From fighting sandwiching to democratizing trading opportunities, the multiple concurrent leaders model is full of promises. But there is one underappreciated advantage: it makes geography an explicit dimension of competition and so organically incentivizes a decentralized chain. To those who dream of expanding validator sets to the frontiers of the world, this upgrade over the single leader model of today is the solution.
Centralizing forces on validators
Every validator faces the same calculus: where should it locate? There are three factors in answering this question.
- Validators want to locate in areas that offer cheap electricity, available hardware, favorable tax and regulatory regimes, etc.
- Validators want to be sufficiently close to one another, so that they will be able to participate in consensus without skipping slots or missing votes.
- Validators want to locate near sources of information shocks, to maximize the value of the transactions in their blocks.
The first two factors are explicitly centralizing forces. Data centers that serve one validator well are likely to serve all validators well. Moreover, as validators accumulate in one area, the marginal validator is incentivized to remain relatively close. Indeed, approximately half of Solana’s stake is in the three neighboring cities of London, Amsterdam, and Frankfurt.
But the third factor — on sources of information — is neither intrinsically centralizing nor decentralizing. It critically depends on the model of block proposers.
Distributed information, distributed leaders
When it comes to information shocks, the leader paradigm matters. In a single leader model, validators generally make identical decisions. In a multiple leader model, validators explicitly spread out to find new terrain.
To illustrate, suppose there are two cities, separated by a link that takes 200ms to traverse (one-way). The cities generate market-moving news in random ways, but their distributions are identical. In a single leader model (where, say, the leader slot is 400ms long), the leader is indifferent to where they situate themselves. If they situate themselves in the middle, as in the diagram below, they capture all shocks for the first 300ms of their block (plus residual uncaptured shocks from the previous slot); any shock generated after that point does not make it to the leader in time. If they situate themselves closer to one city, they capture shocks for more than 300ms in that city and less than 300ms in the other city. Any solution gives the same answer and so they are indifferent. Only the other factors — cheap power and co-location with peers — matter, and they are explicitly centralizing factors.
But in a two-leader model, a unique solution emerges: regardless of where one validator sits, the other validator is incentivized to migrate to co-locate in the further city. This quickly settles into the equilibrium below, wherein the two validators have spread out from one another. This factor now is an explicitly-decentralizing force, in contrast to the other factor.
The model in this section may be stylized, but it illustrates an important new dimension of competition. In the single leader model, a validator has a local monopoly on information and so its exact location — as long as it is in some geographically central spot — is not particularly important. It captures all shocks that take place during its slot, regardless of where it sits.
In the multi-leader model, that monopoly is gone. Validators directly compete with one another, finding new informational founts that are under-covered by their peers. This allows them to include novel transactions to the system, which (assuming a reasonable model of merging blocks in the same slot) are its primary edge. Two leaders in Frankfurt miss the same shocks and compete away profits on the shocks they don’t miss; whereas two leaders in Frankfurt and Singapore do not. Decentralization is individually profitable.
This effect gets stronger as more leaders compete. Two-leader systems may lead to validators congregating in the same few financial hubs around the globe; but systems with dozens of leaders will provide strong incentives for validators to find new ground, as each one attempts to maintain its edge. Finally, as the validator set spans the globe, one of the centralizing forces — the need to co-locate with other validators to participate in consensus — becomes muted, and decentralization gets yet another tailwind.
Dynamic Beyond Tradfi’s Capabilities
The multi-leader model incentivizes a decentralized network, and so incentivizes a robust network. But there is one final advantage of this model, which is that this decentralization is dynamic, in ways that traditional exchanges struggle to match.
Suppose the toy model is augmented with a third city that rises in prominence and eclipses the first city. The validator in the first city is incentivized to move to the third city, to remain close to information shocks and to maintain an edge in the transactions it submits to the network. The migration is easy — simply spinning down a machine and spinning up a new one — and does not require approval from or coordination with market participants or other validators.
This is a network that is dynamic: as soon as the world landscape changes, validators are incentivized to change with it. Traditional finance is much slower: migrating cities requires substantial investment and coordination, and so too often old stock markets remain anchored in cities long after their financial base has been hollowed out. Multi-leader systems are much nimbler, and that ensures that the particular topology of a network never gets ossified.
Organic Decentralization Beats Any Other Kind
While decentralization may be a lofty virtue, it is all too often neglected in practice. It requires validators to make deliberately uneconomic decisions — situating in places with pricier infrastructure, poorer connectivity, or unfavorable regulatory regimes. Foundations or communities can subsidize these sorts of moves temporarily, but they cannot fix the weak business case.
But multi-leader systems turn that limitation into a strength. They make validators compete on geography, and in doing so, they generate decentralization organically and sustainably. This is the paradigm that can strengthen the core value proposition of blockchains everywhere.