I Analyzed Blob Usage After Pectra—Most L2s Are Using <20% of New Capacity. Future-Proofing or Overbuilding?

General
1

I spent my weekend doing what normal people definitely don’t do - analyzing on-chain blob usage data after Ethereum’s Pectra upgrade. And I found something interesting that I think this community will want to discuss.

TL;DR: Most Layer 2s are using less than 20% of the new blob capacity that Pectra provided. Is this proof we overbuilt, or smart future-proofing?

My Methodology (Feel Free To Verify!)

I pulled blob posting data for the 8 major Layer 2s covering May 7 - July 1, 2025 (the 8 weeks immediately following Pectra’s mainnet launch). I compared blob utilization against the new capacity limits (6 target blobs per block, 9 maximum).

Data sources:

  • Etherscan blob transaction tracker
  • Individual L2 block explorers
  • Dune Analytics L2 metrics dashboard

All of this is public data - I’m happy to share my SQL queries if anyone wants to reproduce the analysis.

What I Found: The Utilization Numbers

Average blob utilization across major L2s (as % of new post-Pectra capacity):

  • Base: 31% - highest utilization, still well below cap
  • Arbitrum: 18%
  • Optimism: 22%
  • zkSync Era: 15%
  • Polygon zkEVM: 12%
  • Scroll: 14%
  • Linea: 11%
  • Starknet: 9%

Even Base, which processes 40% of all L2 transaction volume, is using less than one-third of available blob space on average.

But Peak Usage Tells A Different Story

Here’s where it gets interesting. I dug into the data during the mini bull-run in mid-June 2025 when crypto markets spiked:

Peak blob utilization during 3-4 hour congestion windows:

  • Base: 85% (would have hit 170% under pre-Pectra limits!)
  • Arbitrum: 68%
  • Optimism: 61%

Translation: During normal conditions, we have massive overcapacity. But during volatility spikes, that “overcapacity” prevented blob fee explosions that would have cascaded to user transaction costs.

Historical Context: Ethereum’s Track Record on Capacity

I pulled historical data on Ethereum’s gas limit increases to see if this pattern repeats:

2016 gas limit increase to 4M:

  • Immediate utilization: ~30%
  • 18 months later (DeFi Summer 2017): 95% utilization
  • Assessment: Looked excessive at launch, essential during growth

2020 gas limit increase to 12.5M:

  • Immediate utilization: ~40%
  • 12 months later (NFT mania + DeFi 2021): 98% daily utilization
  • Assessment: Again, looked early but proved necessary

Pattern recognition: Infrastructure upgrades consistently look excessive when measured at launch, but get validated during adoption cycles.

Two Ways To Interpret This Data

Interpretation #1: We Overbuilt

  • 20% average utilization is wasteful
  • Validators store/propagate blobs that aren’t being used
  • Could have shipped a smaller increase and upgraded again later
  • Resources spent on Pectra could have gone to interop standards instead

Interpretation #2: This Is Smart Infrastructure Planning

  • Low average utilization = headroom for growth and volatility
  • Prevents fee spikes during congestion (which we saw avoided in June)
  • Historical precedent shows “overbuilding” gets validated by adoption
  • Better to have capacity before you need it than scramble to add it under pressure

What About The Long Tail?

Here’s another interesting finding: I looked at the 30+ smaller L2s (the “long tail” beyond the top 8).

Collective blob usage of 30+ small L2s: 2.1% of total capacity

This raises @blockchain_brian’s question from another thread - are we subsidizing dozens of L2s that barely anyone uses? Or is that the point of permissionless infrastructure - anyone can build without asking permission?

My Personal Take (Informed by Data, Not Definitive)

I think Pectra was slightly ahead of immediate need but well-timed for growth insurance.

Here’s my reasoning:

  1. Prevented fee volatility - June congestion data shows we would have hit capacity constraints without Pectra
  2. Historical pattern - Ethereum capacity upgrades consistently prove necessary 12-18 months after launch
  3. Low cost of overcapacity - Validators can handle the extra blob storage/bandwidth without major pain
  4. High cost of undercapacity - Fee spikes would have driven users away or fragmented them further across L2s

But I also think we’re reaching a point where more L1 capacity isn’t the highest priority for the next 12-18 months. The data suggests we have headroom. Time to focus on interoperability, UX, and solving the fragmentation problems that actual users care about.

Questions for This Community

  1. Do my numbers match what you’re seeing? (Please verify - I could have made mistakes!)
  2. How should we measure “optimal” capacity utilization? Is 20% too low, or is that healthy headroom?
  3. What should Ethereum prioritize next? More L1 capacity (PeerDAS), or cross-L2 infrastructure?

