 that started burning fees, a hard cutover to proof-of-stake in 2022, the opening of staking withdrawals in 2023, and a 2024 upgrade that quietly flipped net issuance from durably negative to mildly positive. If you have ever squinted at ultrasound.money and wondered whether ETH is "deflationary," the honest answer is: it depends on which week you look, and why. This piece — Part 1 of our Ethereum Deep Dive — walks the arc so the dashboard starts telling you something real.
Before jumping in, one piece of shared vocabulary: ETH is the native coin of the Ethereum chain, not a token issued on top of it. If that distinction is new, our primer on coins vs tokens: ETH is the native coin of its chain lays out the difference in a page.
ICO, genesis, and the proof-of-work era
Ethereum's supply started with a sale, not a mine. The ether sale ran 42 days, from July 22 to September 2, 2014, priced in bitcoin: 2,000 ETH per BTC for the first two weeks, declining linearly to 1,337 ETH per BTC by the final days. When the Frontier network went live on July 30, 2015, the genesis block recorded 72,009,990.5 ETH — roughly 60 million distributed to presale buyers and about 12 million allocated between the Ethereum Foundation and early contributors.
From that genesis until September 2022, new ETH came from proof-of-work (PoW) — the same consensus family Bitcoin uses, where miners race to solve a hash puzzle and the winner collects a block reward. If you want the mechanics, our explainer on how proof-of-work consensus actually secures a chain covers the cryptography. For our purposes, the number that matters is the issuance rate: roughly 4.61% annualized, or about 13,000 ETH per day between block rewards and uncle rewards (small bonuses for valid blocks that lost the race). By the end of the PoW era, Etherscan's supply accounting shows 47.2 million ETH added from block rewards and another 3.14 million from uncles, stacked on top of the original 72 million.
An analogy helps: think of Ethereum's early monetary policy as a tap on a reservoir. The tap ran at a steady drip for seven years. How much ETH existed at any moment was just genesis plus whatever the tap had added. There was no drain yet.
EIP-1559: adding a drain in 2021
The drain arrived with the London hard fork, which activated at block 12,965,000 on August 5, 2021. London bundled five upgrades; the headline was EIP-1559, a redesign of how Ethereum prices its gas.
Before London, every transaction ran a blind first-price auction — you guessed a gas price, and either it cleared or it didn't. EIP-1559 replaced that with a two-part fee: a dynamically adjusting base fee that every transaction is required to pay, plus an optional priority tip that goes to whoever builds the block. The base fee itself is burned — destroyed by the protocol, permanently removed from circulation. Miners (and later validators) keep only the tip.
This is the first durable sink in ETH's monetary history. Every block of mainnet activity now subtracts ETH from supply in proportion to demand. Cumulatively, about 4.62 million ETH has been burned since London — enough to make "net issuance" a number worth tracking rather than "gross issuance" alone. For a wider look at how burns sit alongside halvings and unlocks across the industry, see our field guide to supply events — halvings, burns, and unlocks.
One parallel worth noting: Bitcoin reaches scarcity by cutting issuance on a fixed schedule, which we walk through in Bitcoin's halving schedule and hard 21M cap. Ethereum chose a different path — demand-driven scarcity, where the harder the chain is used, the more ETH disappears.
The Beacon Chain and The Merge
While mainnet was still running PoW and burning base fees, a second Ethereum was quietly booting up beside it. The Beacon Chain went live at its own genesis on December 1, 2020 at 12:00:23 UTC, after the deposit contract crossed 524,288 ETH from 16,384 validators. For about 22 months it ran in parallel, handling proof-of-stake (PoS) consensus but processing no user transactions — essentially a staged consensus layer waiting to be bolted on.
The bolt-on happened at The Merge: execution block 15,537,393, September 15, 2022. In a single upgrade, proof-of-work mining was retired and the Beacon Chain took over block production. Two numbers describe what changed.
First, issuance. Annualized ETH creation fell from ~4.61% under PoW to ~0.52% under PoS — about 88.7% less ETH minted per year. Daily issuance dropped from roughly 13,000 ETH per day (miner rewards and uncles) to roughly 1,700 ETH per day (validator rewards, at the time of the measurement). Second, energy: estimated electricity consumption fell by approximately 99.95%, because validators secure the chain by posting capital rather than burning kilowatts.
