Ethereum's circulating supply has grown by roughly 950,000 ETH since the Merge in September 2022, reaching approximately 120.7–121.5 million tokens as of April 2026. The network is running a mild annualized inflation rate of roughly 0.23%, reversing the deflationary regime that defined the 18 mon...
"At the current rate of network activity, Ethereum will not be deflationary again. The narrative of 'ultrasound money' has probably died, or would need much higher network activity to come back to life." — CryptoQuant, research note, April 2026
Ethereum's circulating supply has grown by roughly 950,000 ETH since the Merge in September 2022, reaching approximately 120.7–121.5 million tokens as of April 2026. The network is running a mild annualized inflation rate of roughly 0.23%, reversing the deflationary regime that defined the 18 months following The Merge. The "ultrasound money" thesis — coined in 2021 to describe an asset whose supply would shrink as usage expanded — is no longer supported by on-chain data.
The proximate cause is EIP-4844 (Dencun), activated in March 2024. Dencun introduced blob transactions, cutting Layer 2 data-posting costs by over 90%. L2 transaction fees fell to fractions of a cent within days. Mainnet Ethereum, which had been the primary execution venue, became a settlement layer. The base fee collapsed. As of April 7, 2026, average gas on mainnet priced at 0.052 gwei. On April 11, Ethereum's official dashboard displayed a headline reading "0 Gwei." With base fees near zero, the EIP-1559 burn — the mechanism that canceled issuance — effectively switched off.
Issuance did not. Validators continue to receive approximately 1,700 ETH per day, against a daily burn that has fallen to 50–70 ETH in low-demand periods. Net issuance is positive. The burn-versus-issuance gap, combined with ~36–37 million ETH locked in staking (over 30% of supply), produces an asset whose headline float is inflating while its liquid float is shrinking. The economic-value-first question this report examines is whether Ethereum has become a subsidy-supported network in the same sense as Solana and Bitcoin — one where token issuance, not user fees, sustains the validator set.
At the time of The Merge (September 15, 2022), Ethereum's total supply was approximately 120.52 million ETH. Between September 2022 and mid-April 2024 — the 19-month window immediately following the transition to proof-of-stake — the network ran net-deflationary. Burn from EIP-1559 consistently exceeded validator issuance, and total supply fell by roughly 440,000 ETH during the period.
That regime ended on April 14, 2024, roughly one month after the Dencun upgrade went live. From that date, net supply began expanding. By April 2026, total supply stands between 120.7 and 121.5 million ETH, depending on the data provider. That represents an increase of approximately 950,000 ETH relative to the low point.
The annualized inflation rate over the trailing twelve months is approximately 0.23% on an issuance-minus-burn basis, with spikes as high as 0.74% in individual months (September 2024 recorded the highest post-Merge reading). The rate is modest in absolute terms. It is directional in narrative terms: Ethereum is no longer an asset whose supply falls as the network is used.
EIP-4844 introduced a separate data-availability layer — "blobs" — for Layer 2 rollups. Before Dencun, L2s posted transaction data to mainnet as calldata, paying the same gas as any other mainnet transaction. After Dencun, L2s pay a separate blob gas fee, which trades in its own market and does not directly burn ETH from the EIP-1559 mechanism in the same way calldata did.
The impact was immediate and large:
The burn mechanism is a function of gas price multiplied by gas used. When mainnet gas is near zero, even heavy mainnet activity produces negligible burn. During Q1 2026, the daily burn rate frequently dropped into the double digits of ETH per day — against 1,700 ETH per day issued to validators. The gap is structural rather than cyclical.
The foundational framework for this publication treats on-chain fee revenue and token issuance as distinct categories. On-chain fee revenue represents genuine value transfer from users to the network's economic stakeholders. Issuance represents dilution of existing holders to pay the network's security budget. Networks sustained predominantly by issuance are subsidy-dependent. Networks sustained predominantly by fee revenue are self-sustaining.
By that measure, Ethereum's position in April 2026 has shifted materially:
| Metric | Pre-Dencun (Q1 2024) | April 2026 | |--------|----------------------|------------| | Average mainnet gas price | ~30–50 gwei | ~0.05–0.16 gwei | | Daily burn (ETH) | 2,000–3,500 | 50–150 | | Daily issuance (ETH) | ~2,700 | ~1,700 | | Net supply change | Deflationary | +0.23% annualized | | Validator staking yield (gross) | ~3.5–4.0% | ~3.1–3.3% |
The validator yield figures are partly a function of the falling burn as well. Validator rewards in proof-of-stake Ethereum come from three sources: protocol issuance, priority fees (tips), and MEV. Priority fees have fallen alongside base fees. MEV on mainnet has also declined as activity migrated to L2s. The composition of validator income has shifted toward pure issuance — a source that does not represent user willingness to pay.
