Mainnet fees dropped 7% last quarter. Layer 2 activity surged 37%. If the data looks familiar, it should. IBM just warned the market with the exact same pattern: hardware demand is eating software budgets. Ethereum is not a tech giant—it is a multi-layered economy. And that economy is now in a zero-sum war between its base layer and its scaling layers.
Reversing the stack to find the original intent.
The parallel is precise. IBM’s revenue warning came from a structural shift: enterprises rushed to buy AI infrastructure (servers, storage), starving their software and consulting arms. The same dynamic is playing out on Ethereum. Solvent L2s and AI-driven compute networks (EigenLayer, Arbitrum Orbit) are drawing activity—and fee revenue—away from the base layer. The narrative is that this is healthy scaling. But the numbers tell a different story: the base layer, the security backbone, is being financially cannibalized by its own expansion.
Context: The Protocol as a Conglomerate
Ethereum is not a single product. It is a stack: consensus (L1), execution (L2), data availability (Blobs), and application (dApps). Each layer competes for a finite pool of user attention and transaction value. In Q2 2025, total ETH burned through fees fell. According to Ultrasound.money, daily base fee burns dropped nearly 10% week-over-week in late June. Meanwhile, L2 transaction count hit all-time highs, growing 37% year-over-year. The same pattern IBM reported: hardware revenue (L1 fee revenue) down 7%, distributed infrastructure (L2 activity) up 37%.
Core Analysis: The Zero-Sum Fee Market
Scaling was supposed to be a rising tide. Instead, it has become a siphon. Every transaction that moves to an L2 reduces the pressure on L1 blockspace. That reduces base fees, which reduces ETH burn, which reduces staking rewards (through lower issuance offset). The network still pays stakers in inflation, but the economic security budget—the actual fee income stakers expect—is being squeezed.
Let the data speak. At the end of 2024, L2s accounted for roughly 60% of all transactions on the Ethereum ecosystem. In Q2 2025, that number crossed 75%. The L1’s share of fee revenue dropped from ~40% to ~30% in the same period. That is a direct wealth transfer from the security layer to the execution layer.
Now, consider the quality of that security. Ethereum’s security is funded by two streams: new issuance (inflation) and transaction fees. Issuance is fixed per epoch; it does not adjust with usage. Fees are variable and sensitive to demand. When demand shifts to L2s, the variable component shrinks. Stakers rely on that variable income to offset the risk of capital lockup. If fee income drops below the opportunity cost (e.g., DeFi yields or bond yields), rational stakers will exit. Reduced stake means reduced security—a lower economic cost to attack the network.
This is not a hypothetical future. It is happening now. The 7% drop in mainnet fees is not noise; it is a leading indicator. Combined with the 37% surge in L2 activity, it signals a structural reallocation. IBM’s CEO explicitly stated that hardware demand “squeezed out” software spending. The same language applies here: L2 activity is squeezing L1 fee budgets.
Contrarian: The Blind Spot of Complexity
The market celebrates L2 growth. Analysts point to total activity across rollups as a sign of Ethereum’s dominance. But this ignores a fundamental trade-off: the L1 is the sovereign finality layer. Every L2 depends on it for settlement and security. If the L1 becomes underfunded, the entire stack becomes brittle.
Here is the counter-intuitive angle: L2s are not saviors; they are competitors. They compete for the same user transactions, the same developer mindshare, and the same value. The Ethereum Foundation promotes L2s as an overflow valve, but valves can also reduce pressure on the main engine. When an L2 like Base or zkSync launches its own native token for fees, it further fragments the ecosystem’s economic unity. ETH’s role as the native gas token is being eroded.
IBM’s Red Hat grew 11%—healthy. But the rest of the software business only grew 5%. The growth story masked the stagnation. Similarly, L2 growth at 37% masks the base layer’s stagnation. The most optimistic projections for Ethereum’s fee revenue ignore that a large portion of value is now being captured by L2 tokens, not ETH. The network’s security budget is being hollowed out.
Takeaway: The Vulnerability Forecast
If L1 fee revenue continues to shrink relative to L2 activity, staking yields will decline. Lower yields will cause stake to outflow. Lower stake will reduce security. At some point, the cost to attack Ethereum drops below $1 billion. That is the line where institutional confidence wavers. The next 12 months will test whether the ecosystem can rebalance—perhaps through forced L2 fee sharing, or a baseline fee on L2 withdrawals. Otherwise, the network risks an abstraction leak: the layers above are dependent on a layer below that is growing weaker by the day.
