Expand capacity vs reduce speed: Why Ethereum prioritizes bandwidth to ensure security

Vitalik Buterin recently shared on social media a technical analysis arguing that Ethereum’s strategy should focus on increasing bandwidth rather than solely pursuing latency reduction. His argument is based on considerations of decentralized security in the long term.

The feasibility of scaling through bandwidth

With the implementation of PeerDAS and zero-knowledge proofs (ZKP), the community has clearly identified a scalable path. Theoretical calculations suggest that, maintaining the current model, capacity can be expanded by a thousand times. Buterin emphasizes that both security parameters and decentralization remain favorable under this approach, and physical laws do not impose fundamental restrictions to coexist “extreme scalability + robust decentralization.”

The physical limitations of reducing latency

The situation is radically different when the goal is to minimize propagation times. The speed of light represents an insurmountable barrier, but beyond that, there are deep practical constraints affecting the network’s economic viability:

Geographical distribution of nodes: Operators (validators and proposers) must operate in rural areas, homes, and conventional commercial facilities, not exclusively in highly optimized data centers.

Resistance requirements: The network requires anonymity and resistance to global distributed censorship, impossible to guarantee with concentrated infrastructure.

Economic sustainability: If staking outside major centers yields 10% less performance, economic pressure will inevitably centralize operations in those few locations. In the long run, this dynamic would erode decentralization.

Proof of abandonment: Ethereum must pass a critical test: operate without relying on continuous social coordination to maintain decentralization. Economic incentives should assume “most” of the responsibility, not all.

A realistic path: incremental improvements without sacrificing decentralization

Without making fundamental compromises, Buterin suggests significant optimizations:

Innovations in P2P networks, particularly optimized erasure codes, can reduce propagation times without forcing individual nodes to increase their consumed bandwidth.

A more efficient availability chain, using fewer nodes per slot (512 instead of 30,000), would eliminate aggregation steps, allowing the critical path to complete within specialized subnets.

These improvements could achieve latency reductions of 3 to 6 times. Under this scenario, reaching latencies of 2 to 4 seconds is entirely feasible.

Why Ethereum cannot be a global gaming server

Buterin posits a fundamental conceptual distinction: Ethereum functions as the “global heartbeat” of the digital economy, not as an entertainment infrastructure. Applications demanding speeds faster than this “heartbeat” must necessarily incorporate off-chain components.

This reality explains why layer 2 solutions will persist even if Ethereum achieves massive scalability. Beyond technical reasons (virtual machine customization and demand for even more extreme scale), there is a fundamental reason: AI-driven applications will require “city-level” or even building-level blockchains.

If an AI processes 1000 times faster than humans, its “subjective speed of light” would be only 300 km/s, enabling almost instant communication within a metropolis, but not between cities. These needs cannot be met at the base layer. Similarly, running staking validators on Mars would be prohibitively expensive—a consideration that even Bitcoin does not pursue.

Conclusion: Ethereum remains grounded on Earth

Vitalik’s conclusion is that Ethereum fundamentally belongs to Earth. Its architecture should be optimized for coexistence of maximum decentralization and maximum capacity through scalable bandwidth, while its layer 2 solutions serve both ultra-localized needs within cities and applications operating on a planetary scale.