Fusaka Nine Months Later: Did ETH Fees Really Drop?

Quick answer: Ethereum’s Fusaka upgrade went live on December 3, 2025, and finished its phased rollout on January 7, 2026, when blob capacity settled at a target of 14 and a maximum of 21 per block — up from 6 and 9 before the upgrade. Nine months later, Layer-2 transaction costs remain extremely low: Base processed roughly 292 million user operations in the 30 days ending September 7, 2026, while paying Ethereum mainnet only about $8,800 in data-posting fees for the entire month. But most of that fee collapse actually traces back to March 2024’s Dencun upgrade, not Fusaka. What Fusaka mainly delivered was headroom — capacity for L2 activity to keep growing without fee spikes — plus a new mechanism, EIP-7918, designed to stop the blob fee market from collapsing to zero and starving Ethereum’s base-layer fee revenue. The unglamorous truth: L2 users got cheaper, more reliable transactions, but ETH itself picked up a genuine, still-unresolved question about where its value actually accrues.

Nine Months Later: Why This Retrospective Matters

A lot of crypto coverage lives and dies on launch day. Fusaka got its wave of “everything changes now” headlines back in early December, and then, like most protocol upgrades, it faded from the news cycle. That’s normal. It’s also exactly why a check-in matters more than the original announcement did.

Launch-day pieces are built on promises and testnet numbers. What we can do now, nine months in, is look at what actually happened on mainnet — real blob counts, real L2 fee data, real ETH issuance figures — and separate what Fusaka delivered from what got credited to it by mistake. Some of that credit, it turns out, belongs to an earlier upgrade entirely.

What Fusaka Actually Is, in Plain Terms

Fusaka is Ethereum’s most recent major protocol upgrade, named (per Ethereum’s now-standard convention) by combining a Devcon city with a star. It bundles several EIPs, but the headline feature by far is PeerDAS — “peer data availability sampling,” formally EIP-7594.

To understand why that matters, you need one piece of background: Ethereum’s Layer-2 networks (Arbitrum, Base, Optimism, and dozens of others) don’t process most of their transactions on Ethereum’s main chain anymore. They batch transactions off-chain and then post a compressed summary back to Ethereum as proof that the work happened correctly. Since March 2024’s Dencun upgrade, that summary gets posted in a data structure called a “blob” — a temporary storage slot that holds roughly 128 kilobytes of data and gets deleted after about 18 days, once it’s no longer needed for verification.

Blobs are cheap by design. But there’s a ceiling on how many blobs fit in each Ethereum block, and that ceiling determines how much L2 activity the network can absorb before blob space gets scarce and fees start climbing again. Before Fusaka, that ceiling sat at a target of 6 blobs per block, with a hard maximum of 9. Fusaka’s entire point was raising that ceiling — safely, in stages — without asking every Ethereum node to shoulder a heavier bandwidth burden.

PeerDAS, Explained Without the Jargon

Here’s the problem PeerDAS solves. If Ethereum just cranked blob capacity up by 3x or 8x overnight, every node on the network would need to download and store 3x or 8x more data to verify it. That’s fine for a well-funded data center. It’s a real problem for someone running a node on a home internet connection, and Ethereum’s decentralization depends on those smaller operators staying in the game.

PeerDAS gets around this with a trick borrowed from error-correction math. Blob data gets mathematically “extended” using something called erasure coding — think of it like adding redundant checksums to a file so it can be reconstructed even if pieces go missing. That extended data gets split into 128 pieces, called columns. Instead of downloading everything, a typical node only needs to grab a handful of those columns — roughly 8 out of 128 — and cryptographically verify that its sample matches the whole. If enough nodes across the network are each sampling a different slice and reporting no problems, the entire dataset is provably available, even though no single node stored all of it.

It’s a bit like a library checking whether a shipment of a thousand books arrived intact by having a thousand different volunteers each spot-check one random page, rather than asking one person to read every book cover to cover. Nobody does the full job alone, but collectively, the check is airtight. That’s the piece that let Ethereum scale blob capacity without quietly turning into a network only large institutions can fully validate.

