Saving ETH staking: the consequences we have not modelled
Ethereum’s staking ratio is rising because ETH finally became a productive asset with a forecastable cash flow, a regulated wrapper, and a working credit market on top of it. The tapered issuance burn treats that as a pathology to be corrected. I think it is the most consequential economic change proposed since the Merge, and that its second-order effects have not been modelled at all.
Every figure below is derived from the proposal’s own formulas: issuance of 64·√D Gwei per epoch, burn fraction b = (D/60,250,000)^1.5, and the stated upper bound of 78,300 ETH in annual execution-layer rewards. The modelling reproduces every number the proposal publishes (1,051,200 ETH annual issuance at 40M staked, 2.62% gross yield, 93% issuance share of income, the 19.84% issuance peak, ~1.2% net yield), so it should be straightforward to reconcile.
| Staked | % supply | burn b | gross CL | net CL | MEV | all-in | yield sensitivity vs today | outage recovery vs today |
|---|---|---|---|---|---|---|---|---|
| 39.0M | 32.0% | 0.521 | 2.661% | 1.275% | 0.201% | 1.476% | 4.3x | 3.6x |
| 42.0M | 34.5% | 0.582 | 2.565% | 1.072% | 0.186% | 1.258% | 5.2x | 4.3x |
| 48.0M | 39.4% | 0.711 | 2.399% | 0.693% | 0.163% | 0.856% | 8.4x | 6.8x |
| 54.0M | 44.4% | 0.849 | 2.262% | 0.343% | 0.145% | 0.488% | 17.8x | 14.2x |
| 60.25M | 49.5% | 1.000 | 2.141% | 0.000% | 0.130% | 0.130% | infinite | infinite |
At an unchanged 39M staked, all-in validator income falls from 2.862% to 1.476%, a 48% cut, delivered on a published 548-day schedule in 64 steps of 8.6 days each.
1. Zero-yield staking selects for institutional staking
This is the objection I would most like the authors to answer, because it inverts the entire logic of the proposal.
The stated purpose is to stop the ETH supply being captured by a handful of custodians, exchanges and ETF issuers. So consider who is still staking when net consensus yield is 0.7%, or 0.3%, or zero.
Not solo stakers. They stake because of the yield and because of conviction, and conviction does not pay a $500 annual electricity and hardware bill. Their costs are fixed rather than proportional, so they leave first.
Who stays? Everyone who stakes for reasons other than yield. An ETF issuer that has to show a distribution yield on the same screen as a Solana product will stake at 0.13% if the alternative is 0.00%. A corporate treasury stakes to book reported income. An exchange stakes because it is a product line, not an investment. A sophisticated operator stakes because MEV, which this proposal does not touch, still pays.
A zero-yield regime accelerates the capture it means to deter. It filters out everyone who stakes for economic return and leaves the field to entities that stake for structural, regulatory, or product reasons, which describes exactly the KYC’d, jurisdiction-bound, coercible operators the proposal fears. Revenue cuts favour whoever holds the lowest cost per validator and the best MEV capture, so cutting revenue cannot fix a composition problem.
2. What this does to a home validator
A 32 ETH home validator, ETH at $1,900, $500 a year in amortised hardware and European electricity, 30% marginal rate assessed on gross credited rewards:
| - | gross credited (taxable) | net received | after tax and opex |
|---|---|---|---|
| Today | 0.916 ETH ($1,740) | 0.916 ETH ($1,740) | $718 |
| At activation, BRF 128 | 1.768 ETH ($3,358) | 0.881 ETH ($1,673) | $165 |
| End of transition | 0.916 ETH ($1,740) | 0.472 ETH ($898) | -$125 |
Look at the middle row, because I do not think anyone has noticed it. The transition works by doubling BASE_REWARD_FACTOR to 128 and decaying it. That doubles the gross credited reward while burning 52% of it. Net yield is roughly unchanged, which is the stated intent, but taxable receipts double. The mechanism designed to cushion the impact is the mechanism that creates the exposure. A home staker in a jurisdiction that taxes rewards on receipt sees after-tax income fall 77% at the exact moment the proposal says nothing has happened to them.
This risk exists because the reduction is implemented as a post-credit burn rather than a smaller reward. The spec credits the full amount through the existing machinery, then applies a separate decrease_balance. If a revenue service assesses income on the credit and treats the burn as a capital loss, deductible only against gains and capped annually in many jurisdictions, a flawlessly performing validator runs at a loss. Build the reduction into the reward curve and the question never arises.
I do not know how any tax authority will treat this, and neither does anyone else. That is the problem. It is a several-hundred-dollar annual swing on the exact cohort the proposal claims to protect, and it cannot be settled by argument on this forum. It needs written opinions.