One subtlety: validator issuance scales with the square root of total staked ETH. More validators means less ETH each — so the 0.52% figure is a snapshot, not a fixed rate. The more ETH gets staked, the lower the per-validator yield, and the lower the gross issuance.
After The Merge, Ethereum's monetary picture looked like this: a small steady drip from validators, offset by a demand-driven burn. For the rest of 2022 and into 2023, active weeks were net deflationary.
Shapella and the post-Merge normal
The Merge retired mining, but it did not let stakers withdraw. ETH posted to the deposit contract since December 2020 was locked until the protocol shipped withdrawal logic. That upgrade was Shapella — Shanghai on the execution layer plus Capella on the consensus layer — which activated on April 12, 2023 at 22:27:35 UTC (epoch 194048, block 17,034,870).
Shapella enabled two kinds of withdrawals: partial (automatic sweeping of rewards above the 32-ETH validator balance) and full (exiting a validator entirely and reclaiming the 32-ETH principal). For the first time, staking was round-trippable. "Finality" took on a second meaning: not just transaction finality (~12.8 minutes after inclusion under PoS), but the ability to actually liquidate a staking position back to spendable ETH.
The practical consequence was predictable. Staking became the default passive use for ETH. By early 2026, more than 30 million ETH — closer to a third of total supply — sits staked, with the Beacon Chain issuing new ETH to those validators at an ongoing rate that scales with the staked pool. That flow is the floor of new supply creation.
Post-Dencun: net issuance oscillates around zero
The most recent chapter is the one people misread most often. The Dencun upgrade activated at epoch 269568, block 19,426,587, on March 13, 2024 at 13:55 UTC. Its centrepiece was EIP-4844, also called proto-danksharding, which introduced a new kind of short-lived data object called a blob. Blobs have their own fee market, separate from regular gas, and expire from node storage after about 18 days.
Layer-2 rollups — the Optimisms and Arbitrums that post batched transactions back to Ethereum — were the target audience. Before Dencun, rollups paid L1 gas prices to publish their data. After Dencun, they switched to blobs. L2 fees collapsed roughly an order of magnitude, which is good news for users — and is also what DeFi on L2s runs on, a dynamic our primer on what DeFi actually is and why L2s matter for the burn unpacks further.
The side effect showed up on the supply chart. Blob fees don't contribute to the mainnet base-fee burn in anywhere near the same way L1 calldata did. With rollups moving to blobs, L1 gas demand — and therefore the burn — dropped sharply. Post-Dencun, ETH has been structurally closer to mildly inflationary than deflationary on a 30-day rolling basis, with supply growth oscillating in the 0.0%–0.5% per year band. Quiet weeks see positive net issuance; heavy-activity weeks still occasionally print negative.
Current supply, early 2026: roughly 120.7 million ETH per Etherscan (72.0M genesis + 47.2M PoW rewards + 3.14M uncles + 2.94M PoS staking rewards minus 4.62M burned), or closer to 121.6 million on ultrasound.money's broader accounting that includes Beacon Chain balances. Either way, still inside the narrow band ETH has hovered in since The Merge.
The takeaway for anyone reading supply dashboards: "ultrasound money" is a design property, not a guarantee. It's what happens when burn exceeds issuance. Whether that's true in any given month depends on activity demand — and since Dencun moved the bulk of that demand onto blobs, the net-zero oscillation is the new normal, not the old one.
Key takeaways for ETH holders
- There is no fixed cap. ETH supply is the output of a policy — validator issuance minus base-fee burn — not a preset schedule.
- The Merge did the heavy lifting. A ~88.7% cut to annualized issuance and a ~99.95% drop in energy use happened in one upgrade in September 2022.
- Shapella made staking liquid. Roughly a third of ETH is now staked; validator issuance is the floor of new supply and scales with how much is staked.
- Post-Dencun, net issuance hovers near zero. Supply growth sits in the 0.0%–0.5% band because L2 blob data doesn't burn L1 gas the way calldata did.
- The dashboards are honest, but context-dependent. A "deflationary" print on ultrasound.money reflects last week's activity, not a structural guarantee.
With the supply side mapped, Part 2 of this series turns to how Ethereum actually tracks state — the account model, the EVM, and why "smart contract" is a more specific thing than the headlines suggest.
