Six Months After Fusaka: How PeerDAS Quietly Rewrote Ethereum's L2 Fee Equation

Six months ago, Ethereum executed one of its most consequential hard forks in years. On December 3, 2025, at epoch 411392, the Fusaka upgrade went live — combining the Fulu consensus layer with the Osaka execution layer and delivering PeerDAS, the technology that blob-scaling advocates had been waiting for since EIP-4844. Today, the numbers are in, and they tell a story more nuanced than the hype promised.

L2 fees dropped 40–60% overnight. Arbitrum, Optimism, and Base are processing 3.5 times more transaction data than before Fusaka. And Ethereum's blob capacity — the critical throughput measure for rollups — grew from 6 blobs per block to 48, an 8× jump that is still scaling upward. But this is not purely a success story. The same economics that lowered user fees are forcing L2 operators to rethink their entire business model.

What PeerDAS Actually Is (and Why It Matters More Than "More Blobs")​

Before Fusaka, every Ethereum full node downloaded every blob in every block. At 6 blobs per block, this was manageable. Scale to 48 or 256 blobs and the math breaks: storage and bandwidth requirements would price out smaller operators, centralizing the validator set.

PeerDAS (Peer Data Availability Sampling), standardized as EIP-7594, solves this with an elegant probabilistic trick. Rather than downloading entire blobs, nodes download a random 50% sample of erasure-coded blob data from peer subnets. Because the data is erasure-coded, any 50% of the shards is sufficient to reconstruct the original. A node that successfully samples its portion can cryptographically attest to availability without having stored the full dataset — reducing bandwidth requirements for full nodes by up to 80%.

The result: Ethereum can carry far more blob data per block without requiring each validator to handle proportionally more load. Validators now follow new fork-choice rules, voting only for blocks after independently verifying blob availability. Security is preserved; storage costs are not socialized.

This is not danksharding — full danksharding remains a future milestone. PeerDAS is the foundational networking layer that makes danksharding architecturally possible.

The Blob Parameter Forks: A Phased Ramp​

Fusaka did not deliver 48 blobs at launch. It delivered the infrastructure to safely increase blobs, then executed that increase in a phased schedule:

• Fusaka launch (Dec 3, 2025): PeerDAS live, baseline blob infrastructure deployed
• Blob Parameter Only Fork 1 (Dec 17, 2025): Target 10 blobs/block, maximum 15
• BPO 2 (Jan 7, 2026): Target 14 blobs/block, maximum 21
• Current trajectory: Approximately 48 blobs per block as the phased ramps continue

Each BPO fork is a low-risk, non-contentious upgrade that requires no EIP process — just a coordinated parameter change across Ethereum clients. This is a deliberate engineering philosophy: prove PeerDAS works at scale incrementally, then ramp the parameters. It avoids the big-bang risk of jumping from 6 blobs directly to 48.

L2 Economics: The Fee Compression Paradox​

The user experience improvement is real and measurable. Layer-2 transaction costs fell 40–60% in the weeks following Fusaka, with some use cases seeing 95% reductions. For microtransaction-heavy applications — gaming, tipping, high-frequency DeFi — the economics shifted from marginal to viable.

But here is the paradox facing every L2 business: the same upgrade that makes L2 transactions cheaper per user compresses revenue per transaction for L2 operators. Pre-Fusaka, posting transaction data to Ethereum was a meaningful cost that L2s priced into their sequencer fees. Post-Fusaka, data availability costs dropped dramatically, and competitive pressure forces L2s to pass those savings to users.

Arbitrum and Optimism, which have already leveraged Fusaka to process 3.5× more transaction data, now must grow volume 3.5× faster than they shrink per-transaction margins just to maintain revenue. The L2 business model is shifting from margin-per-transaction to volume-at-scale.

This dynamic is not new — it mirrors every infrastructure cost compression cycle in technology. It selects for L2s with strong distribution advantages, sticky liquidity, and proprietary application layers. Pure sequencer plays with no vertical differentiation will face margin erosion.

The Gas Limit Expansion: A Separate Scaling Vector​

Alongside PeerDAS, Fusaka raised Ethereum's block gas limit from 45 million to 150 million — a 3.3× increase that directly expands execution throughput on mainnet. This is independent of blob scaling; it allows more transactions and more complex smart contract interactions per block on L1 itself.

The gas limit raise comes with a new constraint: a 16.7 million gas cap per individual transaction. This is a denial-of-service hardening measure. Large transactions that previously could theoretically consume most of a block are now bounded, preventing single actors from monopolizing block space and improving consistency for other users.

