Verify one number before you read anything else.
Over a 282-day window ending in early October 2026, the Ethereum network issued 796,623.377 ETH to its validators. In that same window, it destroyed 16,524.553 ETH through execution-layer base fees and blob fees combined. Divide the second number by the first. You get 2.07%.
That is the entire story. Everything else in this article is just explaining why that ratio exists, why it is not a bug, why it is not temporary, and why a large cohort of holders is going to be very unhappy when they finish the math.
For five years, the dominant Ethereum thesis was supply scarcity. EIP-1559 introduced a burn. The Merge cut issuance by roughly 88%. The community coined a term โ ultrasound money โ and treated it as settled physics. Burn more than you mint, and the asset becomes deflationary. Simple. Elegant. Verifiable on any block explorer.
Except the burn was never a constant. It was a function of demand for block space. And in 2026, demand for Ethereum Layer 1 block space is not what the thesis assumed. It is a rounding error against issuance. The gap is not 10% or 30%. It is a factor of roughly forty-eight. If you hold ETH because you believe it is becoming scarcer every day, the ledger has been quietly telling you a different story for most of the past year.
I have watched this movie before. In 2022, I exited UST forty-eight hours before the mint mechanism ate itself, because I sat down and modeled the seigniorage loop instead of reading the tweets about it. The lesson was not that I am clever. The lesson was that when a system's survival depends on a demand assumption nobody is stress-testing, the demand assumption eventually gets tested. Ethereum's burn mechanism is a demand-dependent scarcity machine. Nobody wants to say that out loud. The data says it anyway.
Context: How a Burn Becomes a Rounding Error
To understand the 2.07%, you have to understand what Ethereum actually does when it processes a transaction.
Since EIP-1559 activated in August 2021, every transaction pays a base fee. That base fee is burned โ destroyed, removed from supply, sent to an address nobody controls. On top of the base fee, users can attach a priority fee, which goes to the block producer, the validator who includes the transaction. Only the base fee burns. The priority fee is compensation, not destruction.
Separately, since EIP-4844, there is a second market: blob space. Blobs are chunks of data that Layer 2 rollups post to Layer 1 for availability and settlement. Blobs have their own fee market, and blob fees are also burned.
So the destruction side of Ethereum's ledger has two real components: execution-layer base fees and blob fees. There is also a tiny third category โ SELFDESTRUCT and similar opcodes โ which is negligible. There is a fourth category that is not destruction at all: slashing. When validators misbehave, they lose stake. That ETH is not burned in the fee sense; it is penalized. The distinction matters enormously, and I will come back to it.
The issuance side is simpler. Proof of stake pays validators a reward for proposing and attesting to blocks. That reward is newly minted ETH. There is no hard cap. The issuance rate scales with the total amount of ETH staked, because more validators means more rewards distributed across a larger base.
Here is the structural problem. Issuance is a function of how much ETH is staked. Destruction is a function of how much people are willing to pay for block space. Those two functions are not linked. They can move in opposite directions. And in 2026, they have.
Staking participation sits at roughly 43,754,000 ETH across 853,325 active validators. Total supply is approximately 122,116,000 ETH. So about 35.8% of all ETH is staked. That is a large, committed base generating a continuous issuance stream. Meanwhile, the execution-layer base fee on mainnet has collapsed to around 0.335 gwei.
Sit with that number. 0.335 gwei. In the 2021 bull market, base fees routinely exceeded 100 gwei during congestion events. A gwei is one-billionth of an ETH. The base fee is the price of the scarcest resource in the network, and it is trading at fractions of a cent. That is what happens when the demand for L1 execution migrates to Layer 2 and the L1 becomes a settlement and data-availability layer. The settlement layer does not need to be busy. It needs to be secure. Busy was always the application layer's job.
But the burn mechanism was designed when L1 was still expected to be busy. And that is the gap the 2.07% exposes.
Core: Reading the Ledger Line by Line
Let me walk through the actual supply accounting, because this is where most commentary gets lazy.
Over the 282-day window, the numbers break down like this. Total issuance: 796,623.377 ETH. Execution-layer and blob destruction combined: 16,524.553 ETH. Slashing: 1,685.919 ETH. Other destruction: 0.059 ETH, which rounds to nothing.
Net supply increase: 778,412.846 ETH.
That is the number. Net positive. Nearly 780,000 ETH added to the float in under ten months. Annualized, issuance runs around 1,031,000 ETH, destruction runs around 21,400 ETH, and net increase lands near 1,007,000 ETH per year.
Now the ratios. Fee destruction against issuance: 2.07%. Total destruction against issuance โ including slashing: 2.29%.
