Share
Last week’s brief but historic fee inversion, when Polygon out-earned Ethereum in daily transaction fees for the first time, highlighted more than a fleeting anomaly. It offered a glimpse into where on-chain activity is migrating. On February 14, 2026, Polygon generated $407,121 in daily fees versus Ethereum’s $211,790. By February 15, the gap had narrowed to $303,923 against $285,470. The flip lasted only a couple of days, not weeks, but the forces behind it point to longer-term structural shifts.
Polymarket was the most visible catalyst. Matthias Seidl, co-founder of Ethereum analytics platform growthepie.com, confirmed in an X post on February 16, 2026, that the fee surge was “fully driven by Polymarket,” which alone generated over $1 million in Polygon fees over the preceding seven days.
More fees were paid on Polygon PoS than on Ethereum Mainnet 🤯
This is fully driven by Polymarket
-> see screenshots of app breakdown by fees paid https://t.co/EU5zDofxiO pic.twitter.com/toqo346VEs— matze | growthepie 🥧 (@web3_data) February 16, 2026
However, a broader structural trend was also unfolding. Payment-focused AI agents, producing high volumes of smaller ‘micro’ transactions, contributed significantly to Polygon’s spike. According to Dune Analytics, monthly payment transfers on Polygon nearly doubled from under one million to around two million between November 2025 and January 2026.
Polygon shows monthly growth in payment transfers, with the number of transfers reaching almost 2M in January.
I would like to note that just two months ago, this figure was less than 1M transfers.
The reason for this was the rapid use of payment AI agents on Polygon, which… pic.twitter.com/ThXsTwGQEw
— Alex (@obchakevich_) February 14, 2026
During the same period, Polygon also led all chains in x402 activity, an emerging standard for agentic micropayments, recording $1.2 million in weekly organic micropayments and 358,000 transactions. This is further corroborated by PIP-82, a Polygon governance proposal filed February 12, 2026, which confirmed Polygon had attracted 20.3% of all x402 transactions from the beginning of the year.
Last week in stablecoins on @0xPolygon was madness:
→ Polygon reached a new weekly high of 28M @USDC transactions, surpassing Solana’s 22M to become the most active USDC chain across the ecosystem (h/t @AlliumLabs)
→ USDC transfers also hit a weekly record 103M, giving… pic.twitter.com/lUZ8vohyzx
— Peter (📖, ✍️, 🔑) (@petertherock) February 16, 2026
As autonomous agents increasingly route high-frequency, low-value transactions to cheaper chains, Polygon’s low-cost infrastructure positions it as a structural beneficiary.
At the same time, market conditions reinforce the divergence. As of February 20, 2026, ETH trades near $1,940 on $19 billion in daily volume, down over 35% year-to-date (at the time of writing). In contrast, POL trades at $0.104, with a $1.12 billion market cap, up roughly 22% over the past two weeks at the time of writing.
ETH’s sustained price decline in early 2026 has reduced USD-denominated fee revenue, while on-chain DeFi activity has continued to expand. Ethereum DeFi TVL rose from around 23 million in January 2026 to over 27 million in the final week of February 2026, despite ETH’s weaker price performance. This dynamic reflects a growing structural decoupling between ETH price and network utilization, a trend that deepened throughout 2025 and has carried into 2026. Notably, beyond ETH’s price decline, the fee revenue impact also came from protocol upgrades such as Dencun, Pectra, and Fusaka, which reduced L2 fees by more than 90%.
On the other hand, Polygon’s recent fee revenue spike, despite trading 91% below its $1.29 all-time high, underscores how on-chain activity and token price can diverge entirely.
Ethereum’s protocol-level response to fee compression is the upcoming Glamsterdam upgrade. In its February 18, 2026, Protocol Priorities Update, the Ethereum Foundation confirmed the upgrade is targeted for the first half of 2026. The upgrade aims to increase the gas limit from 60 million toward and beyond 100 million, introduce parallel execution via Block-level Access Lists (EIP-7928), and enshrine proposer-builder separation (EIP-7732) directly into the protocol. This would reduce reliance on external MEV relays while strengthening censorship resistance. Vitalik Buterin summarized the direction in an X post in January, stating, “2026 will be the year we reclaim lost ground in self-sovereignty and trustlessness.”
2026 is the year that we take back lost ground in terms of self-sovereignty and trustlessness.
Some of what this practically means:
Full nodes: thanks to ZK-EVM and BAL, it will once again become easier to locally run a node and verify the Ethereum chain on your own computer.…
— vitalik.eth (@VitalikButerin) January 16, 2026
Greater execution capacity combined with lower per-unit fees could expand Ethereum’s fee base through higher transaction volume rather than higher costs. However, Glamsterdam does not eliminate the agentic web’s natural pull toward cheaper chains. Ethereum’s fee leadership will increasingly depend on use cases that require its base-layer security, i.e., high-value settlement, complex DeFi, and applications where trust minimization is critical, rather than those that remain by default.
February’s fee inversion was short-lived, but its implications are not. Even temporary displacement signals a deeper shift; the composition of on-chain activity is evolving, and not all of it will return to Ethereum’s L1 as capacity expands.
Share
