Skip to main content

Ethereum's BPO2 at 100 Days: 40% More Blob Space, 25% Used, and a Tokenomics Reckoning

· 12 min read
Dora Noda
Dora Noda
Software Engineer

Ethereum quietly shipped one of its most consequential scaling upgrades in years on January 7, 2026, at 1:01:11 UTC. There was no Devcon stage. No countdown clock. No price pump. BPO2 — the second "Blob Parameter Only" hard fork — raised the per-block blob target from 10 to 14 and the maximum from 15 to 21, expanding rollup data capacity by 40% in a single coordinated client release. By every technical measure, it worked.

It also created a problem nobody is talking about loudly enough: Ethereum now has more blob space than its L2s know what to do with. Blob utilization sits at 20-30% of the new ceiling. Blob fees have collapsed toward the floor. ETH issuance has crept back ahead of burn. And the next two upgrades on the roadmap — Glamsterdam in H1 2026 and another BPO targeting 48 blobs by mid-year — will pour even more capacity into a market that hasn't absorbed what it already has.

This is the awkward middle of Ethereum's rollup-centric thesis: the engineering is shipping on time, the user fees are falling on schedule, and the token's "ultrasound money" narrative is quietly cracking under the same mechanism that made it credible in the first place.

What BPO2 Actually Changed​

BPO2 is defined by EIP-8135, the meta-EIP documenting parameter changes, executed under the framework of EIP-7892, which authorizes Blob Parameter Only hard forks as a class of lightweight upgrades that touch only three values: blob target, blob max, and the base-fee update fraction.

The numbers tell a clean story:

  • Cancun (March 2024): target 3, max 6
  • Prague: target 6, max 9
  • Osaka / Fusaka activation (December 3, 2025): target 6, max 9
  • BPO1 (late 2025): target 10, max 15
  • BPO2 (January 7, 2026): target 14, max 21

At 21 blobs of 128 KB each, an Ethereum block can now publish 2,688 KB of rollup data — up from 768 KB at Dencun's launch in March 2024. That is a 3.5x expansion of the network's data availability layer in 22 months, achieved without changing a single line of execution-layer code or asking node operators to modify their software beyond a configuration update.

The "no code change" property is the quiet innovation here. Traditional Ethereum hard forks coordinate Geth, Nethermind, Besu, Erigon, and Reth across hundreds of operators on a single block height. BPO forks ship through configuration alone, which means the network can ratchet capacity in response to observed mainnet behavior instead of waiting 12-18 months for the next named hard fork. EIP-7892 turned blob scaling from an event into a dial.

L2 Fees Followed the Blob Curve Down​

The user-visible result was immediate. A typical L2 transaction that cost roughly $0.50 in late 2025 dropped to between $0.20 and $0.30 within weeks of Fusaka and BPO1, and Ethereum Foundation modeling projected an additional 40-60% reduction during Fusaka's first months — with potential 90%+ drops as capacity ramps further.

Base, Arbitrum, Optimism, and zkSync all peg sequencer fees against EIP-4844 blob costs. When BPO2 raised the supply curve, the marginal price of blob space dropped, and L2 sequencer pricing followed within days. By April 2026, simple swaps on Base routinely settle for under $0.05. Bridge transactions on Arbitrum One that cost $1.50 a year ago now cost $0.10. The "Ethereum is expensive" narrative that dominated 2023-2024 retail discourse has lost its remaining substantive claim against L2s — even if mainnet itself remains a $0.50-$3.00 venue for simple transfers.

The competitive read against Solana is more interesting than the headline numbers suggest. Solana's median fee sits near $0.00025 — roughly 100-1000x cheaper than Ethereum L2s in most scenarios. But the gap that mattered for builder choice was never the absolute number. It was the order of magnitude. At $0.50 per L2 transaction, consumer apps couldn't ship the kind of high-frequency interaction patterns Solana enabled. At $0.05, they can. The remaining cost differential matters for narrow workloads (HFT bots, micro-tipping) but stops being a deal-breaker for the vast majority of consumer DeFi, social, and agent-economy use cases.