Notice what I just did. I gave you two different numbers for what looks like the same thing. This is the single most important methodological distinction in the entire dataset, and the tracker that produced it deserves credit for making it explicit. If you include slashing in the destruction figure, you get a higher offset ratio โ 2.29% instead of 2.07%. That looks slightly better. But slashing is not a user paying for block space. Slashing is a validator being punished. Counting it as if it were fee revenue is exactly the kind of cosmetic accounting that makes a bad number look marginally less bad.
The honest number is 2.07%. User demand destroyed 2.07% of what the protocol printed. The other 0.22% was validators losing their own money.
Why does the tracker separate these? Because the methodology tracks issuance and destruction as distinct line items: validator reward issuance, execution-layer base fee burn, blob fee burn, slashing, SELFDESTRUCT destruction. Each category is tagged. When you aggregate naively, you inflate the user-driven offset. When you separate them, you get the truth. I spent the better part of 2017 doing this kind of line-item verification by hand on ERC-20 contracts, and the discipline has not changed: the number that flatters the thesis is almost always the number that requires the loosest definition.
Now let me put the burn in the context of what Ethereum needs to reach neutrality.
At the current network configuration โ a 60 million gas limit per block โ the base fee needs to average about 13.85 gwei for fee destruction to fully offset issuance. That is the break-even threshold.
The current base fee is 0.335 gwei.
That is a gap of approximately 41 times. Ethereum's base fee would need to rise by a factor of forty-one, sustained, just to make the burn equal the mint. Not to make ETH deflationary โ just to make the burn neutral.
The Glamsterdam upgrade introduces a conditional target of a 200 million gas limit. Read that carefully. It is conditional. It is not an automatic new ceiling. The actual gas limit that takes effect depends on what validators configure. The protocol suggests; validators decide. And if the gas limit does expand, the required break-even base fee does not stay at 13.85 gwei.
Here is the counterintuitive part. Expanding the gas limit makes the burn problem worse, not better, in a low-demand environment. More gas per block means the same transaction demand is spread across more units of capacity. Spreading demand across more capacity lowers the price per unit. If gas consumption stays constant while the gas limit rises, the equilibrium base fee falls. The target of 200 million gas, if actually adopted by validators, pushes the break-even base fee down to roughly 4.16 gwei.
That sounds like good news โ a lower bar to clear. But it is the wrong frame. The base fee is not at 13.85 gwei and it is not at 4.16 gwei. It is at 0.335 gwei. Whether the break-even target is 13.85 or 4.16 is irrelevant when the actual price is forty-one times below the lower one and roughly twelve times below the higher one. The expansion does not close the gap. It moves the finish line closer while the runner has already stopped.
This is the technical reality that the expansion narrative obscures: scaling capacity and increasing burn are not the same objective, and under low demand they are in direct tension. More capacity does not create more fees. It dilutes the price of the capacity that exists. The burn is a demand function, not a capacity function. Every article that treats a gas limit increase as bullish for the burn is confusing two unrelated variables.
Let me quantify the staking economics, because this is where the holder's opportunity cost actually lives.
Annual issuance of roughly 1,031,000 ETH against a staked base of 43,754,000 ETH implies a gross staking reward rate of about 2.36%. Gross. Before costs. Running a validator means hardware, a VPS or cloud instance, monitoring, and the operational burden of staying online. Net, a solo staker might keep 1.5% to 2.0%. Through a liquid staking protocol, take another cut for the operator.
So the honest yield on the security budget is low single digits, and it is being paid in newly minted ETH. That is the part people skip. The staking reward is not a dividend from network revenue. Protocol revenue โ the fee burn โ covers 2.07% of issuance. The other 97.93% of the reward is dilution. Validators are being paid in the debasement of everyone who holds ETH and does not stake. Staking yield on Ethereum in 2026 is, in the aggregate, an inflation tax transferred from passive holders to active validators.
Is that a Ponzi? No. I want to be precise here, because the word gets thrown around carelessly. A Ponzi pays early participants with later participants' principal. Ethereum pays validators with protocol issuance for providing a real service โ consensus security. There is no principal pool being drained. The structure is legitimate.
But there is an endogenous spiral worth watching. Issuance scales with staked ETH. More staked ETH means more issuance. More issuance means more reward paid out, which can attract more staking, which raises issuance again. It is not fraud. It is a feedback loop that leans inflationary as participation grows. The tracker's own data makes the point: the amount of ETH staked is a direct determinant of issuance. High participation does not reduce issuance. It increases it.
Now the value-capture question. Where does ETH actually capture value in 2026?
Channel one: L1 fee burn. At 2.07% of issuance, holding ETH as a claim on network revenue returns approximately nothing. The burn is too small to matter.