In absolute dollar terms, Ethereum's mainnet fee revenue for Q1 2026 ran at annualized rates well below the peaks of prior cycles. At current gas levels and transaction volumes, annualized mainnet fee revenue is in the low hundreds of millions of dollars — down from over $2 billion during earlier activity peaks. Solana, by comparison, generates approximately $55 million in annualized base-layer fees against $4.5–5 billion in annual inflation subsidies. Ethereum's ratio is less extreme but is moving in the same direction.
The counterargument to a straightforward "Ethereum is now inflationary" framing is that the tradeable float is smaller than the headline supply suggests. As of April 2026:
Staking lock-up produces a form of synthetic scarcity. If 30% of supply is operationally unavailable and the remaining 70% carries 0.23% annual dilution, the effective inflation on the liquid float is different from the headline figure. Bulls emphasize this. Bears note that exit queues can process stakers in days, not years, and that the synthetic scarcity can reverse if staking yields underperform alternative uses of capital.
A related point: the SEC and CFTC issued a joint interpretive release on March 17, 2026, classifying staking rewards as non-securities across 16 digital commodities, with ETH named explicitly. That regulatory clarity enabled ETH staking ETFs — BlackRock's ETHB structure passes approximately 82% of gross staking rewards to investors, implying net yields of roughly 2.6% before the fund's expense ratio. The ETF channel is likely to pull additional ETH into staked form, further compressing tradeable float.
Dencun did not merely lower fees. It changed which entities capture value from Ethereum usage. Economic activity that previously paid mainnet gas now pays L2 sequencer fees. Layer 2 sequencer revenue is captured by the L2 operators (Base, Arbitrum, Optimism, and others), not by ETH holders through the burn mechanism.
Base, operated by Coinbase, extracts all sequencer revenue and is profitable at current volumes. Optimism operates a superchain architecture that captures portions of fees from approximately 40 dependent L2s but has not yet reached breakeven. Arbitrum runs a similar model with its own fee distribution. The base-layer Ethereum mainnet collects only the residual — blob fees plus a small share of settlement activity — which is a fraction of what calldata posting previously generated.
From an economic-value-distribution perspective, Ethereum has become the reserve settlement layer for a set of L2s that extract most of the end-user fee revenue. That is not inherently bad for the network. It is a different business model from the one implied by the "ultrasound money" thesis. The thesis assumed that network usage would continue to generate mainnet gas pressure sufficient to burn supply faster than issuance added it. The Dencun upgrade broke that assumption deliberately — the stated goal was to cut L2 costs, and it did.
The open question is whether the L2 ecosystem will eventually generate enough settlement-layer activity (larger-value transactions, cross-L2 bridging, institutional flows) to revive mainnet burn. CryptoQuant's April note takes the position that this is unlikely at current network activity levels. The alternative view, held by firms including 21Shares, is that Ethereum's 2026 outlook should be characterized as "staked, slightly inflationary, levered by scalability" — accepting mild inflation as the price of L2 adoption and hoping that long-term settlement-layer demand catches up.
The ultrasound money framing was always conditional on a specific model of Ethereum usage: mainnet-centric, high-gas, fee-driven. Dencun shifted the network to a different model: L2-centric, near-zero mainnet gas, issuance-driven. The April 2026 data confirms the shift is persistent rather than transitory. Supply is expanding. Burn is not keeping pace. Validator income is increasingly composed of pure issuance rather than user fees.
Whether this outcome is problematic depends on the observer's framework. For holders who bought ETH on the deflationary thesis, the data is a reversal. For those who see Ethereum as an L2-anchored settlement layer whose long-term value accrues through the broader ecosystem rather than through mainnet fees, the mild inflation is acceptable collateral damage for the scalability gains.
The economic-value-distribution framework applied in this publication treats the question more strictly. Networks that pay their security budget primarily through issuance are categorized as subsidy-dependent. By that standard, Ethereum in April 2026 is closer to the subsidy-dependent category than it was in April 2024. The absolute subsidy rate remains low relative to Solana or Bitcoin, but the direction of travel is the same. The next meaningful test will be whether sustained L2 adoption eventually pulls enough settlement activity back onto mainnet to restore a fee-driven burn regime. The data through Q1 2026 does not yet show that happening.