Really curious what builders, users, and other data nerds think about this.

Context: I’m a data engineer who works on blockchain analytics. I love finding patterns in on-chain data. This analysis took me about 10 hours and way too much coffee. Happy to discuss methodology, share queries, or defend my conclusions!

P.S. - Yes, I know analyzing blob utilization on a Saturday night is peak nerd behavior. No, I don’t regret it. :sweat_smile:

2

Mike, this is exactly the kind of analysis I love to see! As someone building L2 infrastructure, your data confirms what I’ve been experiencing in production.

Why Overcapacity Is A Feature, Not A Bug

Your 20% average utilization number might sound wasteful, but from an infrastructure perspective, this is exactly where we want to be. Let me explain why:

1. Capacity Planning 101: Always Build For Peaks

You don’t size infrastructure for average load - you size it for peak load with headroom. Your June congestion data (85% peak utilization on Base) proves this.

If we had built Pectra for “just enough” capacity based on May 2025 averages, we would have hit breaking point during the June spike. Users would have experienced:

  • Blob fee explosions
  • L2 transaction costs spiking 5-10x
  • Confirmation delays as L2s competed for limited blob space

We didn’t experience any of that, precisely because we overbuilt.

2. The Cost of Undercapacity > Cost of Overcapacity

Let’s talk about what it costs to be wrong in each direction:

Scenario A: We overbuild (current reality)

  • Validators store extra blobs (storage is cheap and pruned after ~18 days)
  • Slightly higher bandwidth costs (marginal)
  • Some development resources “wasted”
  • Total harm: Minimal

Scenario B: We underbuild

  • Fee spikes drive users away from Ethereum ecosystem entirely
  • L2s have to implement their own expensive data availability solutions
  • Network effects fracture as users spread across non-Ethereum chains
  • Total harm: Existential threat to Ethereum’s L2 strategy

The asymmetry is huge. Better to overbuild.

3. Growth Curves Are Nonlinear

You showed historical precedent - gas limit increases that looked excessive proved necessary within 12-18 months.

But here’s the thing: Crypto adoption doesn’t grow linearly. It spikes during bull markets, viral moments, major app launches. If we wait until we’re at 80% utilization to start planning the next upgrade, it’s already too late (because upgrades take 12-18 months to ship safely).

My Prediction

Based on your data and historical patterns, I predict:

  • By Q4 2026, we’ll see 50-60% average utilization
  • By Q2 2027, we’ll see 70-80% average utilization
  • By late 2027, we’ll be talking about PeerDAS being “late” just like we’re now debating whether Pectra was “late”

Current overcapacity will look prescient within 18 months.

Where I Agree With You

You’re absolutely right that more L1 capacity shouldn’t be the priority for next 12-18 months. We have headroom now. Time to focus on:

  1. Cross-L2 messaging standards (EIP-7683 and beyond)
  2. Shared sequencer research and implementation
  3. Better L2 onboarding UX (the Base advantage is mostly UX, not tech)
  4. L2 interoperability tooling (bridges, intent-based systems)

These solve actual user pain points while blob capacity remains adequate.

Methodology Question

One thing I’d love to see: Can you break down blob usage by type of transaction?

I’m curious if we’re seeing:

  • DeFi protocols using X% of blobs
  • NFT mints using Y% of blobs
  • Gaming/social using Z% of blobs

This would help us understand whether certain use cases are capacity-constrained while others have plenty of headroom.

Great work on this analysis! This is exactly the kind of data-driven discussion we need more of in the Ethereum ecosystem.

3

Mike’s data provides important empirical grounding for this discussion. Let me add some historical and technical context.

Ethereum’s History of “Premature” Upgrades That Proved Necessary

Mike showed the pattern - let me add more examples:

2015: Homestead upgrade

  • Removed Canary contracts, increased gas costs for some operations
  • Critics: “Why optimize for scale we don’t have yet?”
  • Result: Enabled the ICO boom of 2017 which would have broken pre-Homestead Ethereum

2019: Istanbul upgrade

  • Reduced costs for ZK-proof verification (EIP-1108)
  • Critics: “Nobody’s using ZK-proofs on mainnet, why optimize for it?”
  • Result: Enabled ZK-rollups which are now critical to scaling

Pattern: Ethereum consistently upgrades infrastructure before demand fully materializes, and gets validated by subsequent adoption.

Why Low Utilization Is Actually Optimal

From a protocol design perspective, Mike’s 20% average utilization is close to ideal. Here’s why:

Queuing Theory Basics:

When utilization of a shared resource exceeds ~70%, you start seeing exponential increases in:

  • Wait times (confirmation delays)
  • Fee volatility (blob pricing spikes)
  • System instability (failed transactions)

Target utilization for stable systems: 30-50% on average, with headroom for 2-3x spikes.