Channel two: blob fee burn. New, real, and tiny. In one sampled epoch, blob destruction totaled 0.000442 ETH. That is not a typo. Less than a thousandth of an ETH across an entire epoch. Blob space is so cheap and so underutilized relative to its capacity that its fee market is effectively dormant. The data-availability revenue stream that was supposed to flow to L1 from the rollup ecosystem is, at present, a trickle measured in fractions of a cent.
Channel three: priority fees. These go to block producers, not to the burn. Which means validator income and ETH-holder interest are structurally diverging. The validator captures priority fees whether or not the base fee burns. The holder benefits only from the burn. As priority fees become a larger share of the fee pie relative to the burned base fee, the value accrues to validators and away from holders.
Channel four: the necessity use case. Validators must stake 32 ETH to participate. That is a genuine lockup. ETH is also the unit of account for gas. These are real, non-speculative demands. But a necessity use case does not automatically overcome issuance. It sets a floor on utility, not a ceiling on supply.
Put it together. Ethereum in 2026 is running a structural mild inflation, annualized net increase around 0.8%, with value capture that is weak on every channel that matters to a long-term holder.
And here is the comparison that should sting. After the 2024 Bitcoin halving, BTC's annual issuance rate dropped to roughly 0.8%. For years, the Ethereum community argued ETH was the superior monetary asset because it could burn and become deflationary while BTC inflated. In 2026, ETH is inflating at approximately the same rate as post-halving BTC, and it has no halving schedule to look forward to. The scarcity edge that ETH was supposed to have over BTC has, on current data, evaporated. The ultrasound money pitch and the bitcoin inflation pitch now describe the same number. Only one of them has a programmed supply schedule, and it is not Ethereum.
Contrarian: What the Expansion Crowd Gets Wrong
Now the part where I disagree with most of the timeline.
The standard bull response to this data goes like this: low fees are fine because they mean the network is scaling; when adoption returns, fees will rise, the burn will resume, and ETH will become deflationary again. The 2.07% is a trough, not a trend.
That argument has a hole in it, and the hole is the migration of execution to Layer 2.
The reason the base fee is at 0.335 gwei is not that the market is quiet. It is that the high-value execution activity โ swaps, lending, liquidations, the fee-generating transactions โ moved to rollups. Arbitrum, Base, Optimism, and the rest process the activity that used to bid up L1 gas. Those transactions pay fees to their own networks. They pay L1 only for settlement and data availability, and that payment is routed through blob space, which is nearly free.
So the L2 ecosystem is growing, and L1 is collecting almost nothing from it. The rollups post data to Ethereum the way a tenant mails a token rent check to a landlord who never raises the rent. The tenant is thriving. The landlord is watching the building fill up while the revenue line stays flat.
I have heard this described as "L2s are parasitic." I would frame it more precisely: L2s are economically decoupled from L1's fee base. Their success does not translate into L1 burn. Their growth does not require L1 congestion. In fact, their growth is predicated on keeping L1 cheap. The roadmap that scaled Ethereum by pushing activity to L2 also engineered the collapse of the fee burn. This was not a surprise to anyone who modeled it. It was a consequence everyone chose not to model.
And the dozens of Layer 2s competing for the same finite user base make it worse. This is not scaling. It is slicing an already thin liquidity pool into fragments, each with its own bridge, its own liquidity mining subsidy, its own token that dilutes the whole. The competition among L2s drives fees down further, which is good for users and terrible for the L1 burn. Every new L2 is another reason the base fee stays at a fraction of a gwei.
Then there is the validator coordination problem, which nobody prices.
The 200 million gas target is conditional. It depends on validators choosing to raise their configured gas limit. The protocol does not force it. The developers propose; the validator set disposes. This creates a genuine execution risk: a target that exists on paper but never takes effect because the operators who must adopt it have no incentive to. Raising the gas limit increases hardware load โ larger blocks to propagate, validate, and store. It raises the operational cost of running a validator. And in a low-fee environment, the additional priority fee revenue from bigger blocks may not compensate for the added cost.
So we have a scenario where the expansion is announced, the narrative turns positive, and the actual parameter change never fully arrives because the people who have to implement it are quietly indifferent. The market prices the headline. The chain runs the old number. This is the gap between protocol intent and validator action, and it is the kind of gap that makes price action lie.
If the gas limit does rise, though, the effect is the opposite of what the bulls want. As I showed, more capacity at constant demand means a lower equilibrium base fee. The expansion pushes the break-even threshold down but the actual base fee down faster. The net result in a weak-demand environment is a worsening burn-to-issuance ratio, not an improving one. The "scaling is bullish" trade and the "burn is bullish" trade point in opposite directions, and the data resolves the argument in favor of scaling being bearish for the burn.
