L2 Fees Are Almost Free. Your Remittance Isn’t.
Rollup fees have fallen more than 99% and the next cut is stuck in a testnet audit. A forecast to 2028 on why sub-cent L2 fees still will not make your cross-border transfer cheap.
Sub-Cent Rollups, Six-Percent Remittances
The cheapest way to move a dollar on Ethereum has never been cheaper. A token swap on Base cost about $0.0034 in early November 2026, a transfer on Optimism settled for well under a hundredth of a cent, and the average fee across the three largest rollups has fallen more than 99% in two years, from roughly $0.18 in the first quarter of 2024 to about $0.0015 by early 2026, according to Token Terminal and the per-transaction feeds at growthepie. If the only thing standing between a worker in Texas and her family in Michoacan were a blockchain fee, that corridor would already be closed.
It is not closed. Sending $200 across a border still costs around 6% on average, per the World Bank’s Remittance Prices Worldwide database, and the busiest money-transfer route on the planet, the United States to Mexico, moved roughly $65 billion a year at prices that have barely budged, hitting a record near $64.7 billion in 2024 by the published bilateral totals. The distance between what a rollup charges and what a remittance charges is the single most important number in this forecast, because it exposes what the fee-compression story keeps getting wrong: the blockchain was never the expensive part.
This is a forecast, not a victory lap. Anyone who followed the earlier entries in our fee-compression series, from how low rollup fees could go to the fight over who keeps the sequencer margin, knows the mechanical story is nearly finished. The data-availability lever that gutted rollup costs after 2024 has hit a hard floor, and the execution upgrade meant to cut them again is parked in a testnet audit queue, as the sections below lay out. So 2026 is the year to ask the question the hype skips: if the on-chain rail is now effectively free, does that finally let stablecoins beat the banks at sending money home? The short answer this piece defends is yes, but only in specific corridors, and the win is decided almost entirely off-chain.
A sub-cent fee is a real engineering achievement. It is also, for the average remittance sender, close to irrelevant. What follows traces where the money actually goes, corridor by corridor, between now and 2028, and why the last leg of the journey, the one that touches cash, is the only part the compression wave has not reached.
How Low L2 Fees Actually Went
Numbers first. According to growthepie, which tracks per-transaction costs across rollups, the median token swap in early November 2026 cost about $0.0034 on Base, roughly $0.0105 on Arbitrum One, and less than $0.0001 on OP Mainnet. Simple transfers sit in a similar band: a few cents at most, with Base near $0.02, Optimism near $0.03, and Arbitrum near $0.04 on a typical day. Stretch the window and the trend is brutal for anyone who built a business on charging for blockspace. Token Terminal pegs the average fee across the three biggest rollups down 99.16% over two years, from $0.180219 in the first quarter of 2024 to $0.001512 in the first quarter of 2026.
That collapse has one cause and one limit. The cause was Dencun, the March 2024 upgrade that gave rollups a dedicated data lane called blobs and cut their largest expense, posting transaction data back to Ethereum, by more than 90%. The Fusaka upgrade in December 2025 and its follow-on blob-parameter forks raised blob capacity again through early 2026. More blobs, cheaper data, lower fees. On-chain foreign exchange between stablecoins rides exactly the same rails: the automated market makers that price a USDC-to-local-stablecoin swap now clear for a rounding error.
Liquidity followed the cheap fees. Arbitrum One holds about $13.8 billion in value locked and Base about $11.2 billion, together more than half the rollup total, per L2BEAT. For payments the takeaway is simpler than the usual chain beauty contest: the rail is no longer a cost center. A company can move a dollar of USDC from one rollup address to another for less than the fee a card network charges to authorize a coffee.
