TL;DR — Key Takeaways
- EIP-4844 blob transactions have reduced L2 fees by 90%+, with an average swap on Optimism dropping from $0.87 to $0.03.
- Combined L2 throughput has reached 245 TPS — a 16× multiplier over Ethereum mainnet — and is still growing ~18% month-over-month.
- Cumulative user fee savings since Dencun are estimated at $4.8B; L2 TVL has grown from $18B to $47B in two years.
- ZK-rollups are closing the gap with innovations like Validium (100× compression) and Volition (per-transaction DA choice).
The Dencun upgrade — Ethereum's most significant protocol change since the Merge — went live on March 13, 2024. Its centerpiece, EIP-4844 (proto-danksharding), introduced a new transaction type called "blobs" that fundamentally changes the cost structure of Layer 2 rollups. Two years later, the data is in: Dencun worked. Perhaps better than even its most optimistic proponents expected.
Before Dencun, rollups posted transaction data to Ethereum as CALLDATA, which was priced identically to regular transaction data — expensive, permanent, and competing directly with user transactions for block space. Post-Dencun, rollups can post data as blobs, which are stored temporarily (~18 days) and priced independently via a separate fee market.
The cost reduction has been dramatic. In the month before Dencun, the average cost for an Optimism user to execute a token swap was approximately $0.87. Today, that same swap costs approximately $0.03. For Arbitrum, the drop was from $0.72 to $0.02. Base, which launched post-Dencun, has never known the pre-blob era — and it shows in their growth trajectory.
The Numbers That Matter
The blob fee market has evolved in ways that surprised even close observers. Here are the key metrics as of March 2026:
L2 transaction volume: Combined L2 throughput has reached 245 transactions per second (TPS), compared to Ethereum mainnet's 15 TPS. That's a 16× multiplier — and it's growing at roughly 18% month-over-month.
Blob utilization: The target of 3 blobs per block is frequently hit during peak hours (US/EU overlap), but rarely exceeds it. The surge pricing mechanism, which increases blob base fees exponentially when usage exceeds the target, has only been triggered in 4% of blocks since the upgrade.
Fee savings: Cumulative L2 user fee savings since Dencun are estimated at approximately $4.8 billion. This is revenue that would have been spent on data availability costs — instead, it stayed in users' wallets or was redirected to protocol usage.
L2 TVL: Total value locked across the top 5 rollups (Arbitrum, Optimism, Base, zkSync Era, StarkNet) has grown from $18B pre-Dencun to $47B today — a 2.6× increase that correlates almost perfectly with the fee reduction curve.
The Developer Flywheel
Lower fees don't just benefit existing users — they expand the design space for what can be built on-chain. Pre-Dencun, applications with per-transaction fees above ~$0.50 were simply non-viable for most use cases. The post-Dencun L2 landscape has unlocked several categories that were previously impossible:
On-chain gaming: Fully on-chain games (FOCG) that require dozens or hundreds of transactions per session are now viable. Titles like Dark Forest, Primodium, and Sky Strife are demonstrating that crypto-native game mechanics — not just asset ownership — can attract real player bases.
Micro-payments: Streaming payments, per-article micropayments, and API monetization at the cent level are now practical. The unit economics finally work without requiring a separate payment channel or state channel architecture.
DeFi composability: Complex multi-protocol DeFi strategies that might involve 5–10 transactions per position adjustment are now economically rational for portfolios as small as $1,000. This dramatically expands the addressable market for automated yield strategies.
The ZK vs. Optimistic Dynamic
Dencun has disproportionately benefited optimistic rollups (Arbitrum, Optimism, Base) in the short term because their existing architecture could immediately leverage blobs. ZK-rollups, which produce cryptographic validity proofs rather than waiting for fraud proofs, have a more complex data posting pipeline.
However, the ZK ecosystem is catching up rapidly. zkSync Era's "Validium" mode can post state diffs rather than full transaction data to blobs, achieving compression ratios of up to 100× relative to naive posting. StarkNet's "Volition" architecture allows applications to choose between on-chain data availability (blobs) and off-chain data availability (DAC) on a per-transaction basis. These innovations suggest that the long-term cost advantage may actually favor ZK architecture — but the market hasn't priced this in yet.
What Comes Next: PeerDAS and Full Danksharding
Dencun's blob architecture is an interim step. The long-term roadmap includes PeerDAS (Peer Data Availability Sampling), which distributes blob verification across the network using erasure coding and KZG commitments, and eventually full danksharding, which will increase the data bandwidth available to rollups by another 10–50×.
The timeline for PeerDAS is approximately Q4 2026–Q1 2027, subject to the results of ongoing Devnet testing. Full danksharding is likely a 2028–2029 milestone. In the meantime, the blob market will continue to evolve: the target of 3 blobs per block may be increased through a simple parameter change if the network can sustain it. Several client teams are already advocating for a target of 6 blobs per block in the upcoming Pectra fork.
What's clear is that Dencun has fundamentally transformed Ethereum's scalability landscape. The rollup-centric roadmap is no longer theoretical — it's working, at production scale, with real economic activity. The question now isn't whether Ethereum can scale, but how quickly the next phases of the data availability roadmap can arrive.
Maya is a protocol researcher and former core developer at the Ethereum Foundation. She holds a PhD in distributed systems from MIT and has been writing about Ethereum's technical evolution since 2019.