The Actual Rollout Timeline

Fusaka activated on mainnet December 3, 2025. Rather than jumping straight to a new blob ceiling, Ethereum’s developers used a new mechanism called Blob-Parameter-Only (BPO) forks — smaller, faster-moving forks that adjust just the blob numbers, without the months of testing a full upgrade requires. The first one landed December 9, 2025, raising the target to 10 and the max to 15. The second and, so far, final one arrived January 7, 2026, pushing the target to 14 and the max to 21 — where it has sat for the past eight months.

That’s a meaningful jump: roughly 2.3 times the pre-Fusaka target, and Ethereum’s developers have discussed eventually pushing toward a much higher ceiling in future forks. But nothing beyond 21 has actually shipped on mainnet as of this writing, which is worth flagging, because some coverage conflates “the long-term roadmap number” with “what’s live today.”

What Actually Changed for Everyday Users

Here’s where the retrospective gets honest rather than celebratory. If you’ve read that Fusaka “cut L2 fees by 60% overnight,” treat that claim skeptically. The bulk of that fee collapse happened back in March 2024, when Dencun first introduced blobs and dropped typical Arbitrum transactions from around 37 cents to roughly a penny, and Optimism from about 32 cents to under a cent. Fusaka arrived 21 months later, into a world where blob fees were already close to negligible for most of the year, because blob demand wasn’t yet bumping against the old 6-9 ceiling most of the time.

So what did Fusaka change? Mostly headroom and resilience, not a fresh cliff-edge drop. Base, the largest L2 by several usage measures, processed about 292 million user operations — swaps, contract calls, batched account-abstraction transactions — over the 30 days ending September 7, 2026. The total it paid Ethereum for that entire month’s worth of data posting: around $8,800. Spread across those operations, that works out to roughly $0.00003 per transaction in L1 settlement cost. That’s not a post-Fusaka miracle number; it reflects a fee environment that was already extremely cheap and that Fusaka helped keep cheap as usage climbed, rather than letting blob demand outrun supply the way it periodically did before the upgrade.

Independent tracking from L2BEAT’s Onchain Costs dashboard, covering a window in April and May 2026, put typical optimistic-rollup fees (Arbitrum, Optimism, Base) at roughly 3 to 8 times cheaper than transacting directly on Ethereum mainnet, with zero-knowledge rollups like Starknet and Linea running 40 to 100 times cheaper. Those multiples are consistent with a market that’s been broadly stable rather than one experiencing a sudden new step-change from Fusaka specifically.

Put simply: if you were already using an L2 in early 2026, you probably haven’t noticed a dramatic new drop in what you’re paying. What you likely have gotten is more consistency — fewer moments where a busy day pushes blob demand against a hard ceiling and fees creep up in response.

What Changed for ETH Itself — and Why It’s Complicated

This is the part launch-day coverage mostly skipped, and it’s arguably the more consequential story nine months on.

Ethereum burns a portion of every transaction’s base fee under EIP-1559, permanently removing that ETH from circulation. When mainnet activity and fees were high, that burn was substantial — thousands of ETH some days before Dencun. After Dencun moved most rollup activity onto cheap blobs instead of expensive mainnet execution, that burn fell off a cliff: current tracking puts typical daily burn around 50 to 70 ETH, a small fraction of pre-2024 levels. Base-layer fee revenue followed the same path, dropping from roughly $40 million in early 2025 to around $10 million at points in 2026.

The mechanism matters here, and it’s a real, verifiable dynamic rather than speculation: blob transactions pay their own separate fee, the blob base fee, which is also burned — but only in proportion to how scarce blob space actually is. When blob space is abundant relative to demand, as it largely has been since Fusaka expanded the ceiling, that blob base fee sits near its floor, and burn from it stays minimal even as L2 transaction volume keeps climbing. In other words: more usage doesn’t automatically mean more ETH burned anymore, because the cost per unit of that usage keeps falling.