| Network | Median swap fee (Nov 2026) | Median transfer | Value locked |
|---|---|---|---|
| Base | ~$0.0034 | ~$0.02 | ~$11.2B |
| Arbitrum One | ~$0.0105 | ~$0.04 | ~$13.8B |
| OP Mainnet | <$0.0001 | ~$0.03 | n/a |
The Floor Is Real, and the Next Cut Is Stuck
Rollup fees cannot keep falling forever, and Ethereum wrote that limit into its own rules. A change shipped with Fusaka, EIP-7918, sets a floor under the blob base fee by anchoring it to the cost of ordinary execution on Ethereum, roughly one sixteenth of the L1 base fee. Before that floor, blob fees spent long stretches at a vanishing 0.000000001 Gwei. Afterward they jumped. The researcher known as Kydo captured the shift bluntly, noting the blob base fee had gone up 15,000,000x since Fusaka. That is not a bug; it is the network refusing to give data away for free, much as miners’ power bills quietly set a cost floor under Bitcoin that no model can wish away.
With the data lever pinned, the next leg of compression was supposed to come from execution, the part of a transaction that runs code rather than posts data. That is what the Glamsterdam upgrade is for. It is not arriving on schedule. The upgrade cleared the Sepolia testnet cleanly on 6 October 2026, but its second testnet step, the Hoodi fork, was pulled before its 26 October slot and remains frozen. On 6 November, Ethereum Foundation developer Marius van der Wijden confirmed the fork stays on hold while teams work through audit findings. The auditors at ChainSecurity had flagged that two gas-accounting proposals, EIP-8037 and EIP-8038, could change how contracts behave, warning that even contracts written before the upgrade can respond differently under the new rules. No Hoodi date and no mainnet date have been set.
Read those two facts together and the forecast narrows fast. The data-availability motor is capped by design, and the execution motor is stuck in review. For the horizon that matters to a payments decision, the next 12 to 24 months, L2 fees are structurally near their floor. The on-chain leg of a transfer is not going to get materially cheaper on any near-term calendar, which means the question of whether stablecoins win payments no longer hinges on gas at all. It moves to the parts of the journey Ethereum does not control.
The Blockchain Was Never the Expensive Part
The most useful document published on this subject in 2026 did not come from a crypto company. It came from a central bank. In July, the Bank of Italy ran what it called a mystery-shopping exercise: researchers sent 200 real USDC transfers across ten corridors, routing money from Italy to Argentina, Brazil, South Africa, the United Arab Emirates, and Japan, then measured the full cost of getting value from one person’s pocket to another’s. CoinDesk summarized the result that the industry would rather ignore: stablecoins showed no systematic cost advantage over conventional money transfer operators once the whole payment chain was counted.
The detail inside the study is what makes it devastating for the fee-compression thesis. Total end-to-end costs ranged from 0.3% to nearly 9% of the amount sent. The blockchain leg itself averaged about 0.4%. Everything else, the exchange charges, the currency conversion, the local banking steps, drove the bill. In the Bank of Italy’s own words, the on-ramp and off-ramp frictions are the main source of cost and transfer duration. Three corridors beat a conventional operator like Wise; four cost more. Where an instant domestic payment system existed on the receiving end, money arrived in under 20 minutes; where it did not, conventional bank steps still took one to two business days.
Now apply that arithmetic to the headline. Suppose Glamsterdam eventually ships and cuts the on-chain leg from 0.4% to 0.04%. The total cost of a transfer that ran 7% falls to roughly 6.96%. The sender does not notice. The engineering that crypto Twitter celebrates is optimizing the one component of the journey that was already close to free, while the components that actually hurt, fiat conversion and cash-out, sit untouched. Fee compression is real, and for remittances it is almost beside the point.
Anatomy of a Cross-Border Transfer
To see why the rail barely matters, follow a single $200 transfer from a sender in the United States to a recipient in a developing-market corridor. The money passes through five distinct steps, and only one of them runs on a rollup.
First comes the on-ramp: turning dollars from a bank account or debit card into a stablecoin. That conversion happens at an exchange or a licensed issuer, and it carries a fee plus, often, a quiet spread. The venue matters more than most senders realize, which is why the economics of the big exchanges feed straight into remittance costs. Second is the currency spread: a dollar stablecoin still has to become local purchasing power, and the price of that swap, whether through a local stable token or a withdrawal desk, is where the real margin hides. Third is the on-chain hop itself, the sub-cent, sub-half-percent step this whole industry has spent two years optimizing. Fourth is the off-ramp: converting the stablecoin into spendable local currency, the part that depends on local liquidity, agent networks, and bank integrations. Fifth is compliance, the know-your-customer and limit checks that add time and sometimes cost at both ends.
