Blob count hits 80% of daily capacity. Fees already up 4x on Arbitrum. The cheap-L2 narrative is about to crack.
I spent the last 72 hours scraping blob data from Etherscan, Dune Analytics, and L2beat. The results are ugly for anyone betting on sustained low-cost rollups. Post-Dencun euphoria blinded the market. Everyone cheered the 90% fee drop. No one watched the ceiling.
Here’s the raw data: - April 15: 1,200 blobs used in 24 hours. Capacity cap: 1,536. - May 10: 1,322 blobs. That’s 86% utilization. - May 20: 1,401 blobs. Blobs are now the bottleneck, not calldata.
Context: Why this matters now
Dencun (EIP-4844) introduced proto-danksharding. Rollups moved from permanent calldata to ephemeral blobs. Fees dropped instantly. Transaction costs on Arbitrum fell from $0.12 to $0.01. Optimism hit $0.008. The market celebrated. NVDA-level euphoria. But blobs are a finite resource — max 6 per block, ~1,536 per day. Each blob holds ~128KB. That’s it.
The hidden assumption everyone made: Blob supply scales with demand. It doesn’t. Not until full danksharding, which is years away (EIP-7623, maybe 2026). Meanwhile, every new L2 — Base, Blast, Linea, zkSync, Scroll, Polygon zkEVM — competes for the same 1,536 slots.
Let’s break the math down: - 1,536 blobs/day = 192 MB of total L2 data capacity per day. - As of May 20, L2s are pushing 1,401 blobs/day ≈ 175 MB. - Growth rate: L2 transaction volume has increased 340% since Dencun went live in March 2024.
At current growth, we hit full capacity by October 2024. Not two years. Six months.
Core: The fee pressure is already visible
I pulled data from Dune’s blob tracker. Here’s the trend line.
Arbitrum: - March 15 (pre-Dencun): median tx fee $0.23 - March 20 (post-Dencun): $0.01 - May 20: $0.045
4.5x increase in two months. Not because gas prices went up. Because blob space became scarce. Rollups bid for inclusion. Base is already paying 2.3 gwei per blob where Optimism pays 1.1 gwei. The market is fragmenting.
Optimism: - Fee floor: $0.008 → $0.025 (3x increase)
zkSync Era: - Post-Dencun drop to $0.02, now $0.06 (3x increase)
What happens at 100% blob utilization?
Simple: rollup sequencers will outbid each other for blob inclusion. Gas fees will spike non-linearly. I modeled this using historical calldata cost data before Dencun.
At 80% blob utilization, fees double compared to 60% utilization. At 90%, fees jump 5x. At 95%+, we see 20x+ spikes. The L2 fee market will look like a Dutch auction where the highest-paying rollup wins block space.
Which L2s are eating the most blobs?
- Base: 30% of total blobs. Their user base is retail-driven and highly elastic. Fee sensitivity is high. They’ll feel the squeeze first.
- Arbitrum: 25% of blobs. They have a deeper treasury and may subsidize fees longer, but that’s not sustainable.
- Optimism: 18%. Superchain thesis increases their blob demand.
- zkSync: 12%. They’re more efficient per tx, but volume is growing.
- Starknet: 8%. Low volume. Not a concern yet.
- Linea/Scroll/BLAST: the rest.
The Contrarian Angle: The cheap L2 narrative is a marketing gimmick
Everyone thinks L2s solved the scalability problem. They didn’t. They shifted the bottleneck from calldata to blobs. It’s the same problem with a different name. The only difference is blobs expire after ~18 days, so full nodes don’t need to store them forever. But during the active window, competition is real.
Here’s what no one is talking about:
1. Blob price auctions are going to create L2 fee volatility. We saw this with calldata in 2021 — gas wars drove L2 costs to $0.50 per tx. Blobs will create the same dynamic, only faster because supply is rigid.
2. Full danksharding (DAS) is not coming anytime soon. The core devs are focused on Pectra (EIP-7600ish) for late 2025. Even then, DAS is not in the roadmap until at least 2026. That’s two years away. In crypto, two years is an eternity of fee spikes.
3. L2s are incentivized to waste blobs. Many sequencers batch transactions suboptimally to reduce latency. The result: more blobs per user tx than necessary. Base, for example, sometimes posts 3 blobs per block when 2 would suffice. That’s 50% inefficiency.
4. Institutional inflow will accelerate blob demand. When BlackRock and Fidelity start deploying on Ethereum via L2s for tokenized assets (their RWA push), institutional volume will swamp the blob market. Each institutional transaction needs its own blob space for state diff. This isn’t retail — it’s high-value, high-blob tx.
I ran a stress test scenario:
Assume Tesla tokenizes its stock on Ethereum via Base (rumored). They post 10,000 tx/day. Each tx consumes 0.1 blob equivalent. That’s 1,000 blobs/day — 65% of current daily capacity. By itself. That would push blob utilization to 156%. Impossible. Sequencers would have to compress or delay. Either way, fees explode.
What about data compression?
L2s could compress blob data, reduce blob count. But compression ratios are already near theoretical max. zk-rollups are already highly efficient. Optimistic rollups have worse compression. The low-hanging fruit is gone.
Takeaway: This is a time bomb
If you are a trader: watch blob utilization daily. When it hits 90% consistently, short L2-native tokens (ARB, OP). The fee spike will hurt their TVL and user growth.
If you are a builder: budget for fee increases. Your dApp will become unprofitable for small users if blob fees rise 10x. Consider alt-DA solutions (Celestia, Avail, EigenDA) — but those have their own trust and finality tradeoffs.
If you are an investor: the real winner of Dencun might be L1 Ethereum itself, as blob scarcity pushes activity back to mainnet for high-value tx. Also, look at DA-layer tokens (TIA, AVAIL) as the scarcity hedging plays.
The bottom line: Post-Dencun was not the final solution. It was a Band-Aid that created a new bottleneck. And the data shows that bottleneck is already near bursting.
My next steps: I’m building a real-time blob dashboard that tracks 7-day moving average utilization and predicts fee surge windows. First version will be public by end of this week on my station. Gas up or get left behind.
— Jacob Hernandez
Sources: - Etherscan Blob Tracker - Dune Analytics: Blob Space Dashboard by @dshoy - L2Beat: Data Availability Dashboard - EIP-4844 Spec - EIP-7623 Proposal for Calldata Cost Increase