Fusaka’s developers weren’t blind to this. One of its included changes, EIP-7918, sets a minimum blob base fee tied to Ethereum’s execution-layer base fee, specifically to stop the blob fee market from staying permanently near zero regardless of demand. Analyst modeling suggests that if this floor had been in place since blobs launched in 2024, it would have added something on the order of $78.6 million in cumulative burn. That’s a meaningful number, but it’s a floor under future burn, not a restoration of pre-2024 levels — and because Fusaka has only been live for nine months, the real-world impact so far is a fraction of that modeled figure.

Ethereum’s net issuance in 2026 has generally run mildly inflationary, somewhere in the range of 0.2% to 0.8% annually depending on the measurement window, a notable shift from the “ultrasound money” narrative that circulated when burn regularly outpaced issuance. That’s not a crisis for the network’s security or function — validators still get paid, transactions still settle — but it’s a real change in ETH’s monetary story that Fusaka’s scaling success arguably deepened rather than fixed.

The Underlying Tension: Scaling Success vs. Mainnet Value Capture

This is where legitimate criticism lives, and it’s worth stating plainly rather than glossing over. Ethereum’s roadmap has, for years, deliberately pushed activity off mainnet and onto L2s, on the theory that mainnet becomes a settlement and security layer while L2s handle volume. Fusaka is squarely in that lineage — it’s a scaling upgrade, and by that measure it worked. Blob capacity is up, L2 costs are low and stable, and PeerDAS did what it was designed to do without compromising node decentralization.

But some analysts, including research from Fidelity Digital Assets, have pointed out the flip side: the additional blob-fee cost Fusaka’s floor imposes on a large L2 like Base runs somewhere around $30.6 million a year, against roughly $5.2 million in blob fees that L2 actually paid over the prior year and about $94 million in fees it collected from its own users. That gap illustrates the real tension — L2s capture most of the economic value generated by their own usage, while Ethereum mainnet, and by extension ETH holders, see a shrinking slice of it. Whether that’s a design flaw or simply the intended division of labor between L1 and L2 depends on who you ask, and it’s an honest disagreement rather than a settled one.

There’s a second, more practical concern too: if L2 transaction fees keep compressing toward zero, the business model question becomes unavoidable. Several rollup teams have acknowledged, in various public forums this year, that fee revenue alone doesn’t obviously sustain a rollup’s operating costs indefinitely, which is part of why sequencer decentralization, MEV capture, and alternative revenue models remain active areas of L2 development rather than solved problems.

Common Misconceptions Worth Correcting

A few claims circulating about Fusaka deserve a direct correction. First, the “up to 60% fee drop” figure attached to Fusaka in some recent coverage largely describes Dencun’s effect from 2024, not a fresh Fusaka-era cut. Second, headlines citing a future 48-blob maximum are describing a longer-term roadmap target discussed by Ethereum researchers, not a number that has shipped — the live maximum, as of this writing, is 21. Third, PeerDAS reducing bandwidth requirements for regular nodes does not mean large staking operations or institutional node runners saw comparable savings; the efficiency gain is concentrated among smaller, home-staking-style operators, which was the explicit point of the design.

What to Watch Next

Ethereum’s roadmap doesn’t pause. The next scheduled upgrade, Glamsterdam, is targeting a mainnet launch in the fourth quarter of 2026 (no activation date has been set), with two headline features selected through Ethereum’s developer governance process: enshrined proposer-builder separation (ePBS, EIP-7732), which builds the block-builder relationship into the protocol instead of relying on outside relays, and Block-Level Access Lists (BALs, EIP-7928), aimed at speeding up how nodes process and verify entire blocks. FOCIL (fork-choice enforced inclusion lists), designed to make it harder for block builders to censor specific transactions, was dropped from Glamsterdam and deferred to the following upgrade. Ethereum researchers have already floated a name, Hegota, for whatever comes after Glamsterdam — a sign that the twice-yearly upgrade cadence Fusaka helped establish is now the operating norm rather than the exception.

For anyone holding or using ETH and its L2 ecosystem, the practical takeaway from nine months of Fusaka data is this: the scaling roadmap is working as designed for users, and the open question for ETH’s long-term value proposition — whether mainnet activity and fee capture matter as much as they used to — is one Ethereum’s own developers are actively debating, not one this article can resolve for you.

