Ethereum's most foundational economic thesis quietly died over the past two years, and nobody held a funeral. The panda has been watching the numbers slide while every L2 launch press release recycled the same line about scaling Ethereum to billions of users. The thesis here is simple. L2s scaled Ethereum's users. They did not scale Ethereum's value accrual. They reversed it.
Where did the L2 fees actually go?
According to CoinGecko's global market data, ETH dominance sits at 8.93% on June 12, 2026, with a market cap of $201.92B against BTC's $1.28T. Four years ago, that ratio was closer to one in three. ETH did not get smaller in raw user terms. The Ethereum L2 stack, by L2Beat's own tracking, processes more daily transactions than the L1 ever did at peak. So where did the value go?
The simple answer: into rollup treasuries. Arbitrum, Optimism, Base, and ZkSync each operate centralized sequencers. Sequencer revenue (the spread between user gas paid on the L2 and the cost of posting compressed data back to L1) flows to the operator, not to ETH stakers. In 2025, that spread cleared into the hundreds of millions of dollars annually for Base alone. Coinbase pockets it. ETH stakers see none of it.
The same pattern holds across the stack with minor variations. Arbitrum's sequencer revenue flows into the Arbitrum DAO treasury, denominated in ARB and ETH, governed by ARB holders. Optimism's Retroactive Public Goods Funding gets funded by sequencer profit, technically pro-social, structurally still a redirect away from ETH. ZkSync's economics route through the Matter Labs treasury. The unifying feature is not how the money is spent but which token holders get to spend it. None of those token holders are ETH stakers in their capacity as ETH stakers.
This was meant to be temporary. The "decentralize the sequencer" roadmap was meant to redistribute value back to ETH. As of June 2026, not a single major L2 has shipped a fully decentralized sequencer with ETH-aligned fee burn. Base still settles to Coinbase. Arbitrum still routes profits to ARB holders. Optimism still feeds OP's economic security with sequencer revenue that bypasses ETH entirely.
Vitalik wrote in 2020 that L2s would "scale Ethereum and ultimately strengthen the value of ETH." He was sincere. The execution diverged. Every major L2 became a competing platform, optimized for its own token, with ETH demoted to a settlement guarantee that does not collect economic rent on the activity it secures. The Ethereum brand expanded. The Ethereum asset did not. Those are two different bets, and in 2026 only one of them is paying off.
The blob subsidy and what it cost ETH stakers
EIP-4844 (proto-danksharding) shipped via the Dencun upgrade in March 2024. It added blob data: a cheaper way for L2s to post compressed transaction data to Ethereum. It worked, in the sense that L2 fees dropped roughly 95% overnight. It also gutted ETH's fee burn mechanism.
Before 4844, L2s paid L1 gas, which routed into the EIP-1559 burn. ETH became net deflationary. Sound money narrative intact. After 4844, blob data uses a separate fee market that burns far less ETH per unit of work. According to ETH supply data tracked on Etherscan, ETH supply flipped net inflationary in mid-2024 and has stayed that way for 24 consecutive months as of June 12, 2026.
The math is brutal: Ethereum subsidized the L2 ecosystem with its own monetary tightness. The blob discount is the subsidy. The discount comes out of ETH staker revenue. Stakers got the consolation prize, since blob fees still flow to them, just at a fraction of the previous rate. The L2 operators got the actual prize: sub-cent transactions for end users while keeping the sequencer markup.
Where the panda raises an eyebrow: this trade was framed as a long-term play. Cheaper L2s, more users, eventually more demand for blockspace, restored deflation. By June 2026, none of the second-order effects have arrived. L2 daily active users are up, but the marginal user moved value off-L1, not onto it. Aggregate ETH-denominated activity on L1 (DEX volume, NFT settlement, contract calls) has stayed flat or fallen since 4844 shipped. The user growth happened on platforms that do not pay ETH stakers.
There is a second-order issue worth naming. Blob pricing was designed around the assumption that L2 capacity would saturate quickly, pulling blob fees up as a normal supply-demand market. That has not happened. Target blob count remained well below the per-block cap for most of the post-Dencun period, meaning the minimum base fee for blobs stays near floor for long stretches. ETH stakers do not get auctioned value because the auction barely activates. The fee market is, mechanically, doing what the spec said it would do. The economic result is a bigger transfer to L2 operators than anyone modelled in 2023.
What is the counter-argument?
The bull case has not disappeared. Three objections deserve a fair hearing.