The 25% Utilization Problem​

Here is the awkward fact: Ethereum L2s are not actually using the blob space BPO2 created.

A MigaLabs analysis of more than 750,000 slots since Fusaka's activation found that blocks routinely carry fewer than the 14-blob target. Blob utilization across Q1 2026 averaged 20-30% of the new ceiling, with the distribution heavily skewed toward lower counts. Some analysts reading the cryptoslate data have argued that Ethereum "solved the wrong problem" with Fusaka — that scaling supply ahead of demand collapsed the price signal that was supposed to fund ETH burn.

There are two competing readings of this:

The optimistic read is that we are in an absorption window. L2s only switched their data-availability layer from calldata to blobs after Dencun in March 2024, and re-architecting sequencer batching, fraud-proof systems, and ZK provers to fully exploit blob throughput takes quarters, not weeks. Demand will catch up as L2 transaction volumes compound and as use cases (high-frequency on-chain agents, fully on-chain games, social protocol activity) ramp.

The pessimistic read is that Ethereum is shipping capacity faster than its rollup ecosystem can absorb it, and the "10x cheaper than mainnet" pricing was the only forcing function getting L2s to fill blobs in the first place. Once blobs are essentially free, the marginal incentive to batch aggressively, compress rigorously, or migrate calldata-mode workloads disappears. The system reaches an equilibrium where L2s pay near-zero for data, users pay near-zero for L2 transactions, and Ethereum L1 captures near-zero in burn.

Both readings can be partially correct. The 20-30% utilization figure is real today; the demand curve for autonomous-agent workloads, app-chain rollups, and consumer applications is also real and growing. The question is the shape of the catch-up curve.

The ETH Value Capture Tension​

This is where the engineering success and the tokenomics failure intersect. Lower blob fees mean less ETH burned per L2 transaction. Less ETH burned means net issuance can exceed net burn. Net issuance exceeding burn means the "ultrasound money" thesis — the post-Merge claim that ETH is structurally deflationary — stops working.

The data has already shifted. Post-Dencun, ETH inflation reached 0.74% in September 2024 as blob fees collapsed and L1 burn dropped with them. The ChainCatcher and CoinLedger analyses both note that the question "is Ethereum still ultrasound money in 2026?" no longer has a clean yes.

Fusaka attempted a fix. EIP-7918, the "Blob Base Fee Bound," establishes a minimum price floor for blob transactions tied to the execution base fee. Even during low L2 data demand, rollups now pay a minimum fee proportional to L1 activity, creating a guaranteed minimum stream of ETH burn during quiet periods. The Liquid Capital projection is that as L2 transaction volumes grow, blob fees could contribute 30-50% of total ETH burn by mid-2026, returning the asset to a deflationary trajectory.

Whether that actually happens depends on three variables nobody can model precisely:

  1. L2 volume growth rate. If on-chain agents, app-chain rollups, and consumer applications drive 10x L2 volume growth in 2026-2027, blob demand will saturate the new ceiling and burn will recover.
  2. Blob target trajectory. Core developers are already planning further BPOs targeting 48 blobs per block by mid-2026, and Danksharding's long-term destination is 128 blobs per slot. Each capacity bump pushes the absorption finish line further out.
  3. L1 demand resilience. Mainnet activity (deep DeFi, institutional RWA settlement, high-value transfers) still generates execution-layer fees that fund burn directly. If institutional flows continue to settle on L1 — as the BlackRock BUIDL and Centrifuge V3.2 patterns suggest — the burn floor holds even with weak blob revenue.

The honest framing is that Ethereum is running an experiment. The hypothesis is that aggressive blob scaling unlocks enough total economic activity that 30% of a much larger pie generates more burn than 100% of a small pie. BPO2 is a midpoint data point, not a verdict.

What Comes Next: Glamsterdam, Hegota, and the 48-Blob Horizon​

The roadmap from here gets denser, not lighter.