Lay the five steps side by side and the shape of the problem is obvious. The step crypto made free is the small one in the middle. The steps that dominate the total, the on-ramp spread and the off-ramp cash-out, are fiat-world functions that no Ethereum upgrade touches. They compress only when competition, licensing, and local payment infrastructure improve, and those move on a regulatory clock, not a hard-fork calendar.
| Step in the journey | What it does | Typical cost | Does L2 compression help? |
|---|---|---|---|
| On-ramp (fiat to stablecoin) | Exchange or issuer converts dollars to USDC | 0.1% to 1.5% plus spread | No |
| Currency spread (USD to local) | Dollar value becomes local purchasing power | 0.5% to 4% | No |
| On-chain transfer | Stablecoin moves across a rollup | Under 0.4%, often sub-cent | Yes (already floored) |
| Off-ramp (stablecoin to cash) | Recipient converts to spendable local money | 2% to 5% | No |
| Compliance and limits | KYC, caps, reporting at both ends | Time, occasional fees | No |
The Corridors Where Stablecoins Actually Win
None of this means stablecoins lose. It means they win selectively, and the selection rule is legible once you stop looking at gas. Stablecoins beat the incumbents where three conditions stack: the existing service is expensive and poorly contested, the receiving country has real demand to hold dollars, and the local instant-payment rail is weak or missing. Sub-Saharan African corridors, where the World Bank still records average costs above 8% and the UN’s own target of under 3% by 2030 looks distant, are the clearest case. So are high-inflation economies like Argentina, where holding a dollar stablecoin is not a transfer feature but the entire point, because the alternative is a currency that loses value by the week.
They win less, or not at all, where the fiat system already works. Inside the eurozone, instant euro transfers now settle in seconds at no extra cost, and a dollar stablecoin that has to round-trip through two conversions simply adds steps. The same logic increasingly applies to United States domestic transfers, where instant bank rails have closed much of the speed gap. The Bank of Italy data makes the point empirically: savings were highly corridor-specific, with some routes beating a conventional operator and others losing to it. A forecast that says stablecoins will take over payments is wrong. A forecast that says they will capture the expensive, dollar-hungry, badly-served corridors first is defensible, and that is where the volume is heading.
The United States to Mexico corridor, the largest in the world at roughly $65 billion a year, sits in the middle and is therefore the one to watch. Mexico runs a capable instant-payment system, so the rail-speed advantage is muted, but dollar demand is strong and the sheer size of the flow makes even a one-point cost improvement worth billions. Whoever solves the Mexican peso off-ramp at scale, in cash and in bank deposits, captures the defining prize of this decade in payments, and they will do it with licensing and liquidity, not with a cheaper blob.
The United States Just Taxed the Competition
A policy change that landed on 1 January 2026 reshaped the incentive map, and most crypto coverage missed it. The One Big Beautiful Bill Act created a new federal excise tax on remittance transfers, and the Treasury and the IRS issued proposed regulations spelling out how it works. The tax is 1% of the transfer, and crucially it applies to transfers funded with cash, a money order, a cashier’s check, or a similar physical instrument. Transfers funded from an account at a qualifying United States financial institution can fall under an exception.
Read the structure, not just the rate. The tax, as drafted, bites hardest on exactly the cash-in, cash-out model that defines traditional money-transfer storefronts, and it leaves account-funded digital transfers a path to the exception. That is a quiet tailwind for any rail that starts from a bank account or a regulated wallet rather than a stack of twenties handed across a counter. Whether a stablecoin-funded transfer counts as cash or as an account transfer is precisely the kind of question the IRS guidance still has to resolve, so no one should overstate the edge. But the direction is clear: Washington has put a thumb on the scale against physical-cash remittances, and the sender most likely to migrate is the one already comfortable funding a transfer digitally.