Let me also dismantle the Solana comparison that gets trotted out. Solana's fee revenue has overtaken Ethereum's on some measures. People cite this as proof that Ethereum is losing. I read it differently. Solana is a high-throughput, low-fee chain where real user activity generates real fee flow to a small validator set. Its economics are internally coherent: activity pays, and payment flows to the network. Ethereum's L1 economics are internally incoherent: activity pays almost nothing, and the payment flows to L2s and away from the burn. The comparison is not embarrassing because Ethereum's fees are lower. It is embarrassing because Ethereum's fee base no longer tracks its own usage.
Here is the blind spot I want to name directly. For two years, the market treated the gas limit and the burn as the same lever. They are not. One is capacity. One is price. Confusing them produced a consensus view that Ethereum's supply was tightening when it was loosening. The retail trader saw the burn happening on the block explorer and assumed scarcity. The smart money looked at the ratio of burn to issuance and understood the burn was cosmetic. That divergence โ retail reading the burn, professionals reading the ratio โ is the whole trade. It always is.
I built and ran an automated arbitrage agent in 2026 across three L2 networks, processing tens of thousands of transactions a day. I can tell you from operational experience that the fee environment on L2s is nothing like L1. The arbitrage that keeps prices aligned happens on the rollups, pays rollup fees, and touches L1 only through the cheapest possible settlement path. The agent's entire economics depended on L1 staying cheap. That is not a coincidence. It is the design. Every sophisticated participant has optimized for low L1 fees because high L1 fees destroy their margins. The sophisticated money is structurally short the burn.
And that is the part that should worry a long-term holder. The people with the most capital and the best infrastructure have arranged their operations to keep the base fee low. They have no reason to want it high. The holder's deflationary thesis requires the base fee to rise by a factor of forty. The marginal, most powerful user of the network actively benefits from it staying where it is.
Takeaway: What to Watch, and What the Number Means
The 2.07% is not a one-off. It is the arithmetic of a mature settlement layer operating in a bear market with activity pushed to Layer 2. Absent a structural change in where fees are paid, it persists.
So watch four things, and watch them like a diagnostic log.
First, the base fee itself. The break-even threshold at the current gas limit is 13.85 gwei. At a 200 million gas limit, it falls to roughly 4.16 gwei. The current reading is 0.335 gwei. Any sustained move toward those thresholds is the only thing that actually matters for the burn thesis. Until the base fee is measured in single-digit gwei on a persistent basis, the burn is noise.
Second, validator adoption of the higher gas limit. The Glamsterdam target is conditional. Track the actual configured gas limit across the validator set, not the announced target. If the limit stays at 60 million, the expansion never happened, and any price move on the announcement was narrative, not mechanism.
Third, blob utilization and blob fees. If rollup data posting ever saturates blob space, blob fees rise and the burn gets a second engine. Current blob destruction is measured in ten-thousandths of an ETH per epoch. Watch for the moment that number stops rounding to zero. That is the earliest signal that L1 data availability is becoming scarce.
Fourth, the staking participation curve. Issuance scales with staked ETH. If the staking rate keeps climbing, issuance keeps climbing, and the burn has to run faster just to stand still. Falling net staking yield โ currently low single digits gross, lower net โ could eventually reverse participation and bring issuance down. That would be a supply tailwind nobody is modeling.
I want to be clear about what I am and am not claiming. I am not claiming Ethereum is broken. The network is secure, the settlement layer works, the rollup ecosystem is real. I am claiming that the specific monetary narrative attached to ETH โ the one about guaranteed scarcity โ was always a bet on demand for L1 block space, and that bet has not paid off. The mechanism did exactly what it was designed to do. It burned fees when fees existed. In 2026, the fees are elsewhere.
Code doesn't lie. It just stops burning when demand leaves. The ledger printed nearly 780,000 net new ETH in under ten months, and it will keep doing so until either demand returns to L1 or the issuance curve bends. Trust is a variable; verify the proof, then sleep. The proof, right now, is a 2.07% offset ratio and a base fee of 0.335 gwei, and neither of those numbers is a rounding error you can afford to ignore if your thesis depends on ETH getting scarcer every day.
So here is the question I would leave with anyone holding ETH for the scarcity story. If the burn covers 2% of issuance, the base fee sits forty-one times below break-even, the most sophisticated users are economically short the burn, and the scaling roadmap mechanically suppresses it further โ what exactly is the mechanism by which ETH becomes scarce? Not someday. On what schedule, at what base fee, adopted by which validators, with demand from which users who are currently paying L2s instead? Answer that, and you have a thesis. Fail to answer it, and you have a narrative with a number attached that no longer supports it.
The chart shows the price. The supply ledger shows the truth. They have been telling different stories for a year. Only one of them has to reconcile eventually.