Together, PeerDAS and the gas limit expansion deliver two orthogonal scaling improvements: more data availability for L2 rollups, and more execution capacity on L1. Ethereum's 2025 research team characterized this combination as unlocking a theoretical 100,000+ TPS across the full L2 ecosystem — compared to roughly 12,000 TPS pre-Fusaka.

Vitalik's Framing: Structural Change, Not Just Incremental Improvement​

In January 2026, Vitalik Buterin explicitly called out PeerDAS and ZKEVMs together as marking "structural changes in Ethereum" — a distinction he draws carefully to separate protocol-level architecture shifts from feature additions.

His argument: PeerDAS changes the data availability architecture permanently. Future upgrades that increase blob counts further — including full danksharding — are now ramps on an established foundation rather than new engineering challenges. ZKEVMs similarly change Ethereum's execution verification model at a structural level. The combination of these two shifts positions Ethereum to scale vertically (more throughput) without sacrificing the decentralization properties that differentiate it from alternative L1s.

This framing matters for developers and investors evaluating the Ethereum ecosystem. The roadmap is no longer aspirational — it is executing on a proven technical trajectory.

What Glamsterdam Brings Next​

Ethereum's next major hard fork, Glamsterdam, is targeting the first half of 2026, with a tentative timeline around mid-year. It is already shaping up to be the most significant upgrade since The Merge in terms of execution-layer impact.

Key proposals under consideration for Glamsterdam:

• Enshrined Proposer-Builder Separation (EIP-7732): Formalizes the MEV supply chain at the protocol level, improving fairness and reducing the power of external relay infrastructure
• Block-Level Access Lists (EIP-7928): Enables parallel transaction execution by declaring state dependencies upfront, potentially enabling true multi-core block processing
• Gas repricing: Adjusts costs to better reflect actual resource consumption, which affects smart contract optimization strategies

The projected outcome is a 10,000 TPS expansion at L1 and 78% lower gas fees — on top of the Fusaka baseline. Following Glamsterdam, Hegotá is already planned as the subsequent upgrade later in 2026.

The pace of Ethereum protocol development has materially accelerated. Where the Merge-to-Shanghai interval measured in years, Ethereum is now shipping major hard forks in 6–12 month cadences.

What This Means for Builders​

For developers deploying on Ethereum or its L2 ecosystem, Fusaka's actual impact has already been absorbed. The fee environment today is the post-Fusaka baseline; it is not going back to pre-PeerDAS levels. Builders should:

1. Reassess data posting strategies: The cost of calldata-vs-blobs tradeoffs has shifted. Blobs are now overwhelmingly cheaper for rollup data, and that gap widens with each BPO fork.

2. Design for lower transaction costs: Use cases that were economically marginal pre-Fusaka — micropayments under $0.01, on-chain gaming moves, high-frequency settlement — are viable today. Product roadmaps should reflect this.

3. Watch Glamsterdam timelines: EIP-7928 (block-level access lists) could enable parallel execution, which would require tooling and sequencer-level changes for L2 operators. Building awareness now prevents late-cycle scrambles.

4. Evaluate vertical differentiation: On the L2 layer specifically, sequencer margin compression means application-layer value capture matters more. The L2s that thrive post-Fusaka will be those with differentiated DeFi ecosystems, gaming communities, or institutional integrations — not those competing purely on fees.

Conclusion​

Fusaka delivered on its primary promise. PeerDAS is live, blob throughput has grown 8×, and L2 fees have fallen by as much as 95% in the six months since December 2025. The Ethereum scaling roadmap — long criticized as aspirational — is now executing in real time. Glamsterdam is next, bringing L1 execution improvements that complement Fusaka's data availability gains.

The harder question is what this means for the L2 competitive landscape. Fee compression is a gift to users and a margin squeeze for operators. The L2s that emerge strongest from this environment will be those that turned cheap data availability into differentiated applications, not those that competed on sequencer economics alone.

The danksharding era is no longer theoretical. It is already changing how Ethereum works — and how builders should think about what they build on top of it.

bex.co provides enterprise-grade node API infrastructure for Ethereum, Base, Arbitrum, and 20+ other chains. As the Fusaka and Glamsterdam upgrades continue to reshape L2 economics, our infrastructure scales alongside — so builders can focus on applications, not uptime. Explore our API marketplace to get started.