Did L2s break Ethereum’s ultrasound money?

Ethereum’s best marketing line was that using it destroyed it, that every transaction burned ETH and shrank the supply. Then the network solved its scaling problem, activity fled to layer 2s, and the burn collapsed. The scaling worked. The scarcity did not survive it.

For about eighteen months, Ethereum had the best story in crypto, and the story was a paradox: the more people used the network, the rarer its token became. Every transaction burned a little ETH, and when the network was busy enough, it burned more than it created. Supply went down. The community called it ultrasound money, a deliberate jab at Bitcoin’s “sound money,” complete with a bat emoji and a movement.

For a while, the data backed it up. Then Ethereum did the thing it had promised to do for years, which was to scale, and scaling broke the story. Activity moved to layer-2 networks that pay almost nothing to the base chain, the burn collapsed, and ETH quietly went inflationary again. This is the story of how Ethereum’s greatest technical success dismantled its best economic narrative, and whether a December upgrade can put the pieces back.

What ultrasound money actually meant

The mechanism is worth getting exactly right, because the whole debate turns on it.

In August 2021, Ethereum activated EIP-1559, which changed how transaction fees work. Instead of paying miners directly, every transaction now pays a base fee that is burned, permanently removed from circulation. The busier the network, the higher the base fee, and the more ETH destroyed. On its own, that is just a fee-burning mechanism. It became a monetary thesis when Ethereum switched from proof-of-work to proof-of-stake in the September 2022 Merge, which cut new ETH issuance by roughly 90%, because the network no longer had to pay energy-intensive miners.

Put the two together, and you get the ultrasound thesis. Issuance dropped to a trickle after the Merge. Burning continued with every transaction. If burning exceeded issuance, total ETH supply would shrink over time, making the asset deflationary. And a deflationary asset with growing demand should, in theory, appreciate. Ethereum would become harder money than Bitcoin, whose supply still grows, hence “ultrasound.” The tracking site ultrasound.money existed to display exactly this: supply ticking down, day by day.

For a stretch after the Merge, it happened. Supply fell back toward and below the level it sat at during the Merge itself. Burns outpaced issuance. The narrative was not hype; it was, for that window, an accurate description of the data. That is what made it powerful, and what made its reversal so awkward.

How scaling broke it

The break came from Ethereum solving its most famous problem, and the irony is total.

Ethereum’s scaling strategy is to push transactions off the expensive base layer and onto layer-2 rollups, networks like Arbitrum, Optimism, and Base that process transactions cheaply and then post compressed data back to Ethereum for security. The base layer becomes a settlement and data-availability layer; the rollups handle the actual activity. This is the roadmap Ethereum has pursued for years, and it works.

The March 2024 Dencun upgrade was the pivotal moment. It introduced EIP-4844, “blob” transactions, a separate and far cheaper data channel for rollups to post their data. Costs for layer 2s dropped by a factor of 10 to 100. Activity that used to happen on mainnet, paying mainnet fees and burning mainnet ETH, moved to rollups paying blob fees that were, in practice, close to zero because blob space was massively oversupplied relative to demand.

The effect on the burn was immediate and severe. Before Dencun, Ethereum burned thousands of ETH per day during busy periods. After Dencun, daily burn dropped to as low as 50 to 70 ETH. The base layer had lost its primary fee source. With issuance running around 1,700 ETH per day and burn collapsing well below that, the equation flipped: Ethereum began creating more ETH than it destroyed. By various measures across 2025 and into 2026, net annual inflation ran somewhere between roughly 0.2% and 0.8%, depending on the window. ETH supply crossed back above its Merge-era level. The deflation was over.

The mechanism that made ultrasound money true, EIP-1559 burning at scale, had not been removed. It had been bypassed. The activity simply moved to a layer where the burn does not happen in any meaningful amount. Ethereum scaled successfully and, in doing so, severed the link between usage and scarcity that the entire thesis depended on.

The bull case: it still works, just differently

The response from Ethereum’s defenders is not denial. It is reframing, and parts of it are genuinely strong.

The first point is that elastic scarcity is the actual feature, not permanent deflation. Ethereum was never designed to deflate forever at a fixed rate. It was designed to burn in proportion to demand, which means it becomes deflationary when the network is busy and mildly inflationary when it is quiet. During periods of high mainnet activity, above roughly 16 gwei average gas, burn still exceeds issuance, and ETH still goes net deflationary, temporarily. The mechanism works exactly as designed; it is just that a scaled network spends more time in the quiet regime. In this reading, ultrasound money was always conditional, and the condition is demand, not a promise.

The second point is that issuance is still radically lower than before. Even mildly inflationary, Ethereum issues roughly 90% less ETH than it did under proof-of-work. Compared to Bitcoin, which currently inflates at around 0.8% annually on a fixed schedule, Ethereum’s roughly 0.2% net inflation in calmer periods is actually lower. Both assets inflate in 2026; Ethereum, by some measures, inflates less. The “harder than Bitcoin” claim survives in a narrow, technical form even without net deflation.

The third point is that the supply figure overstates the sell pressure. Roughly 28% to 30% of all ETH is locked in staking, earning yield and not circulating. The tradeable float, ETH actually available on exchanges, is meaningfully smaller than the headline supply number, and it shrinks as more ETH is staked. A modestly inflating total supply with a large and growing staked portion is a very different pressure than the raw inflation number suggests. Demand from ETFs, treasury companies, and staking can absorb 0.2% inflation without difficulty.

