Ethereum Gas at 5-Year Low: The L2 Revenue Problem

The question every platform lead building on Ethereum should be putting in front of their CFO this quarter is straightforward: if L1 fee revenue collapses 98% and your rollup vendor's economics depend on data blob pricing that just got engineered down to near-zero, who exactly is paying to secure the chain your product sits on? Median gas dropped to 1.9 gwei, a 5-year low, and the celebration in engineering Slack channels is masking a very real treasury conversation that a lot of teams haven't had yet.

The Numbers

Start with the headline figure. As Yellow.com reported, Dune Analytics data shows Ethereum's median gas fees bottomed at 1.9 gwei on August 10th, a 98% drop from the March year-to-date high of 83.1 gwei. Low-priority transactions now clear at roughly 1 gwei or less, about seven cents per transaction. Fees this low have not been seen since mid-2019, before DeFi Summer, before NFTs, before the entire ecosystem that current L2 economics assume.

The proximate cause is the Dencun upgrade in March, which introduced data blobs (proto-danksharding) with the explicit goal of slashing transaction costs for layer-2 blockchains. Dencun did exactly what it was designed to do. The EIP process shipped a working scalability primitive, and rollups immediately captured the cost savings. That is engineering success by any reasonable definition.

Now look at the volume distribution. L2Beat data shows Ethereum's layer-2 activity hit 33 million transactions in the last 30 days. The base blockchain processed 109 million transactions over the same window. Arbitrum and Taiko together handled 97 million transactions in the past month. That is a lot of throughput, and it is running at a fraction of pre-Dencun cost.

Then the number that should make treasury teams nervous. Ultra Sound Money reports the Ether supply increased by approximately 13,400 ETH in the last week, roughly $34.1 million at the time of reporting. Supply went up despite reduced transaction usage and lower staker payouts. Translation: burn is not offsetting issuance. The deflationary narrative that carried a lot of institutional pitch decks in 2022 and 2023 is, for this window at least, empirically wrong.

Martin Köppelmann, co-founder of Gnosis, put the concern plainly on X on August 10th: "Ethereum needs to get more L1 activity again." He was explicit that financing staking incentives is the worry. When an L1 co-founder is publicly flagging the revenue side of the ledger, the CFO of any protocol paying validators, sequencers, or bridge operators in ETH should be listening.

What's Actually New

Every cycle has a "gas is cheap" moment. This one is different for two structural reasons, and platform leads should be clear on both before they make procurement decisions in the next 90 days.

First, the cheapness is not cyclical, it is architectural. Prior gas-fee troughs came from bear markets, from users leaving. This trough coincides with over 100 million monthly transactions on L1 and another 33 million on L2s. Demand did not disappear. Supply of block space expanded via blobs, and the marginal cost of settlement collapsed. That is a permanent shift in the cost curve, not a temporary dip you can wait out. Any capacity planning that assumes fees revert to 2023 norms is broken.

Second, the value-accrual assumption for ETH holders has quietly inverted. For roughly two years the pitch to institutional allocators, and to engineering VPs justifying ETH-denominated treasury reserves, was that EIP-1559 burn plus staking would compound deflationary pressure. Ultra Sound Money's own dashboard is now showing net issuance. The Gnosis co-founder is asking the question out loud. If you are a VP Engineering who signed off on holding ETH as an operational asset because of the burn thesis, this is the week that thesis needs a formal review.

The build-vs-buy calculus for rollup infrastructure also shifts. Six months ago, running your own L2 or paying a rollup-as-a-service vendor made sense partly because L1 was economically prohibitive for high-frequency use cases. At 1.9 gwei median, a lot of applications that migrated to Arbitrum, Base, or a Taiko-style based rollup for cost reasons could technically move back. They won't, because the operational tooling has followed them, but the pricing use in vendor contracts has shifted. Renegotiate. Your rollup provider's cost basis dropped. Yours should too.

What's Priced In for Crypto and DeFi

The market already understood that Dencun would compress L2 costs. That was in every research note in Q1. What is not fully priced in, and what the engineering community in particular has been slow to internalize, is the second-order effect on validator economics and therefore on the security budget of the chain everything else depends on.

DeFi protocols with ETH-denominated liabilities (lending markets, LSTs, restaking layers) built their risk models on a specific range of staking yield. If issuance keeps outpacing burn and the community responds by cutting issuance to defend the supply curve, staking yields compress. That flows directly into liquid staking token spreads, into EigenLayer-style restaking economics, and into any protocol paying incentives in stETH or equivalents. None of this is broken today. All of it is exposed if the trend from this reporting window persists into Q4.

The GC at any protocol touching staking derivatives should be asking this week whether the marketing materials and disclosures around expected yield still match the on-chain reality. If your product page still cites yield ranges from a burn-heavy regime, that is a documentation liability, not just a marketing one. Fix it before somebody else points it out.

What is genuinely surprising, and I do not think most desks have modeled it, is the political dimension. A chain whose L1 becomes a settlement layer for L2s that capture most of the user value is going to face internal governance pressure to change fee-sharing arrangements. Expect proposals. Expect debate. Expect the core dev discussions to increasingly reference L1 revenue as a first-class concern rather than a side effect.

Contrarian View

The consensus reading is that low L1 fees plus booming L2 activity equals a scaling win with a manageable revenue problem. I'd argue the opposite framing is worth taking seriously: this is exactly the outcome Ethereum's roadmap targeted, and the "problem" is a feature the market is misreading as a bug.

If you believe Ethereum's long-term role is settlement and data availability for a fleet of application-specific rollups, then L1 fee revenue was always going to compress relative to L2 activity. The value accrual mechanism in that world is not per-transaction gas, it is blob demand at scale. Right now blob usage is well below capacity, which is why blob fees are near zero and burn is anemic. As L2 activity grows from tens of millions to hundreds of millions of monthly transactions, blob pricing becomes the pressure valve. The current dip could be a transitional artifact, not a structural break.

The Köppelmann concern is real, but it is a concern about the next 12 to 18 months, not about the terminal state. Teams overreacting to this week's numbers by migrating away from Ethereum settlement could be selling the bottom of a J-curve.

Key Takeaways

• Median Ethereum gas at 1.9 gwei is a 98% drop from March's 83.1 gwei high and a 5-year low. This is the new baseline, not a temporary dip. Capacity planning models that assume mean reversion are wrong.
• Ether supply grew by roughly 13,400 ETH ($34.1 million) in a single week despite reduced usage. Any treasury policy or product disclosure built on the "ultrasound money" burn thesis needs a formal review this quarter.
• L2 volume (33 million) plus L1 volume (109 million) shows demand did not disappear. Dencun expanded supply of block space. This is architectural, not cyclical, and it changes vendor negotiation use for rollup-as-a-service contracts.
• A Gnosis co-founder publicly flagging L1 revenue and staking incentive financing is a governance signal. Expect fee-sharing and issuance-schedule proposals to move up the core dev agenda.
• Teams evaluating whether to settle on Ethereum L1, migrate to an L2, or run their own rollup should now be asking themselves a different question: not "which venue is cheapest today" but "which venue's economic model survives if blob demand takes 18 months to catch up with block space supply."