On May 15, 2025, Ethereum’s average gas price hit 150 gwei, not a historical outlier but a steady climb over the last six weeks. Daily active addresses hovered at 600,000, near all-time highs. Conventional wisdom says high fees kill adoption—Layer 2 solutions should drain usage. Yet the data stubbornly refuses to confirm that narrative.
I’ve spent a decade dissecting protocol-level metrics, and this pattern echoes something I saw in traditional markets: Apple’s pricing power during inflation. Just as consumers kept buying iPhones at higher prices due to brand lock-in, Ethereum users continue transacting on L1 despite cheaper alternatives. The question isn’t whether fees are high—it’s why demand remains inelastic.
Context: The Fee Market Architecture
Ethereum’s fee mechanism, governed by EIP-1559, creates a base fee that adjusts algorithmically based on network congestion. Users can also add a priority fee to incentivize miners (or validators post-merge). The base fee is burned, reducing supply and creating deflationary pressure during high activity. This design was meant to make fees predictable, but in practice, it only guarantees that demand shocks translate directly into cost spikes.
Layer 2 solutions—Optimistic Rollups and ZK-Rollups—offer transaction costs that are often 10–100x lower. Arbitrum’s median tx fee in May was $0.15; Optimism was $0.09. Meanwhile, a simple ETH transfer on L1 costs $3–$5, and a Uniswap swap can hit $20–$50. Rational users should migrate. Yet the L1 demand curve hasn’t flattened.
Core: The Inelastic User Cohorts
I ran a script over the past 12 months of on-chain data (January–May 2025) filtering by transaction type, value, and gas paid. Three clear cohorts emerged where users consistently choose L1 over L2:
1. Whale DeFi Interactions – Transactions involving >$1M in value overwhelmingly stay on L1. The reason is twofold: liquidity depth and composability. A whale executing an arb on Uniswap V3 needs to interact with multiple pools in a single block. L2s introduce latency and fragmentation. The gas cost ($200–$500) is negligible relative to the capital at stake. These users are not price-sensitive—they’re speed- and safety-sensitive.
2. NFT Mints and Rare Auctions – High-profile mints (e.g., from Bored Ape Yacht Club derivatives or new blue chips) still happen on L1. The social proof of minting on the “main chain” carries weight, and floor prices on L2s lag. In one case, a mint on Arbitrum sold out in 10 minutes with a 2 ETH floor, but the same collection on L1 traded at 4 ETH. The premium is irrational but real.
3. MEV Searchers – Maximal Extractable Value bots operate primarily on L1 due to the complexity of mempool analysis and bundle submission. L2 centralised sequencers reduce MEV opportunities, but the largest searchers still prefer L1 where they can front-run opaque order flows. Gas fees are just an operational cost.
These cohorts mirror Apple’s high-end user base: they’re not buying a phone; they’re buying into an ecosystem with deep integration. Ethereum L1 is the “luxury tier” where brand trust, security, and composability command a premium.
Check the math, not the roadmap.
The roadmap promised L2s would decongest L1. Instead, L1 has become a settlement layer for high-value txs, while L2s handle everyday transfers. The total fee revenue on L1 in Q1 2025 was $1.2B—higher than any quarter in 2024 despite lower txs. The burn mechanism turned high fees into a feature, not a bug. But this introduces a structural vulnerability.
Contrarian: The Fragility of a K-Shaped Fee Market
Just as the Apple analysis revealed a K-shaped consumer market—where premium and discount segments diverge—Ethereum is experiencing a K-shaped usage bifurcation. High-value users stay on L1; retail moves to L2s. This seems sustainable in a bull market, but the asymmetry creates a single point of failure.
If a systemic shock hits L1—a smart contract exploit, a validator collusion attack, or a governance crisis—the high-value users have no immediate escape. L2s rely on L1 for finality. A disruption in L1 settlement would cascade, freezing billions in DeFi positions. The market is pricing L1 as a safe haven, but that safety is an illusion if the top 10% of txs (by value) control 80% of the fee revenue. Those txs are the minority of activity but the majority of economic value. Lose them, and the fee burn drops, undermining Ethereum’s monetary premium.
“Complexity is the enemy of security.”
The very composability that whales love also means that a single vulnerability in a high-value contract (e.g., a new lending protocol) can drain liquidity that underpins hundreds of other protocols. Audits are snapshots, not guarantees. I’ve personally uncovered edge cases in DeFi protocols that were missed by three independent audits. The current fee structure rewards complexity because it enables unique arbitrage, but each new interaction layer adds risk.
Takeaway: A Vulnerable Equilibrium
The Ethereum L1 fee market is not broken—it’s functioning exactly as designed for a subset of users. But this equilibrium is fragile. It relies on a continuous inflow of high-value txs to justify the premium. If the narrative shifts—say, a major exploit on L1 that makes users question security—the premium could evaporate overnight. The L2 exodus would not be gradual; it would be a cliff.
Will a black swan event expose the loneliness of the premium segment? The math says yes. The timing is uncertain, but the invariant is clear: a fee market driven by a small, inelastic cohort is a ticking bomb. Check the math, not the roadmap.
