Cheap blobs, thin margins: three scenarios for the Layer 2 business model
Fusaka made data cheap for rollups, yet many chains still cannot cover their costs. Starknet's L1 vote and Robinhood's fee days point to very different futures.
LAYER 2ETHEREUMANALYSIS
Sofia Marchetti
10/28/20263 min read
Ethereum's scaling roadmap has largely delivered what it promised. Rollups now post data to mainnet for very little, transactions on the biggest networks cost fractions of a cent, and capacity is no longer the bottleneck. The problem for 2026 is different. Cheap infrastructure has not translated into healthy businesses for most Layer 2 networks, and the sector is now splitting into a few models that look less and less alike.
What Fusaka changed
The Fusaka upgrade, activated in December 2025, introduced PeerDAS, a system that lets validators check data availability without downloading every blob in full. It was followed by two blob-parameter-only forks. According to The Block, the second of these, on January 7, 2026, raised the target to 14 blobs per block and the maximum to 21, up from 6 and 9 before the upgrade. For rollups, which buy blob space to publish their transaction data, that meant a lot more supply at a lower price.
The effect on margins shows up in the extremes. Robinhood Chain, which runs on Arbitrum technology and launched its mainnet in July, collected about $4.5 million in user fees on September 3, according to Bitquery estimates cited by KuCoin Research, while spending roughly $400 on Ethereum data and proof costs that day. A single day is not a run rate, but the ratio shows how little settlement now costs a busy chain. At the other end, Blast generated $1,793 in network revenue in September, according to DefiLlama figures reported by CoinDesk, before announcing it would shut down. Cheap blobs could not save a chain that had lost its users.
That gap is the core of the new Layer 2 economy. Costs to Ethereum are close to negligible, so profits depend almost entirely on demand. Demand, in turn, depends on distribution, applications and order flow, which only a few networks have.
Scenario one: the corporate chain
The first model is the chain owned by a large consumer company. Base gives Coinbase a way to earn from activity that would otherwise go to other networks. Robinhood Chain does the same for a broker with millions of retail accounts and a growing tokenized stock business. These chains do not need a token or a speculative community to justify themselves; they need to lower costs and keep customers inside one product. Their revenue flows mainly to the parent company, with a share sometimes returned to the technology provider. Under the Arbitrum Expansion Program, chains settling outside Arbitrum One and Nova contribute 10% of net protocol revenue to the Arbitrum ecosystem. If this model dominates, the Layer 2 market starts to look like a set of private platforms that rent Ethereum's security.
Scenario two: the independent exit
The second model is to stop being a Layer 2 at all. Starknet's community voted between October 15 and 22 to research a move to an independent, quantum-resistant Layer 1, with 93.52% of voting power in favour across 3,236 votes. The vote was non-binding, a draft whitepaper is expected in November, and Starknet chief Eli Ben-Sasson has pointed to 2027 as a target. The logic is that a network with its own technical agenda, in this case post-quantum cryptography, may gain more from controlling its own consensus than from settling on Ethereum. The early market response was harsh: value secured on Starknet fell sharply in the following week. Other mid-sized chains will be watching whether that loss is temporary.
Scenario three: a stronger base layer
The third model puts Ethereum itself back at the centre. Core developers are working on Glamsterdam, the next major upgrade, although testing on the Hoodi network has been postponed. Vitalik Buterin has backed EIP-8288, a proposal to aggregate signatures and proofs in the mempool, which is part of a broader push to make the base chain cheaper to use directly. If mainnet becomes good enough for many applications, the case for small general-purpose rollups weakens further. What would remain are specialised chains for trading, payments or games, and rollups tied tightly to Ethereum's own block production.
These scenarios are not mutually exclusive. The most likely outcome over the next year is a mix: a few corporate chains with real revenue, a handful of independent projects testing their own path, and a long tail that either merges or closes, as Blast and Abstract already have.
For users and builders the practical signals are clear enough. Watch revenue rather than incentives, check whether a chain's activity depends on one company or one reward programme, and pay attention to how each network describes its future relationship with Ethereum. In 2026, that relationship has become a strategic choice rather than a given.