Mike’s data shows we’re at 20-30% average with room for 3-4x spikes before hitting limits. That’s textbook optimal capacity planning.

The Long Tail Question

Mike asked about the 30+ small L2s using just 2% of capacity. Are we “subsidizing” them?

My view: That’s exactly what neutral infrastructure should do.

Ethereum doesn’t know which L2 will become the next Base. Maybe one of those tiny L2s with 100 users today will find product-market fit and scale to millions of users tomorrow.

Providing cheap blob capacity to everyone - including speculative/experimental L2s - enables permissionless innovation. That’s a feature, not a bug.

Comparison: AWS provides the same infrastructure pricing whether you’re a 2-person startup or Netflix. That’s how infrastructure platforms should work.

Technical Nuance: Not All Blobs Are Equal

One thing Mike’s analysis doesn’t capture (and can’t easily capture from on-chain data): blob utilization efficiency.

Some L2s are better at compressing transaction data into blobs than others:

  • Optimistic Rollups (Arbitrum, Optimism, Base): Less efficient, use more blob space per transaction
  • ZK-Rollups (zkSync, Scroll, Starknet): More efficient, use less blob space per transaction

So when we see zkSync at 15% utilization vs Base at 31%, that doesn’t necessarily mean Base is “using the infrastructure more” - it might mean zkSync is using it more efficiently.

This is another argument for overcapacity: It gives room for less-efficient-but-faster-to-develop Optimistic Rollups while also supporting more efficient ZK-Rollups.

What Should Come Next (My Technical Priorities)

  1. PeerDAS (2026) - Already in development, will further increase blob capacity. Given 12-18 month development timelines, we should proceed with this even though current utilization is low.

  2. Cross-L2 communication (EIP-7683 and beyond) - Agree with Mike and Lisa that this is high priority.

  3. Enshrined rollup standards - Long-term, but important for interoperability.

  4. Blob market improvements - Current blob pricing is simple but crude. Could optimize for better price stability.

Response to Mike’s Questions

How should we measure “optimal” capacity utilization?

30-50% average with headroom for 3-5x spikes is optimal for reliability and fee stability.

What should Ethereum prioritize next?

Both/and, not either/or:

  • Continue PeerDAS development (it’s already in progress)
  • Simultaneously focus on cross-L2 interop (which needs more resources)

We can walk and chew gum at the same time.

Context: Ethereum core contributor since 2016. I’ve seen this “are we building too much capacity?” debate repeat every major upgrade. The answer is consistently: we’re building the right amount of capacity, just earlier than current demand requires - which is exactly when you should build infrastructure.

4

Okay I followed like maybe 60% of the technical stuff but here’s my question:

If blobs are currently at 20% utilization and super cheap… does that mean my transaction fees on Base should be basically free? Because they’re not?

I still pay like $0.10-0.50 per transaction on Base depending on what I’m doing. If the infrastructure cost (blobs) is basically zero now, where is that fee going?

Not trying to derail the technical discussion - genuinely trying to understand if I should be seeing cheaper fees as a result of Pectra or if I’m misunderstanding how this all works.

Also Mike’s comment about spending Saturday night analyzing blob data made me feel better about spending Sunday afternoon trying to understand what “blob utilization” even means :joy:

My Actual User Experience

I’ve been using Base since October 2025 (so before Pectra in May). Trying to remember if I noticed any difference:

Before Pectra (October-April):

  • Typical swap fee: $0.20-0.40
  • NFT mint: $0.50-1.00
  • Adding liquidity: $0.80-1.20

After Pectra (May-now):

  • Typical swap fee: $0.10-0.30
  • NFT mint: $0.30-0.80
  • Adding liquidity: $0.50-1.00

Okay actually writing this out, I guess fees DID drop roughly 40-50%? I didn’t really notice at the time because I wasn’t paying close attention. But that’s actually significant!

So I guess my question is: Is 20% blob utilization why my fees are low? Or would they be low anyway?

The “Peak vs Average” Thing Makes Sense

Mike showed that June had some crazy spikes where Base hit 85% utilization during volatile periods.

I think I actually remember this! Mid-June when everything was pumping, I tried to make some swaps and my transactions were taking FOREVER to confirm. Like 2-3 minutes instead of the usual 15-30 seconds.

So I guess the overcapacity prevented my fees from spiking 10x during that period? That would have sucked.

Question For The Engineers

Lisa and Brian are saying “20% utilization is optimal” and “we should build for peaks not averages.”

This makes intuitive sense (like how roads are built for rush hour, not 3am traffic).