For the forecast, that matters because policy, not protocol, is doing the work. The event that could meaningfully shift corridor economics in 2026 was a tax bill, not a hard fork. It is a reminder that the binding levers in payments sit with legislators and regulators, and the crypto industry’s instinct to treat every problem as an engineering problem keeps aiming at the wrong target.
The Incumbents Are Buying the Off-Ramp, Not Fighting It
The clearest signal that the rail has been commoditized is what the incumbents are doing with their money. They are not defending the old pipes. They are buying the stablecoin stack and bolting it onto the one asset crypto cannot easily replicate: the fiat endpoints. Western Union, a company whose entire history is cash across a counter, launched its own dollar stablecoin, USDPT, built on Solana and issued by Anchorage Digital Bank, in May 2026. It plans to use the coin for treasury and agent settlement and to pair it with a consumer product, Stable by Western Union, across more than 40 countries, all while keeping the 200-country agent network that lets a recipient walk out with physical cash. The rail got modernized; the cash-out moat stayed exactly where it was.
Stripe made the same bet with a bigger checkbook. It paid $1.1 billion for the stablecoin infrastructure firm Bridge, its largest acquisition ever, then helped launch a payments-focused Layer 1 called Tempo, whose mainnet opened in March 2026. Stripe now accepts stablecoin payments worldwide, settles into a merchant’s balance in local currency, charges around 1.5% for the service, and is extending stablecoin-linked cards toward 100 countries. Owning issuance, processing, and settlement in one stack is the strategy, and the on-chain hop is the cheapest, least differentiated part of it.
Circle rounds out the picture from the issuer side. Its Circle Payments Network knits licensed institutions together for cross-border settlement in USDC, and it has tied into Visa Direct so clients can prefund payouts with the stablecoin in near real time. Chief executive Jeremy Allaire has framed the shift plainly, arguing that stablecoins are moving beyond crypto trading into the infrastructure for working capital, collateral, settlement, and everyday payments, a view he repeated as Circle pushed into dozens of emerging markets. When the issuer, the card network, the money-transfer giant, and the payments processor all converge on the same model, the rail is settled business. The fight has moved to the endpoints.
GENIUS, the OCC, and the SEC Line
The United States spent 2026 building the legal scaffolding that will decide who controls those endpoints. The GENIUS Act, the federal law for payment stablecoins, moved from statute to working rulebook when the Office of the Comptroller of the Currency issued its final implementing rule on 21 October 2026, setting standards for permitted payment stablecoin issuers covering reserves, redemption, risk management, audits, and custody. Treasury, acting for the Stablecoin Certification Review Committee, issued interim procedures for certifying state-level regimes. For the first time, a bank-grade path to issuing and holding dollar stablecoins exists in black and white.
The design also draws a clean line with the Securities and Exchange Commission. The GENIUS framework treats a compliant payment stablecoin as a payment instrument rather than a security, which keeps the core product out of the SEC’s lane and removes the jurisdictional uncertainty that stalled stablecoin adoption for years. That clarity is itself a competitive advantage for regulated issuers, though it does not extend to every design: yield-bearing or interest-passing variants remain a contested question the SEC still watches closely. The European Union reached a parallel destination by a different road, slotting stablecoins into its MiCA regime as e-money and asset-referenced tokens, so the global pattern is consistent: payment stablecoins are becoming regulated bank-adjacent products, not wildcat money.
That is the opposite of a neutral development for the fee-compression thesis. A regulated perimeter rewards whoever can operate inside it: banks, licensed issuers, and the payment giants already buying their way in. It does nothing to make a cheap blob more valuable. The clearer the rulebook gets, the more the advantage flows to balance sheets and licenses, the exact resources that control on-ramps and off-ramps, and the less any of it depends on the marginal cost of an Ethereum transaction.
What Fee Compression Changes, and What It Doesn’t
It would be a mistake to read all of this as crypto losing. Fee compression changed real things. It made on-chain settlement a solved problem, so a stablecoin can move between wallets or protocols for a rounding error. It opened a design space that simply did not exist at five-dollar gas: streaming payroll by the second, machine-to-machine payments between software agents, in-game item trades, and creator micropayments too small to have ever cleared a card network. For a crypto-and-gaming audience, that last category is the quietly important one. Sub-cent fees do not replace Western Union, but they make a one-cent tip or a fractional in-game purchase economically possible for the first time, and whole product categories are being built on exactly that.
