Can We Call Ethereum Scaling Done?

Ethereum’s Layer 2 ecosystem is no longer experimental: rollups now process the majority of transactions, hold tens of billions in TVL, and underpin expanding DeFi and consumer activity. Layer 1 has evolved into an anchor for settlement and data availability, while rollups serve as the de facto execution layer.

This transition, formalized in Ethereum’s 2020 rollup-centric roadmap, reached full production scale in 2023–2025 as demand aligned with infrastructure improvements and ecosystem momentum. But what does “done” mean for a technology stack that continues to evolve at breakneck pace?

Optimistic Rollups: Market Leaders by Adoption

Optimistic rollups such as Arbitrum, Base, and Optimism dominate in adoption and usage metrics. Each has carved out a distinct strategic position in the ecosystem.

Base, Coinbase’s implementation on the OP Stack, achieved something remarkable: explosive growth without issuing a token. It regularly hosts millions of daily active users and surpassed $9 billion in TVL by mid-2025 (2x growth in the last ±6months). This demonstrates pure product-market fit—users choose Base for its functionality and user experience, not token incentives.

Arbitrum maintains its position as the DeFi anchor with deep liquidity pools and sustained developer momentum. The ecosystem has matured beyond speculation to support genuine financial infrastructure. Optimism plays a different game entirely, positioning itself as the backbone of the “Superchain” vision where multiple rollups share security and interoperability standards.

ZK Rollups: The Technical Frontier

ZK rollups, zkSync Era, Linea, Scroll, and StarkNet—offer stronger cryptographic guarantees with instant validity proofs. While adoption currently trails optimistic rollups, these networks continue to push technical boundaries.

The gap between optimistic and ZK rollup adoption isn’t just about technology—it’s about go-to-market strategy, developer experience, and ecosystem momentum. Optimistic rollups chose pragmatism and shipped early. ZK rollups chose technical excellence and are now catching up on user experience.

March 2024’s Dencun upgrade delivered EIP-4844, introducing blob space that fundamentally transformed L2 economics. The numbers are striking: median transaction fees dropped from ~$0.30 to <$0.01—a greater than 90% reduction that changes the fundamental economics of on-chain activity.

This isn’t merely about cheaper transactions. It’s about enabling entirely new categories of applications:

• High-frequency updates: Applications can now update state every block without prohibitive costs

• Granular data: Instead of aggregated summaries, protocols can store detailed transaction-level data

• Complex computations: Logic that was previously off-chain due to cost can now execute directly on L2s

Activity metrics tell the story. L2 networks now regularly process over 250 TPS compared to Ethereum L1’s ~15 TPS. This 16x throughput improvement, combined with the 90% cost reduction, creates a 160x improvement in cost-efficiency for data-intensive applications.

The economic shift also reallocates value within the stack. Sequencers now capture more fee revenue, raising important questions about sequencer decentralization and MEV distribution that the ecosystem is actively addressing.

Beyond execution, L2 scalability intersects with the broader modular blockchain thesis. Data availability has become a marketplace with distinct trade-offs:

• Ethereum L1 provides maximum security and complete verifiability but at the highest cost. It remains the gold standard for applications requiring absolute trust minimization.

• Celestia offers purpose-built data availability with significantly lower costs. Its success demonstrates market demand for specialized DA layers.

• EigenDA creates an Ethereum-aligned alternative through restaking, attempting to balance security with cost efficiency.

Rollups increasingly support multiple DA layers, letting applications choose their trust model. A DeFi protocol securing hundreds of millions might require Ethereum DA, while a gaming application optimizes for Celestia’s cost structure. This flexibility represents blockchain architecture maturing from one-size-fits-all to application-specific optimization.

→ Check the Rollup-as-a-Service Map to discover all modular stacks and frameworks

Projects like MegaETH and Rise are pioneering real-time Layer 2 networks with 200ms – 2 second block times and minimal confirmation latency. Base itself now operates with 200 ms blocks, faster than human reaction time.

These aren’t incremental improvements. They’re attempting to match the performance characteristics of traditional centralized systems while maintaining decentralization properties. The technical challenges are substantial:

• State management at high frequency

• Network propagation across global nodes

• Consensus mechanisms that can finalize in milliseconds

• Data availability proofs that don’t bottleneck throughput

If successful, these networks enable use cases previously impossible on blockchain: real-time gaming with thousands of concurrent players, high-frequency trading directly on-chain, and responsive user interfaces that feel like Web2 applications.

Justin Drake’s Lean Ethereum roadmap, unveiled on Ethereum’s tenth anniversary, provides the framework for understanding the long-term trajectory. The vision encompasses three integrated sublayers:

• Beacon Chain 2.0 hardens consensus with post-quantum cryptography and single-slot finality. This isn’t just about security, it’s about preparing for a world where quantum computers could break current cryptographic assumptions.

• Blobs 2.0 introduces granular data structures optimized for rollup efficiency. The goal is data availability that scales with demand while maintaining verifiability on consumer hardware.

• EVM 2.0 reimagines execution for zero-knowledge proving. Every computation becomes provable, every state transition verifiable, creating a foundation for trustless interoperability.

The performance targets, 10,000 TPS on L1 and 1 million TPS on L2, seem ambitious until you examine the trajectory. We’ve already achieved 16x throughput improvement and 90% cost reduction in two years. Another 100x improvement over a decade follows the same exponential curve that has characterized Ethereum’s growth.

Critically, this roadmap assumes continued L2 dominance with L1 as the security backbone. It’s a vision of specialization where each layer does what it does best.