Note also that the proposal cites the nominal-yield tax asymmetry between solo stakers and shielded ETPs and wrapped LSTs as a reason to act. The chosen implementation is the one design that risks widening it.
Then add the proposal’s own downtime figure. Because penalties stay at full magnitude while net earnings collapse, recovery from an outage takes 3.6x longer at today’s stake, 6.8x at 48M, 14x at 54M. That is a transfer of advantage to operators with redundant power, failover and 24/7 monitoring. It is a tax on residential internet, and it compounds everything above.
3. If the premise is right, the mechanism does not work
The proposal argues the staking risk premium is trending toward zero: infrastructure has matured, LSTs and ETFs have stripped out friction, slashing credibility is degraded at high stake. It then argues equilibrium arrives where net yield meets that premium, comfortably below 50%.
Both claims cannot be load-bearing.
If the premium really is collapsing toward zero, equilibrium converges on the point where net yield is zero: 60.25M ETH for issuance, and above it once MEV is included, because MEV is not burned and accrues pro rata at any ratio. The proposal concedes this in one sentence and does not follow it through. The outcome is a ~50% staking ratio with zero issuance, zero net reward for honest validators, negative returns for anyone with imperfect operations, and every concentration risk it set out to prevent, now with no security budget attached.
If the premium is materially positive, stake growth was going to stall on its own and the intervention was unnecessary.
There is no configuration of the world in which this achieves its goal.
4. It removes the reason to hold ETH onchain
Staking yield is the reference rate for ETH-denominated yield. Almost everything onchain that pays a return on ETH is priced off it.
This proposal cuts most of that yield away. All-in income falls 48% immediately at today’s stake, to 0.86% at 48M, to 0.13% at saturation. Whatever is left is not enough to price a lending curve, a fixed-yield market, or a basis trade around. ETH stops being a productive asset onchain and becomes beta only.
That matters because of who was holding it. A large share of onchain capital was comfortable carrying ETH price risk specifically because the position also paid. Remove the yield and the trade no longer makes sense: the same allocator can hold stablecoins earning 4-5% and take beta somewhere cheaper, or not at all. We would be pushing capital out of ETH-denominated positions into dollar-denominated ones inside our own ecosystem, and handing the reference-rate role in DeFi to Circle and Tether.
And if ETH is beta only, it is not obviously the best beta. That is the part I find hardest to look past. ETH’s case against BTC for an allocator was never that it was a superior store of value; it was that it was a productive asset with a cash flow BTC structurally cannot offer. Take the yield away and ETH is left competing on BTC’s home ground, where it has to win on liquidity depth, ETF distribution, institutional acceptance and narrative simplicity. It does not win on any of those. What it does offer is higher volatility, more protocol risk and more governance risk, and the yield was the compensation for exactly that. Strip out the compensation and you have an asset that moves in the same direction as BTC, with more downside, and no income to hold you through it. Anyone who wants crypto beta rotates to BTC, and anyone who wants yield rotates to stablecoins. ETH gets squeezed from both sides at once.
Japan spent two decades demonstrating where that ends. A currency that pays nothing does not become the thing people hold; it becomes the thing people borrow. It turns into the funding leg of everyone else’s trade. If ETH yields near zero while dollars onchain yield 4-5%, the natural position is to borrow ETH and hold stablecoins, and the dominant reason to borrow ETH becomes to be short it. We would be volunteering ETH to be the funding currency of its own economy.
The proposal argues ETH becomes better money by paying nothing. Zero-rate assets are what you hold at the edges of a system, not what you build credit on. “ETH as money” is not served by deleting ETH’s term structure.
5. What this does to the cash flow institutions finally underwrote
US spot ETH ETFs have taken roughly $11.2 billion in cumulative net inflows, and BlackRock’s product now passes staking rewards through to holders. That demand exists because ETH became the first major crypto asset with a regulated, distributable, forecastable yield.
This proposal does two things to it.
It sets the terminal value to zero. Not lower, zero, reached by a mechanism no holder controls. No income desk underwrites an instrument whose coupon is designed to be extinguishable by other people’s behaviour.
It multiplies the yield’s sensitivity to the staking ratio by 4.3x immediately and 8.4x at 48M staked. Today yield scales as D^(-1/2), so elasticity to the staked base is exactly -0.5: a 10% forecast error on the ratio costs 5% of your yield estimate. Under the taper it is -2.13 today and -4.19 at 48M, so the same error becomes a 21% or 42% error in your income projection. The staked base has moved from roughly 25% to 32% of supply in eighteen months. Nobody forecasts it within 10%.