What compression did not change is the list that governs remittances. It did not touch the currency spread, the off-ramp cash-out, the compliance overhead, or the quality of local payment rails. It did not remove the trust and user-experience hurdles that keep a first-time sender on a familiar app. And it did not erase the risk that value takes on when it hops between chains, where every extra bridge is a fresh attack surface rather than a free convenience. Those frictions are off-chain, and they do not compress because Ethereum shipped an upgrade.
The strategic consequence for the next two years is blunt. Cheaper fees are no longer a differentiator, not among rollups and not as a payments pitch. Everyone has them. The variables that will decide who wins corridors are liquidity depth at the off-ramp, breadth of licensed coverage, and an experience a non-crypto user will tolerate. A forecast built on fees going lower is forecasting a variable that has stopped mattering.
The Value-Accrual Bill Comes Due
There is a cost to all this cheapness, and Ethereum the asset is paying it. The same shift that made rollup fees disappear, pushing activity off the main chain and onto subsidized blob space, also starved the mechanism that once made ETH scarce. With less execution happening on L1 and blob fees sitting near their floor, less ETH gets burned, and the deflation that holders treated as a law of nature has broken. Fidelity Digital Assets put concrete numbers on the drain: Base paid only about $5.2 million in blob fees over a year against roughly $94 million in user revenue, so the rollup keeps the economics while the base layer collects scraps, and the firm estimated the EIP-7918 floor, had it existed since the blob era began, would have returned about $78.6 million more to ETH holders.
The market has noticed. ETH changed hands near $1,635 on 21 November 2026, down about 13% on the week and roughly 67% below its all-time high around $4,946, according to CoinGecko, caught in the same broad risk-off move that has dragged the whole market into the December macro reckoning. The money supply that payments depend on softened with it: the total stablecoin market capitalization slipped to about $281 billion by 20 November, shedding more than $16 billion in a month, per DefiLlama. Even as the payment integrations accelerate, the on-chain float contracted when the macro turned, a reminder that stablecoin supply is cyclical, not a one-way escalator.
For the forecast, the lesson is to keep two stories separate. Stablecoins winning specific corridors and ETH appreciating are different bets, and 2026 showed they can move in opposite directions. Cheap rails can carry more payment volume while the token that secures them underperforms. Anyone pricing the payments opportunity off the ETH chart is reading the wrong instrument.
The Forecast: Three Scenarios to 2028
Put the pieces together and the forecast resolves into three paths, separated not by where fees go, because fees are already floored, but by whether the off-chain last mile improves. The base case is the most likely and the least dramatic. On-chain fees stay near zero, Glamsterdam eventually ships in 2027 without transforming the picture, and off-ramp costs grind down slowly as licensing and competition spread. Stablecoins keep taking the expensive, dollar-hungry corridors they already win, global blended remittance cost drifts down toward the mid-single digits, and the UN’s sub-3% target stays out of reach for most routes. Steady, corridor-by-corridor gains, no payments revolution.
The bull case needs the endpoints to crack, not the chain. A major network, some combination of Circle, Visa, Stripe, and the remittance incumbents, solves the Mexican peso and Latin American cash-out at scale, the GENIUS rails pull bank issuers in, and the 1% remittance tax nudges meaningful cash volume onto digital rails. In that world a flagship corridor finally crosses the 3% line, stablecoin remittance volume rises several-fold, and the off-ramp stops being the bottleneck. The bear case is its mirror: Glamsterdam slips deeper into 2027 and stays messy, the macro drawdown keeps the stablecoin float shrinking, IRS guidance treats crypto-funded transfers as cash with no exception, and sticky off-ramp costs leave adoption stranded in today’s niches. In the bear case, free fees change nothing at all, which is the whole point of this forecast in miniature.