Glamsterdam (targeted H1 2026) introduces two structural changes that compound BPO's effects:

  • Enshrined Proposer-Builder Separation (ePBS) via EIP-7732 splits the validation and consensus tasks, expanding the data propagation window from 2 seconds to roughly 9 seconds. That window expansion is what makes much higher blob counts safe for non-supercomputer node operators — it's the precondition for the 48-blob and 72-blob targets that are otherwise bandwidth-prohibitive.
  • Block-Level Access Lists (BALs) require blocks to declare every account and storage slot they will touch before execution, enabling parallel processing on the execution side. Combined with the proposed gas limit raise to 200M, Glamsterdam is targeting "thousands of TPS" on L1 itself.

Further BPOs in mid-to-late 2026 will likely raise blob counts to 48 per block, conditional on observing sustainable performance under BPO2's parameters. The long-term anchor remains full Danksharding at 128 blobs per slot.

Hegota, the late-2026 fork, is expected to layer in additional consensus optimizations and continue the ZK-EVM migration that Vitalik's April 2026 Hong Kong keynote framed as the 2027-2030 endgame.

For infrastructure providers — RPC operators, indexers, archive nodes, app-chain frameworks — this sequence creates a planning challenge. Each BPO incrementally raises the bandwidth and storage burden on full nodes. Each parameter change rebalances the L1 / L2 / app-chain economics that drive RPC demand patterns. Sequencer-driven workloads (predictable batching, deterministic call graphs) increasingly dominate the mix, while human-driven workloads (bursty, irregular) shrink as a percentage of total network activity.

The Builder's Read​

If you are building on Ethereum or its L2s in mid-2026, three things changed because of BPO2 that should change how you architect:

  1. Data availability is no longer a cost constraint for most use cases. If you were rate-limiting on-chain logging, off-chain proof posting, or full-state commitments because blob fees were the bottleneck, you have headroom now. Workloads that looked uneconomic in 2024 — full-on-chain games, agent transaction histories with verifiable provenance, on-chain social graphs at scale — are inside the cost envelope.

  2. The L1/L2 boundary is reshuffling. Glamsterdam's gas limit expansion and BAL-driven parallelism mean L1 will absorb workloads that previously had to go to L2 for cost reasons. Decisions about where to deploy contracts in 2026 should account for the L1-mainnet-as-execution-venue thesis that Vitalik's roadmap explicitly endorses for the late-2020s.

  3. Indexing and RPC patterns are shifting toward sequencer-driven loads. Rollup sequencers post predictable, large blob batches at known intervals. RPC providers, indexers, and archive nodes need to be optimized for the batch pattern — not the bursty human-transaction pattern that defined 2018-2023 infrastructure design.

BlockEden.xyz operates production RPC and indexing infrastructure across Ethereum, Base, Arbitrum, Optimism, and the broader L2 ecosystem most directly affected by BPO2's blob expansion. If you are evaluating data availability cost models or planning for the Glamsterdam transition, explore our API marketplace for chain coverage, sequencer-aware endpoints, and infrastructure built around the rollup-centric roadmap.

The Quiet Inflection​

BPO2 will not be the upgrade Ethereum's marketing team wishes it had. There was no narrative-friendly headline number, no UX win that mainstream users will notice, no new asset class enabled. What it did was confirm that EIP-7892's progressive scaling pattern works — that Ethereum can ratchet blob capacity through configuration changes without coordination crises — and that L2 fee compression is a deliverable engineering outcome, not just a roadmap promise.

It also confirmed that the harder questions are not technical. The 25% utilization figure, the blob fee floor, the ETH burn trajectory, the L1-vs-L2 value capture split — these are economic and behavioral problems that the next two years of Glamsterdam, BPO3, BPO4, and Hegota will either solve or expose. The engineering is shipping. The tokenomics are still being written.

For builders, the practical takeaway is that the "Ethereum is expensive" framing that shaped a decade of architectural choices is now substantively false at the L2 layer, and the L1 layer is on a credible path to following. For ETH holders, the practical takeaway is that price action through 2026-2027 will be driven less by what Ethereum can do and more by whether the demand curve for what Ethereum has already built catches up to the supply curve the protocol keeps producing.

Sources:

Share on Twitter