The question “Is Ethereum scaling done?” misses the point. Technology is never “done”, it evolves continuously. The better question is whether current scaling is sufficient for mainstream adoption.

By most metrics, the answer is yes. Transaction costs have dropped below the psychological threshold where users think about them. Throughput exceeds demand for all but the most extreme use cases. Confirmation times satisfy user expectations trained on traditional financial systems.

But “good enough” doesn’t mean development stops. It means the foundation is solid enough to build upon. We’re past the experimental phase where fundamental viability was in question. We’re in the optimization phase where improvements are incremental but compound over time.

Ethereum and Solana embody two opposing scaling philosophies:

• Ethereum’s modular approach separates settlement and data availability on L1 from execution on L2 rollups, offering flexibility but still facing challenges like fragmentation, sequencer centralization, and the need for further scaling. Ethereum L2s collectively hold over $40 billion TVL despite requiring users to navigate multiple chains, bridges, and gas tokens. This suggests a different value proposition: users accept complexity in exchange for Ethereum’s security guarantees and the experimental diversity of specialized rollups.

• Solana’s monolithic design consolidates all execution, consensus, and data availability on a single chain, enabling fast, low-cost transactions and a unified user experience, albeit at the expense of higher validator hardware demands and potential bottlenecks under sustained load. It processes >1000 TPS in production with 10-second (full) finality*, no bridges, no fragmentation, no sequencer debates. Just pure, unified execution. The $25 billion TVL concentrated on a single chain creates deep liquidity without cross-chain complexity. Users get what they want: it just works.
  

While developer activity remains centered on the EVM, Solana’s performance today shows that a well-optimized monolithic chain can compete strongly. 

Ethereum’s roadmap, including blob scaling, Danksharding, and the new Lean Ethereum vision, aims to match or exceed Solana’s UX in a modular framework that retains maximum decentralization and flexibility. The bet is that fragmentation is a solvable UX problem while global consensus on every transaction is a hard physical limit.

*Solana is expected improve to sub-sec finality with the upcoming Alpenglow upgrade.

Despite remarkable progress, significant challenges persist:

• Sequencer centralization remains the elephant in the room. Most L2s operate with centralized sequencers, creating single points of failure and value extraction. Shared sequencing solutions like Espresso and Astria are promising but unproven at scale. Based rollups seem to be the leading direction.

• Fragmentation across L2s creates friction that monolithic L1s avoid by design. Users must bridge assets between Arbitrum, Optimism, Base, and others, each with different withdrawal periods, security assumptions, and fee structures. Developers deploy the same contract multiple times, managing liquidity across chains. 

This fragmentation is Ethereum’s Achilles’ heel compared to Solana’s unified state or Hyperliquid’s integrated experience. Solutions exist, such as cross-chain messaging protocols, account abstraction, and intent-based bridges, but they add complexity. The question is whether superior scalability and flexibility justify this fragmentation, or whether users will prefer the simplicity of monolithic chains even with their limitations.

State growth threatens long-term sustainability. As L2s process more transactions, state size grows unbounded. State rent, expiry, and statelessness proposals exist but lack consensus on implementation.

Proving costs for ZK rollups remain substantial. Generating proofs for complex computations is expensive, limiting update frequency and increasing latency. Hardware acceleration and proof aggregation help but haven’t fully solved the problem.

The current state of Ethereum scaling enables primitives that were technically impossible or economically unviable just two years ago.

• Decentralized capital markets have become the most exciting use case, enabling users to easily trade stocks, commodities, FX rates and more on-chain, plus their integration into DeFi protocols.

• Decentralized social networks become feasible when posting costs pennies, not dollars. Lens Protocol and Farcaster show early promise, though they’re still finding product-market fit.

• Micro-transactions unlock new business models. Streaming payments, pay-per-use APIs, and granular resource pricing become practical when transaction fees are negligible.

And yes, oracles can finally go fully on-chain – and at DIA have already shipped the first fully onchain, rollup-based oracle. The original vision of transparent, verifiable data feeds was always constrained by L1 economics. With L2 scaling, oracles can publish every price update, show every data source, and make every computation auditable, delivering on blockchain’s promise of transparency for external data. This isn’t about any particular oracle provider winning; it’s about the entire category becoming what it was always meant to be.

Ethereum scaling has reached a crucial milestone: it works. Not perfectly, not completely, but well enough that the conversation has shifted from “can we scale?” to “how do we optimize?”

Layer 2s process the majority of transactions at costs that make blockchain accessible to mainstream users. The infrastructure is production-ready, battle-tested, and improving rapidly. EIP-4844 proved that transformative upgrades can ship successfully. The modular stack provides flexibility for different trust models. Real-time networks push the boundaries of what’s possible.

Yes, alternative L1s offer simpler user experiences without fragmentation. Solana’s unified state and Hyperliquid’s integrated stack show the appeal of monolithic design. But Ethereum’s modular approach enables a Cambrian explosion of experimentation, different VMs, different consensus mechanisms, different trust assumptions, all inheriting Ethereum’s security. The market will ultimately decide whether flexibility or simplicity wins.

Is Ethereum scaling “done”? No, and it never will be. Technology doesn’t reach a finish line; it continues evolving to meet new demands and enable new possibilities. But Ethereum scaling is undeniably real, foundational, and entering its next phase.

The experimental era is over. The building era has begun. The infrastructure can now support applications we’re only beginning to imagine, and the Robinhood Chain is probably one of the best examples today. That’s not an ending, it’s finally the real beginning.