Objection one: blob revenue is structural, not negligible. As L2 transaction volume grows, total blob fees grow with it. Eventually, the absolute dollar amount routed to ETH stakers exceeds what L1-only Ethereum would have generated. This is mathematically possible. The current trajectory does not yet support it. Blob revenue in Q1 2026 was roughly 12% of the pre-Dencun L1 base-fee revenue, growing at a pace that would not match the prior burn for at least three more years. Linear extrapolation of L2 transaction growth gets blob revenue to parity with pre-4844 base-fee burn somewhere in 2029, assuming no further fee market reforms. Three years of subsidy is a long horizon to ask ETH holders to underwrite while the dominance number keeps slipping.
Objection two: the alternative was worse. Without 4844, L2 fees would have stayed high, users would have migrated to Solana, BNB Chain, and Tron, and Ethereum's settlement franchise would have collapsed. There is real truth here. The counter-counter is that the franchise has eroded anyway, just in a slower mode. ETH dominance at 8.93% is not the outcome of a winning defensive strategy. It is the outcome of a subsidy that funded the closest competitors, several of whom share the Ethereum brand.
Objection three: decentralized sequencers are coming. Espresso, Astria, and the Based Rollup movement all promise to re-align L2 economics with L1. Maybe. The industry has been hearing "next quarter" on sequencer decentralization for nine consecutive quarters. The economic incentives push in the opposite direction. Centralized sequencer revenue is too valuable for any individual L2 to give up voluntarily, and no Ethereum-level governance body can force them to.
The objections are real. None of them overturns the central observation: as of mid-2026, the L2 stack functions as a permanent value-extraction layer on top of ETH. The default assumption should be that this continues until concrete economic redesign ships and runs at scale, not at the proposal stage.
Predictions, dated and falsifiable
Three calls. Mark them. Check them.
By end of Q4 2026: ETH dominance will not exceed 11%. The combination of sticky BTC ETF inflows and the L2 value leak makes a sharp dominance recovery structurally difficult. If ETH crosses 11% before December 31, 2026, this thesis is wrong about the magnitude of the drag, even if the mechanism is intact.
By end of Q2 2027: no major L2 (top five by TVL) will have shipped a fully decentralized, ETH-aligned sequencer in mainnet production. "Stage 2" decentralization milestones do not count. The bar is concrete: sequencer revenue either burns ETH or routes directly to L1 stakers, with no central operator gatekeeping the queue.
By end of Q4 2027: ETH supply will not have returned to net deflation on a 30-day rolling basis. If it does, blob volume will be the cause, and that vindicates the original 4844 bet at least partially, even if dominance keeps slipping.
If two of three fail, the thesis is wrong. If two of three hold, the value-tax framing is the right way to read ETH for the rest of this cycle.
What to watch next
The signals worth tracking are narrow and concrete. First, sequencer revenue disclosure: Coinbase quarterly reports now itemize Base economics, and Robinhood's own L2 announcement raises the bar for what every other operator will eventually have to disclose. Watch the line. Second, the fate of account abstraction and Pectra-era execution work coming out of the Ethereum core developer blog, which determines whether L1 reclaims any user-facing utility, not just settlement guarantees. Third, the staking ratio: if ETH stakers exit faster than blob fees can replace burn, the negative feedback loop accelerates. A drop below 25% ETH staked, currently around 28%, would suggest the value-tax is starting to compound on the security side as well as the price side.
Fourth and most underrated, watch the rate at which L1-native demand renews. If the next wave of Ethereum-defining applications (institutional tokenization, AI agent settlement, regulated DEXes) chooses L1 directly rather than launching as a rollup, the dominance argument has a self-healing path. If they launch as L2s by default, the tax persists by structural inertia. So far the default is "launch as an L2," and the few L1-first counter-examples have not gained scale.
A small structural note that matters less for thesis confirmation and more for context. Monolithic chains (BSC at $5.24B TVL per DefiLlama, Solana, Tron) avoid the value-leak problem by design. Validators capture fee revenue directly. No sequencer treasury sits in the middle siphoning the gap. Not advocating, just observing the architectural difference. The BSC ecosystem where Dadacoin lives runs without the same fragmentation tax, which is part of why fee revenue and validator economics track each other linearly there.
Compare and contrast with prior coverage in the Ethereum analysis cluster: the ETH dominance two-asset thesis from June 5 framed the dominance decline as a BTC-ETF demand story. The Layer 2 lens here is the supply-side counterpart: even if demand stabilizes, value still leaks. Similarly, the Optimism Superchain token paradox covered the governance-token side of the same divergence. Together they triangulate the same broken hinge from three angles.
The panda continues to watch. The Ethereum brand will be fine. The Ethereum asset is the one that has the harder argument left to make.
