💻 computer science

Geographical Centralization Resilience in Ethereum's Block-Building Paradigms

This paper develops a formal model and simulations to demonstrate that Ethereum's block-building paradigms create non-neutral geographical incentives, driving validators to strategically co-locate in low-latency regions like the Atlantic corridor, thereby increasing the network's vulnerability to regional shocks and centralization.

Original authors: Sen Yang, Burak Öz, Fei Wu, Fan Zhang

Published 2026-04-07

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Sen Yang, Burak Öz, Fei Wu, Fan Zhang

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine Ethereum not just as a digital ledger, but as a massive, global relay race where thousands of runners (called validators) take turns carrying the baton (creating a new block of transactions). The goal is to pass the baton as fast and fairly as possible so the race never stops.

This paper asks a simple but critical question: Does the location of these runners matter?

The authors argue that it matters a lot. Even though the rules of the race don't say "runners must be spread out," the way the race is structured naturally pushes runners to crowd into the same few cities. This creates a "geographical centralization" that makes the whole system vulnerable. If a natural disaster or a government shutdown hits that one city, the whole race could stop.

Here is the breakdown of their findings using simple analogies:

1. The Two Ways to Run the Race

The paper compares two different ways a runner can prepare their baton (the block):

Local Building (The DIY Runner): The runner gathers their own data, builds the baton themselves, and then runs to the finish line to hand it off.
- The Incentive: To win, you need to be close to the data sources (to get the best info fast) AND close to the judges (to hand off the baton before the clock runs out).
- The Result: Runners flock to the "perfect hub" where data is fast and judges are nearby. This creates a massive crowd in one spot (like the US and Europe).
External Building (The Outsourced Runner): The runner doesn't build the baton. They hire a professional "Builder" to do it for them. The runner just signs the finished baton and hands it over.
- The Incentive: The runner only needs to be close to the Builder. Once the Builder has the baton, the Builder is responsible for running it to the judges.
- The Result: Runners flock to wherever the Builders are located. If the Builders are in a remote, high-latency area, runners will move there too, just to be close to them.

2. The "Hotelling's Law" of the Internet

The authors use a concept called Hotelling's Law. Imagine two ice cream vendors on a beach. To get the most customers, they both move to the exact middle of the beach, right next to each other, even though the customers are spread out everywhere.

In Ethereum, validators do the same thing. They realize that being slightly faster than everyone else (by being in the same city as the data or the builder) means they get more money. So, they all move to the same "middle of the beach," leaving other parts of the world empty.

3. The "Speed Trap" of Shorter Time Limits

The paper also looks at what happens if the race gets faster (shorter time slots).

The Analogy: Imagine a race where you have 12 seconds to finish versus one where you only have 6 seconds.
The Effect: In the 6-second race, a tiny 0.1-second advantage becomes huge. If you are 0.1 seconds closer to the finish line, you win. If you are far away, you lose.
The Consequence: Shortening the time slots makes the runners even more desperate to be in the perfect location. It amplifies the crowding effect.

4. The "Information Black Hole"

The paper found that if the "data sources" (the information the runners need) are all in one place, the runners will all move there.

Local Building: If data is in London, runners move to London.
External Building: If the "Builders" are in a remote, slow internet region, runners might actually move to that remote region just to be close to the Builder, even if it's a bad place to live. This is counter-intuitive but true: being close to the person who does the work matters more than being in a nice city.

5. Why Should We Care? (The "Single Point of Failure")

Why is this crowding bad?

The "All Eggs in One Basket" Problem: If 80% of the runners are in the US and Europe, and a massive power outage hits the Atlantic corridor, or a government bans crypto in those specific countries, the network could freeze.
Unfairness: Runners in Africa or South America are at a permanent disadvantage. They are slower, so they win less often, so they make less money, so they can't afford to upgrade their equipment. It becomes a cycle of inequality.

The Big Takeaway

The authors conclude that Ethereum is not geographically neutral. The rules of the game accidentally encourage everyone to move to the same few cities.

What can be done?

Change the Rules: Adjust the timing rules so that being slightly faster doesn't give such a huge reward.
Diversify the Builders: Encourage Builders to be spread out all over the world, not just in one hub.
Burn the Extra Money: If the extra money gained from being fast is "burned" (destroyed) rather than kept, runners won't be as desperate to move to the perfect city.

In short: Decentralization isn't just about having many people; it's about having them in many places. If they all live in the same neighborhood, the system is fragile. This paper shows us how the current rules of Ethereum are accidentally pushing everyone into the same neighborhood, and suggests how we might fix the rules to spread them out again.

1. Problem Statement

Decentralization is a core security assumption of permissionless blockchains, yet conventional metrics (e.g., stake distribution) often overlook the geographical dimension. The paper addresses the critical issue that validator locations are not encoded in protocol rules but emerge from economic incentives, infrastructure, and regulatory constraints.

The Core Issue: Validators strategically co-locate to minimize latency to payoff-relevant parties (information sources, block suppliers, or other validators). This creates "location games" (reminiscent of Hotelling's law) that lead to geographical centralization.
Risks: High geographical concentration makes the network vulnerable to regional shocks (natural disasters, power outages, government bans) and creates unfair reward disparities due to latency advantages.
Specific Context: Ethereum is transitioning between two block-building paradigms:
1. Local Block Building: Validators construct blocks themselves using distributed signal sources.
2. External Block Building (PBS/ePBS): Validators outsource block construction to specialized "builders" (suppliers).
Gap: Existing studies provide empirical snapshots of current validator locations but lack a unified framework to predict how protocol changes (like enshrined PBS or shorter slot times) will alter future geographical incentives.