But is there a point where we’ve overbuilt so much that it’s wasteful? Like, if blob utilization was 5% average and never went above 20% even during spikes, would you all be like “okay yeah we overdid it”?

Or is the answer just: storage/bandwidth is so cheap for validators that even 5% utilization would be fine?

Context: Still learning this stuff! Really appreciate the patient explanations. This community is great for helping newcomers understand the technical foundations.

Also @data_engineer_mike if you ever want to write a “blob utilization for dummies” guide I would 100% read that.

5

Love the questions Emma! Let me answer from the DeFi economics perspective and also address some of Mike’s data points.

Where Your Transaction Fees Actually Go

@ethereum_emma asked why she still pays $0.10-0.50 on Base if blobs are basically free. Great question!

Your transaction fee is split into several components:

1. L2 Execution Cost (~60-70% of your fee)

  • Compute to run your transaction on Base
  • Paid to Base’s sequencer
  • This is Base’s revenue/profit margin

2. L1 Data Availability Cost (~20-30% of your fee)

  • Cost of posting blobs to Ethereum L1
  • This is what Pectra made cheaper
  • Now nearly zero, but L2s aren’t passing all savings to users

3. Overhead/Margin (~10-20%)

  • L2 operator profit margin
  • Infrastructure costs (servers, devs, etc.)

So when blob costs dropped 90% after Pectra, your total fee only dropped 20-30% because blobs were already a small portion of total cost.

The bigger component is L2 execution, which Pectra doesn’t affect at all.

Why L2s Aren’t Passing All Savings Through

Here’s the uncomfortable truth: L2s are running 95-99% profit margins on transaction fees right now.

Mike mentioned this in another thread. Let me quantify it for Base:

Base’s economics (estimated, weekly):

  • Revenue from user fees: ~$800k/week
  • Blob costs paid to Ethereum: ~$12/week
  • Infrastructure costs: ~$50k/week (servers, devs, operations)
  • Profit margin: ~94%

So Base could reduce fees by another 50-70% and still be wildly profitable. But they don’t, because users are willing to pay current prices and competition isn’t forcing them lower.

This is actually rational behavior - why lower prices if users are happy and switching costs are high?

The DeFi Impact of Low Blob Costs

From a protocol perspective, here’s what cheap blobs enabled post-Pectra:

1. New DeFi Primitives

Strategies that weren’t economically viable before:

  • Micro-rebalancing bots (frequent small adjustments)
  • High-frequency market making (thousands of tiny trades)
  • Automated position management (daily/hourly adjustments)

Our yield aggregator runs 10x more frequent rebalancing post-Pectra because the cost per blob dropped so much.

2. Better Capital Efficiency

When transaction costs are low and predictable:

  • Can move capital more frequently between opportunities
  • Less need to batch operations (better UX)
  • Can chase smaller yield differentials profitably

3. Margin Expansion for L2s (As Noted Above)

L2s are printing money. This is good short-term (they’re profitable and can invest in growth) but might be bad long-term (less pressure to innovate or compete on fees).

Mike’s Data + Economic Interpretation

Mike showed 20% average utilization. From an economics perspective, this means:

We’re in a “supply glut” phase:

  • Blob supply (capacity) far exceeds demand
  • Prices are at floor (~1 wei)
  • L2s capture all economic surplus (not users or Ethereum L1)

This is actually fine for now because:

  • Users still benefit from lower fees (even if L2s could go lower)
  • Ethereum doesn’t need blob fees for revenue (ETH burn from base fees is sufficient)
  • Overcapacity enables experimentation and growth

But if Lisa’s prediction is right (50-60% utilization by Q4 2026), we’d move toward equilibrium where blob costs matter more and L2 margins compress.

My Answer To Emma’s “Wasteful” Question

If utilization was 5% and never spiked above 20%, would that be wasteful?

From a pure economics perspective: No, not wasteful if the cost is low.

Storage and bandwidth for blobs is cheap for validators (few dollars per month). Even at 5% utilization, the social benefit of enabling permissionless L2 innovation > the marginal cost to validators.

From a prioritization perspective: Yes, potentially wasteful if we kept building more capacity instead of solving other problems.

If we were at 5% utilization, I’d say: stop working on PeerDAS, focus 100% on cross-L2 interop.

At 20% utilization with occasional 85% spikes? I’d say: continue PeerDAS development but don’t rush it, focus heavily on interop simultaneously.

Context: Run a DeFi yield aggregator, so I obsess over fee economics and L2 cost structures. Happy to dig deeper into any of the economics here!

P.S. @ethereum_emma I’m also planning to write a “DeFi fee breakdown for users” guide because this is confusing for everyone!