Infrastructure Analysis

The Layer 2 Scaling Landscape: Where We Are in 2025

When Ethereum's network was congested and gas prices made simple token transfers cost more than the transactions themselves, the case for Layer 2 scaling was self-evident. That was the world of 2021. By 2025, the scaling picture looks dramatically different — and considerably more complex. The rollup ecosystem has matured from a set of experimental protocols into a competitive landscape of production-grade systems handling billions of dollars in daily value transfer.

At CloudWorx Capital, we have tracked this evolution closely because our portfolio has significant exposure to infrastructure built at every level of the Layer 2 stack. This piece represents our current thinking on where the technology stands, which bets appear to be paying off, and where we see the most important open questions remaining.

The Rollup Paradigm Has Won

The most important development in Layer 2 scaling over the past four years has not been any single technical breakthrough — it has been the decisive convergence of the ecosystem around the rollup architecture as the dominant scaling paradigm. State channels, plasma, and various other approaches that were competitive with rollups in 2021 have either been abandoned, evolved into rollup variants, or retreated to specialized niches. The debate is now entirely within the rollup family.

This convergence happened for good reasons. Rollups inherit Ethereum's security guarantees in a way that earlier approaches could not match. They allow execution to happen off-chain while anchoring validity proofs or fraud proofs to Ethereum mainnet, preserving the trust assumptions that users and applications care about while eliminating the throughput constraints of executing every transaction on the base layer. The elegance of the architecture — and the fact that it can be implemented without requiring changes to Ethereum itself — made it the natural winner.

Within the rollup paradigm, the major division remains between optimistic rollups (which assume transactions are valid unless challenged) and ZK-rollups (which generate cryptographic validity proofs for every batch of transactions). Both approaches are in production with significant usage, and the competitive dynamic between them has produced better technology on both sides than either would have produced in isolation.

The Optimistic Rollup Ecosystem in 2025

Optimistic rollups achieved production scale faster than ZK-rollups because their architecture was simpler to implement. The core insight — that you can save computation by assuming honesty and only doing work when someone challenges — mapped relatively directly onto existing smart contract infrastructure. This allowed optimistic rollup teams to ship production systems while ZK teams were still working on proving system efficiency.

The major optimistic rollup systems have all reached meaningful maturity in 2025. Transaction costs on these systems have fallen to fractions of mainnet costs. Developer tooling has reached parity with or exceeded the experience of building directly on Ethereum mainnet. User onboarding infrastructure — bridging, fiat onramps, wallet integrations — has improved dramatically.

The most significant ongoing challenge for optimistic rollups is the withdrawal delay. Because the system relies on fraud proofs rather than validity proofs, there is an inherent waiting period before withdrawals from the rollup to Ethereum mainnet are finalized. Various approaches have been developed to mitigate this through liquidity providers who front users capital in exchange for a fee, but the fundamental constraint remains. For applications that do not require fast finality to mainnet — which is most applications — this is not a serious problem in practice. For applications that require rapid cross-system settlement, it remains a limitation.

The ZK-Rollup Revolution

If optimistic rollups won the first generation of the scaling wars by shipping first, zero-knowledge rollups appear to be winning the second generation by being fundamentally better on the properties that matter most at scale. The convergence we anticipated in our early investment theses around ZK technology has materialized — faster than some expected, slower than the most optimistic projections.

The core advantage of ZK-rollups is mathematical certainty. A validity proof that verifies correctly provides the same security guarantee as having every node in the Ethereum network re-execute the transaction. There is no fraud challenge period, no delayed finality, no reliance on the assumption that at least one honest party is watching. The proof either verifies or it does not, and if it verifies, the state transition is final.

The cost of this certainty was, for many years, prohibitive. Generating zero-knowledge proofs is computationally intensive. The early ZK-rollup systems could only handle relatively simple transaction types — transfers and DEX trades — because more complex smart contract logic was too expensive to prove. The history of ZK-rollup development over the past four years is largely the history of making proof generation progressively faster, cheaper, and more general.

By 2025, that work has paid off in a substantial way. The leading ZK-rollup platforms now support arbitrary smart contract execution with proving times and costs that are economically competitive with optimistic rollup alternatives. The hardware acceleration ecosystem — custom ASICs and FPGAs designed specifically for ZK proof generation — has matured to the point where proving costs are falling on a curve reminiscent of Moore's Law. We expect this trajectory to continue.

The Sequencer Decentralization Problem

One of the most persistent criticisms of the current rollup landscape is the centralization of the sequencer — the component that orders transactions and batches them for submission to mainnet. Almost all production rollup systems today rely on a single sequencer operated by the rollup team, which creates a centralized point of failure and control that is at tension with the decentralization values that motivated building on Ethereum in the first place.

The rollup teams acknowledge this problem and have published various roadmaps for sequencer decentralization. The challenge is that decentralizing the sequencer is technically non-trivial: you need a consensus mechanism that can reach agreement on transaction ordering with low latency, at high throughput, without introducing new trust assumptions that undermine the rollup's security model. None of the proposed solutions have yet shipped in production in a way that resolves all of these constraints simultaneously.

From an infrastructure investment perspective, we view sequencer decentralization as one of the most important open problems in the Layer 2 ecosystem — and therefore one of the most interesting investment opportunities. Teams working on shared sequencer infrastructure, decentralized sequencer networks, and based rollup architectures that outsource sequencing entirely to Ethereum validators are all working on facets of this problem. We expect the next three years to produce significant progress.

The Application-Specific Rollup Trend

One of the more interesting structural trends in the Layer 2 landscape has been the emergence of application-specific rollups — chains that are built to serve a single application or closely related set of applications rather than functioning as general-purpose execution environments. The game chain, the perpetuals exchange chain, the social protocol chain: these represent a different philosophy about how rollup infrastructure should be organized.

The argument for application-specific rollups is compelling. When you control the execution environment, you can optimize it entirely for your use case. You can make opinionated choices about gas pricing, transaction prioritization, data availability, and finality assumptions that would be unacceptable in a general-purpose environment but are perfectly sensible for a specific application domain. You can capture MEV internally rather than leaking it to external searchers. You can upgrade more rapidly because you are not beholden to the backwards-compatibility requirements of a general platform.

The argument against is equally compelling: fragmentation of liquidity, complexity for users navigating multiple environments, duplication of infrastructure costs across many chains that each need their own bridging, monitoring, indexing, and security infrastructure. The ecosystem is actively working out the right balance between these forces, and we expect the answer to depend heavily on application category.

Looking Forward: What the Next Three Years Will Determine

The Layer 2 landscape in 2025 is a genuinely competitive market with multiple well-funded, technically sophisticated teams pursuing different architectural bets. The winner will not be determined by marketing narratives — it will be determined by which systems actually deliver the best experience for users and developers over the long term.

The questions we are watching most carefully: Can ZK proving costs continue to fall at their current rate, or will there be diminishing returns that preserve space for optimistic approaches? Will shared infrastructure — shared sequencer networks, shared data availability layers, shared proof aggregation systems — reduce the per-application cost of rollup deployment to the point where every significant application has its own chain? How will the interoperability problem between rollups be solved, and who will build the canonical infrastructure for that?

These questions will have large answers, and those answers will create the next generation of infrastructure investment opportunities. We are paying close attention.

For more on CloudWorx Capital's infrastructure investment thesis, see our about page or browse our portfolio of web3 infrastructure companies.