Meanwhile Solana staking products advertise net rewards above 7%. Yes, that is nominal and Solana’s real yield after dilution is a different number. Institutional screens are run on distribution yield, and the answer to “why hold ETH for income” becomes “you shouldn’t.”
6. MEV share, censorship, and MEV burn
At today’s stake this takes MEV from 7.0% to 13.6% of validator income. At 48M it is 19.1%; at 54M, 29.7%.
Issuance is smooth and scale-neutral. MEV is lumpy, rewards sophistication, and rewards variance-smoothing across large validator sets. Roughly doubling its weight in validator income is a centralising act, treated here as a footnote.
It also carries a neutrality cost. Only a minority of major MEV-Boost relays are non-censoring. When issuance is 93% of revenue, choosing a neutral relay set or building locally is an affordable principle. When it is 70%, it is a business decision you have to justify. This proposal makes credibly neutral block production more expensive relative to compliant block production, which is the opposite of its purpose.
And the roadmap composition should be stated plainly, since the proposal invites it: this EIP plus MEV burn leaves validators with 1.275% at today’s stake and 0.693% at 48M, with no execution-layer income at all. If both ship we have removed essentially the entire validator revenue stream. That endpoint should be argued for explicitly, not arrived at by two separately reasonable increments.
7. The size of the bet
The supply-side benefit is calculable. At today’s stake the burn destroys 540,566 ETH a year, 0.44% of supply, about $1.03 billion at $1,900.
Against that we are wagering the demand channel that produced roughly $11.2 billion of ETF inflows and an entry queue above 3 million ETH. One year of burn is on the order of 18% of one year of that inflow. If removing the yield costs us even a fifth of that demand, the trade is a wash. If it costs more, we have reduced the dollar value of the stake securing the chain, which is the quantity the ETH-as-money argument says it wants to protect.
The security case runs: less issuance, higher monetary premium, higher ETH price, more real value at slashing risk. That channel is speculative and reflexive. The demand-destruction channel is observable and already priced. We would be betting the security budget on the untested one.
There is also no off-ramp. No feedback control, no floor, no stabiliser. If the staking ratio overshoots down to 20% of supply we find out after the fork, and reversal needs another contentious hard fork negotiated in whatever conditions made reversal necessary.
What I am asking for
Dilution is not costless and stake concentration is a real problem. My argument is that issuance is the wrong instrument, that the consequences are unmodelled, and that the burden of proof for a change this large sits with the proposal. Before this goes anywhere we need:
-
A solo-staker impact assessment, with after-tax, after-opex numbers, published by the authors rather than reconstructed by critics.
-
Written tax opinions from at least the US, UK, Germany and Portugal on whether a post-credit burn reduces assessable income or creates a capital loss. If the answer is unfavourable anywhere material, implement the reduction in the reward curve instead.
-
A cascade model for the LST and lending stack, built with Aave, Lido, Etherfi and other DeFi risk teams, covering the window between SFI and activation rather than the steady state.
-
A published demand elasticity, a projected equilibrium with confidence intervals, and a stated response to overshoot. The whole case rests on an unobservable risk premium. Name it, defend the estimate, say what we do if it is wrong.
-
A hard, non-zero floor on net yield. This is the change that would move me furthest. Most of the claimed benefit, removing the marginal incentive to grow the ratio, survives a floor. What does not survive is a terminal value of zero, and that is what breaks institutional pricing, ETH’s onchain yield, and the composition of the validator set.
Targeting concentration directly is available and unexplored here: penalties scaled to correlated failure, which prices the externality large operators impose without touching the yield of being small, and inclusion lists and proposer-set reform, which address the scale advantage that actually exists.
If you run a validator, post your own numbers in this thread: gross rewards, net under the taper, opex, and how your jurisdiction treats staking income. The strongest argument this proposal has right now is that nobody has quantified who pays for it. Let’s fix that.
Ethereum’s staking ratio is rising because staking works. That is what growth looks like, and we should not be engineering a mechanism that punishes it.
I have been building on Ethereum for ten years, and I am writing this because I am concerned about the direction we are taking. It took most of that decade to arrive where we are: staking that attracts serious commitment from every category of participant, from home validators to funds to regulated issuers; a DeFi ecosystem deep enough that ETH is treated as real collateral rather than a speculative chip; and now RWA issuers choosing Ethereum specifically because they can rely on its safety and its settlement guarantees to carry assets that have to exist for decades. Those three things reinforce each other, and each one of them rests on ETH being a productive asset with a dependable yield. None of it happened by accident, none of it is guaranteed to survive a change of this magnitude, and this proposal puts all three at risk simultaneously on the strength of an equilibrium nobody has measured.
Save ETH staking.