Score: 0

Designing Ethereum's Geographical (De)Centralization Beyond the Atlantic

Published: September 25, 2025 | arXiv ID: 2509.21475v1

By: Sen Yang , Burak Öz , Fei Wu and more

Potential Business Impact:

Makes computer networks spread out more evenly.

Business Areas:
Ethereum Blockchain and Cryptocurrency

Decentralization has a geographic dimension that conventional metrics such as stake distribution overlook. Where validators run affects resilience to regional shocks (outages, disasters, government intervention) and fairness in reward access. Yet in permissionless systems, locations cannot be mandated, but they emerge from incentives. Today, Ethereum's validators cluster along the Atlantic (EU and U.S. East Coast), where latency is structurally favorable. This raises a key question: when some regions already enjoy latency advantages, how does protocol design shape validator incentives and the geography of (de)centralization? We develop a latency-calibrated agent-based model and compare two Ethereum block-building paradigms: a Single-Source Paradigm (SSP), akin to MEV-Boost, where proposers fetch full blocks from a relay that also propagates them; and a Multi-Source Paradigm (MSP), where proposers aggregate value from multiple sources and broadcast the block themselves. Simulations show that SSP concentrates around relay placement but more slowly, since proximity mainly affects propagation, and the marginal value of time is relatively uniform across regions. MSP centralizes faster: aggregating across sources makes marginal value location-dependent, amplifying payoff dispersion and migration toward latency minima. Source placement and consensus settings can dampen or intensify these effects, though once validators are already clustered, the impact of source placement on decentralization is marginal. In most cases, North America consistently emerges as the focal hub. These findings show that protocol design materially shapes validator geography and offer levers for promoting geographical decentralization.

Page Count
29 pages

Category
Computer Science:
Cryptography and Security