External References

• Ethereum.org: Fusaka Roadmap Overview
• Ethereum.org: PeerDAS Explained
• Ethereum Foundation Blog: Fusaka Mainnet Announcement
• The Block: Ethereum Rolls Out Final Planned Blob-Parameters-Only Fork
• Fidelity Digital Assets: The Fusaka Upgrade – Scaling Meets Value Accrual
• Ethereum.org: Glamsterdam Roadmap

FAQ

When did Ethereum’s Fusaka upgrade actually happen?

Fusaka activated on Ethereum mainnet on December 3, 2025. Its blob-capacity increases rolled out afterward in two stages, called Blob-Parameter-Only forks, on December 9, 2025 and January 7, 2026 — the second of which set the current maximum of 21 blobs per block.

What is PeerDAS in simple terms?

PeerDAS lets Ethereum nodes verify that blob data is available on the network by checking only a small random sample of it, rather than downloading the entire thing. Using erasure-coded redundancy and cryptographic proofs, a node can be confident the full dataset is intact and available without ever storing it all itself — which is what let Ethereum raise blob capacity without pricing out smaller node operators.

Did Fusaka actually cut Layer-2 fees by 60%?

Not really, at least not on its own. That kind of fee drop mostly happened back in March 2024 with the Dencun upgrade, which first introduced cheap blob space. Fusaka arrived into an environment where L2 fees were already low, and its main contribution has been keeping them low and stable as usage grows, rather than delivering a fresh, dramatic cut.

Does cheaper L2 activity mean less ETH gets burned?

Yes, and this is a verified dynamic, not speculation. Since most transaction activity moved to cheap blob space rather than expensive mainnet execution, daily ETH burn has fallen sharply compared to pre-2024 levels, and Ethereum’s net issuance has run mildly inflationary through parts of 2026. Fusaka included a mechanism, EIP-7918, that sets a floor under blob fees specifically to address this, but it doesn’t fully reverse the underlying trend.

Is Ethereum still “ultrasound money” after Fusaka?

The narrative has gotten more complicated. With burn down and issuance steady, ETH’s supply has behaved in a mildly inflationary way for stretches of 2026, a departure from the deflationary story that circulated in prior years. Whether that matters for ETH’s long-term value depends on whether you weight network usage and scaling success more heavily than direct fee capture by the base layer.

What’s the next Ethereum upgrade after Fusaka?

Glamsterdam, targeted for mainnet in the fourth quarter of 2026, with enshrined proposer-builder separation (ePBS) and Block-Level Access Lists as its headline features. FOCIL (fork-choice enforced inclusion lists) was deferred to the upgrade after it. Ethereum developers have already floated a working name, Hegota, for the upgrade expected after that.

Key Takeaways

• Fusaka activated December 3, 2025, and finished its phased blob-capacity rollout on January 7, 2026, landing at a target of 14 and a maximum of 21 blobs per block via PeerDAS.
• Nine months later, L2 fees remain very low (Base processed roughly 292 million operations in 30 days for about $8,800 in total L1 fees), but most of the original fee collapse traces to 2024’s Dencun upgrade, not Fusaka specifically.
• PeerDAS lets ordinary nodes verify blob data by sampling a small fraction of it rather than downloading everything, preserving decentralization while scaling capacity.
• Cheaper, higher-volume L2 activity has genuinely reduced ETH burned via EIP-1559, pushing Ethereum’s net issuance mildly inflationary in 2026 — a real and verifiable tradeoff, not just a talking point.
• The open debate now is whether Ethereum’s L1/L2 scaling model adequately channels value back to ETH and mainnet, or whether L2s are capturing a growing share of the economic upside.

This article is for informational and educational purposes only and does not constitute financial, investment, tax or legal advice. Cryptocurrency investments carry significant risk, including total loss of principal. Always do your own research (DYOR) and consult a licensed financial advisor before making investment decisions.

If you’re tracking how other major networks are handling their own scaling trade-offs, our look at Solana’s Transaction Format v1 upgrade makes for a useful comparison point. And if the stablecoin side of the ecosystem interests you more, our breakdown of the USDC vs. USDT market-share shift covers where a lot of that cheap L2 activity is actually headed.