And the fourth point is simply that the store-of-value case never rested on deflation alone. As long as demand for Ethereum’s blockspace, its role as settlement for stablecoins, tokenization, and DeFi, grows faster than supply, price can rise regardless of whether supply ticks up 0.2% a year. Scarcity was a nice story. Utility is the real thesis.

The bear case: the narrative was load-bearing

The skeptical reading is that the ultrasound story was not just marketing, that it was doing real work in the investment case, and that losing it matters more than the reframing admits.

The blunt version comes from the on-chain data and the people watching it leave. Daily network fee revenue on Ethereum fell from near $40 million in early 2025 to a local low around $10 million in 2026. That is not just a burn problem; it is a value-accrual problem. If the base layer captures little fee revenue because activity happens on rollups that pay it almost nothing, then holding ETH is a bet on an asset whose own network is monetizing its users poorly. Some analyses have tied this directly to developer attrition and reduced whale support, framing the end of ultrasound money as the end of a period when ETH had a clean, quantifiable reason to appreciate.

The deeper problem is structural and hard to argue away: a scaled, efficient Ethereum is less deflationary than a congested, expensive one. This is the tension at the center of the whole debate. The very thing that makes Ethereum better as infrastructure, cheap transactions, more capacity, activity on fast rollups, is the thing that reduces the burn. Ethereum cannot simultaneously be the cheap, high-throughput settlement layer it wants to be and the fee-burning deflationary asset the ultrasound thesis needed. Those are in direct conflict, and the roadmap chose scaling. The asset thesis was, in a real sense, sacrificed to the technology roadmap.

Then there is the value-capture question that rollups sharpen. Layer 2s use Ethereum for security and pay it a pittance for the privilege. Robinhood’s own chain is an example: analyses of corporate L2s show the base layer capturing a rounding error of the economics while providing the security that makes the whole arrangement credible. If Ethereum’s future is thousands of rollups settling to it cheaply, then Ethereum is providing enormous value and capturing little of it, and no amount of narrative reframing fixes a value-capture problem that lives in the fee structure.

The fix nobody is talking about

Which brings us to December 2025, and the upgrade that was designed, in part, to address exactly this, and that most of the market ignored.

The Fusaka upgrade activated on December 3, 2025. Its headline features were about scaling further, PeerDAS and expanded blob capacity. But buried in it was EIP-7918, the “blob base fee bound,” which is the most direct attempt yet to repair the burn. The problem Dencun created was that blob fees could collapse to near-zero, one wei, when execution costs dominated and blob demand was soft, which meant rollups consumed Ethereum’s capacity almost for free and burned almost nothing. EIP-7918 sets a floor: it ties the minimum blob fee to the execution base fee, roughly the execution base fee divided by 16, so that even in quiet periods rollups pay a meaningful minimum, and a minimum stream of ETH gets burned.

The modeling is striking. Fidelity Digital Assets analyzed what would have happened if EIP-7918 had been active since blobs launched, and found that on 93% of days since the 2024 Dencun upgrade, the adjusted fee would have exceeded the actual fee, generating an estimated additional $78.6 million, roughly 24,641 ETH, in cumulative blob-fee revenue. Blockworks noted that had the mechanism been introduced in June 2025, burnt blob fees would have been nearly 8x higher. The intent is explicit: restore a floor under the burn so that as stablecoins, DeFi, and tokenization migrate to rollups, ETH still captures value from that activity instead of subsidizing it.

The honest caveat is that this is a floor, not a restoration. EIP-7918 prevents the burn from collapsing to zero; it does not recreate the thousands-of-ETH-per-day burn of the congested mainnet era. Whether it produces measurable, sustained deflation depends on how much activity flows through blobs and how high execution base fees run, and the market is still watching. It is a serious, well-designed attempt to reconnect usage and scarcity. It is not a return to 2022.

Sound money versus ultrasound money, honestly compared

Because the entire thesis was built as a shot at Bitcoin, it is worth putting the two monetary models side by side without the tribalism, since the comparison is more interesting than either camp admits.

Bitcoin offers fixed scarcity. The supply schedule is written into the protocol, capped at 21 million coins, and halves on a predictable timetable roughly every four years. A holder knows today, with certainty, what Bitcoin’s issuance will be in 2030 and 2040. That certainty is the entire product. Bitcoin does not react to demand, does not burn, does not adjust; it simply issues on schedule toward a hard cap, and its current inflation runs around 0.8% annually, trending toward zero over decades. The trade-off Bitcoin holders accept is that the base layer offers little native utility and no yield. You hold it for the certainty, and you give up productivity in exchange.

Ethereum offered, and to a degree still offers, elastic scarcity. Supply responds to network demand: high usage burns more and can push ETH net deflationary; low usage burns less and lets mild inflation through. The appeal was a token that becomes scarcer precisely when it is most used, tying the asset’s scarcity to the network’s success. The trade-off, which the L2 era exposed, is that elasticity cuts both ways.

A demand-responsive supply is only deflationary when demand is high on the layer that burns, and Ethereum deliberately moved demand to layers that do not burn. Bitcoin’s rigidity, often criticized as inflexible, turned out to be the thing that made its monetary promise keepable. Ethereum’s flexibility, often praised as sophisticated, turned out to be the thing that made its monetary promise conditional.