2. Methodology

The authors propose a unified analytical and computational framework combining formal game theory with agent-based modeling (ABM).

A. Formal Model

Agents: Validators are autonomous agents with stakes and geographical locations.
Environment:
- Geography: Modeled as discrete regions with log-normal distributed propagation latencies (calibrated to real-world Google Cloud data).
- Information Sources:
  - Signal Sources: Provide partial block value (e.g., price feeds).
  - Suppliers: Provide full blocks (in External Building).
- Consensus: Time is divided into slots. A block becomes canonical if a threshold ( $\gamma$ ) of attestations is received before a cutoff time ( $\tau_{cut}$ ).
Strategies: Validators optimize two variables:
1. Timing: When to release the block to maximize value while meeting the attestation deadline.
2. Migration: Whether to move to a different region to reduce latency, weighing the payoff gain against migration costs.
Paradigms Modeled:
- Local: Payoff depends on latency to all signal sources and all attesters.
- External: Payoff depends on latency to the selected supplier and the supplier's latency to attesters.

B. Analytical Characterization

The authors use a mean-field approximation to derive qualitative predictions about the direction and scaling of location incentives. They prove the existence of pure Nash equilibria and establish monotonicity theorems showing that reducing latency to payoff-relevant parties strictly increases expected rewards.

C. Agent-Based Simulation (ABM)

To quantify these incentives under heterogeneous conditions, the authors built a simulation framework (using Mesa/Python) calibrated with:

Real-world Latency: Inter-region round-trip times from Google Cloud.
Parameters: 1,000 validators, 10,000 slots, varying consensus parameters ( $\gamma$ , slot duration $\Delta$ ).
Experiments: Four distinct experimental setups (EXP) isolating the effects of:
1. Information-source placement.
2. Initial validator distribution.
3. Joint heterogeneity (both validators and sources).
4. Consensus parameter variations.

3. Key Contributions

Formal Model of Geographical Incentives: The first unified model capturing the interaction between Ethereum's block-building paradigms, validator distributions, and information source placement.
Analytical Proofs: Derivation of theorems proving that both paradigms generate location-dependent payoffs, but with different scaling properties:
- Local Building: Incentives scale with the number of signal sources.
- External Building: Incentives are bottlenecked by the single best supplier.
Counterfactual Analysis: Using ABM to simulate how protocol changes (e.g., shorter slots, different attestation thresholds) impact geographical concentration, which empirical data alone cannot predict.
New Metrics: Introduction of geographical-specific metrics including the Geographical Gini, Geographical HHI, Geographical Payoff Coefficient of Variation, and a Geographical Liveness Coefficient (Nakamoto-style).

4. Key Results

A. Non-Neutrality of Architecture

Ethereum's block-building architecture is not geographically neutral. Both paradigms incentivize validators to migrate toward low-latency hubs, but the mechanisms differ:

Local Building: Validators cluster near regions with many signal sources and low latency to attesters. Migration incentives are strong and scale with the number of sources.
External Building: Validators cluster near block suppliers. Interestingly, if a supplier is located in a high-latency (remote) region, the incentive for validators to co-locate with that specific supplier is stronger because the latency penalty for non-co-location is higher.

B. Impact of Information Source Placement

Asymmetry Amplifies Centralization: When information sources are unevenly distributed, centralization accelerates.
Opposing Sensitivities:
- In Local Building, placing sources in low-latency hubs causes rapid centralization.
- In External Building, placing suppliers in high-latency regions causes rapid centralization (as validators rush to close the latency gap with the supplier).

C. Impact of Validator Distribution

Path Dependence: If validators are already concentrated (as in the current Ethereum network), both paradigms lead to rapid convergence to a saturated equilibrium where further migration yields diminishing returns.
Reward Variance: Local building tends to have higher reward variance during the transition, while external building shows sharper disparities once a dominant supplier region is established.

D. Consensus Parameter Effects

Attestation Threshold ( $\gamma$ ):
- Higher $\gamma$ : Dampens migration in Local Building (forces a balance between source proximity and attester proximity) but amplifies migration in External Building (makes the proposer-supplier latency bottleneck more critical).
Slot Duration ( $\Delta$ ):
- Shortening slots (e.g., from 12s to 6s) does not significantly change the degree of centralization but increases the normalized reward disparity. A fixed latency advantage becomes a larger fraction of the available time window, making geographical location relatively more valuable.

5. Significance and Implications

Protocol Design: Geography is an endogenous outcome of protocol design. Parameters like slot time and attestation thresholds act as "levers" that can unintentionally amplify geographical centralization.
Security Risks: The findings suggest that current trends toward external block building (PBS/ePBS) do not eliminate geographical centralization; they merely shift the locus of concentration from "validator hubs" to "supplier hubs."
Mitigation Strategies:
- Decentralized Block Building: Moving toward designs with multiple concurrent proposers (MCP) or decentralized builder networks (e.g., BuilderNet) could reduce the monopoly power of a single location.
- MEV-Burn: Mechanisms that burn MEV rewards could reduce the economic incentive to migrate for latency advantages.
- Supplier Diversity: Protocol-level incentives or governance norms encouraging geographical diversity among builders and relays are crucial.
Conclusion: Geographical decentralization cannot be assumed; it must be actively managed through protocol design and infrastructure diversity to preserve Ethereum's resilience against regional shocks.