| Scenario (to 2028) | On-chain fee | Stablecoin remittance reach | Blended corridor cost | What has to happen |
|---|---|---|---|---|
| Base | Near zero, floored | Keeps high-cost, dollar-demand corridors | Drifts toward mid-single digits | Off-ramp competition grinds costs down slowly |
| Bull | Near zero, floored | Several-fold volume; a flagship corridor under 3% | Under 3% in lead corridors | Endpoints crack: licensing, liquidity, bank issuance scale |
| Bear | Near zero, floored | Stalls in today’s niches | Stays above 6% | Off-ramp stays sticky; macro and tax guidance bite |
What to Watch Into 2027
Because the deciding variables are off-chain, the signals worth tracking are not the ones a block explorer shows. Here is the short list that will tell you which scenario is unfolding.
- Glamsterdam’s revised testnet schedule and a locked mainnet activation block: whether the execution upgrade ships in 2027 or slips again under the weight of its gas-accounting audit.
- Final IRS guidance on whether stablecoin-funded transfers qualify for the account exception to the 1% remittance tax, which decides if digital rails get a real policy tailwind.
- Off-ramp pricing in Latin America: whether Circle, Visa, Stripe, or a remittance incumbent cracks the Mexican peso cash-out below roughly 2% to 3%.
- A single corridor crossing the UN’s 3% cost line, the genuine finish line, rather than another sub-cent fee headline.
- Stablecoin float recovery as the macro cycle turns, and whether GENIUS-era bank issuers add new regulated supply instead of just reshuffling it.
- Any L2 that tries to compete on something other than price, since price has stopped being a differentiator and the next edge has to come from UX, off-ramp partnerships, or compliance.
The fee war is effectively over, and it turned out to be the easy part. Rollups did the hard engineering, drove the cost of moving a dollar on-chain to almost nothing, and in doing so proved that the on-chain leg was never what made sending money home expensive. The war that remains is the last mile, the fiat endpoints where dollars become pesos and pesos become cash, and it will be won with licenses, liquidity, and local rails rather than cheaper blocks. Watch the off-ramp, not the gas meter. That is where the next three years of this story will actually be decided.
Frequently Asked Questions
Why are Ethereum L2 fees so cheap in 2026?
Two upgrades did it. Dencun in March 2024 gave rollups a dedicated data lane called blobs and cut their main cost by more than 90%, and Fusaka in December 2025 raised blob capacity again. The average fee across the three largest rollups fell about 99% in two years to a fraction of a cent. A change called EIP-7918 now sets a floor under blob fees, so they cannot fall all the way to zero.
Do cheap L2 fees make crypto remittances cheaper than banks?
Usually not by much. A 2026 Bank of Italy study that sent 200 real USDC transfers across ten corridors found the blockchain leg averaged about 0.4% of the cost, while the total ran from 0.3% to nearly 9%. The expense sits in the fiat on-ramp, the currency spread, and the off-ramp cash-out, none of which a cheaper blockchain fee changes.
What is the US remittance tax in 2026?
The One Big Beautiful Bill Act created a 1% federal excise tax on remittance transfers, effective for transfers from 1 January 2026. It applies to transfers funded with cash, a money order, or a similar physical instrument, and transfers funded from an account at a qualifying US financial institution can qualify for an exception. How it treats stablecoin-funded transfers is still being clarified in IRS guidance.
Which corridors benefit most from stablecoin payments?
Stablecoins help most where the existing service is expensive, the receiving country has strong demand to hold dollars, and local instant-payment rails are weak. High-cost routes in Sub-Saharan Africa and high-inflation economies like Argentina are the clearest cases. They help least inside the eurozone and for US domestic transfers, where instant bank rails already settle in seconds.
When will Ethereum’s Glamsterdam upgrade go live?
Glamsterdam activated on the Sepolia testnet on 6 October 2026, but its next step, the Hoodi fork, was pulled in October and remained on hold as of 6 November after auditors flagged gas-accounting risks to existing contracts. No mainnet date has been set, and a 2027 activation is now the realistic expectation.
By Priya Reddy, senior markets editor at HOGE Wire, covering Ethereum scaling, stablecoins, and the economics of on-chain payments.