Why Bitcoin Miners Form Companies: What Blockchain Teaches Us About the Nature of Firms
The economics of mining consolidation isn't a bug—it's a century-old theory playing out in fast-forward.
In 1937, Ronald Coase asked a deceptively simple question: if markets are so efficient at coordinating activity through prices, why do firms exist at all? Why don’t we all just freelance, contracting with each other transaction by transaction? His answer—that organising activity inside a firm saves on the costs of using the market—launched an entire field of institutional economics and eventually won him a Nobel Prize.
Nearly a century later, that same question has a new laboratory. Not factories or franchises, but Bitcoin mining.
Between 2009 and 2019, Bitcoin mining went from hobbyists running software on laptops to vertically integrated corporations with industrial-scale data centres, negotiated power contracts, and hierarchical management structures. Critics often frame this consolidation as a failure—proof that decentralisation was always a fantasy. But viewed through the lens of institutional economics, something more interesting is happening. Mining consolidation isn’t a pathology. It’s Coase’s theory of the firm, running in fast-forward, in an environment stripped of all the historical baggage that normally makes firm formation hard to study.
This post draws on a recent paper that applies transaction cost economics and mechanism design to proof of work systems. The paper makes two core arguments: first, that mining consolidation follows inevitably from the cost structure of the environment, producing firms of a determinate optimal size; and second, that the governance legitimacy of these systems rests on a novel concept—dynamic legitimacy—that has implications well beyond blockchain.
Let’s unpack both.
The Problem With Voting When Nobody Has an ID
Before we get to firms, we need to understand the strange world that proof of work systems inhabit. Bitcoin operates under a set of constraints that have no real parallel in conventional institutions:
Anonymity. Participants are identified by cryptographic keys, not legal names. There’s no verified link between a key and a real person.
Permissionlessness. Anyone can join or leave at any time. No application, no approval, no licence required.
Sybil resistance. The system must be robust against one person creating a thousand fake identities.
Oracle independence. There’s no trusted external authority—no court, no regulator, no identity bureau—to verify who’s who.
These four conditions define the design space. They tell us what kinds of governance are even possible before we start asking which ones are desirable.
And the first thing they rule out is democracy.
One-person-one-vote requires knowing how many people there are, and ensuring each one votes exactly once. Under anonymity, that’s impossible. Anyone can generate new cryptographic keys for free. Without an external authority to certify identities, there’s no way to prevent a single actor from creating a million accounts and casting a million votes. This is the Sybil attack, identified by John Douceur in 2002, and it’s not a technical limitation waiting to be solved—it’s a logical consequence of the design constraints.
The implication is profound: any governance mechanism in this environment must weight participation not by identity but by a costly, rivalrous signal. Something you can’t fake cheaply, and something you can’t use in two places at once. Computational work (proof of work) and staked capital (proof of stake) are the two main candidates. The choice between them, as we’ll see, has deep consequences.
Why Miners Form Firms: The Coasian Logic
Now to the central question. If Bitcoin was designed for decentralised participation, why did miners consolidate into firms?
The answer is transaction costs. Consider the cost structure facing a miner deciding whether to operate alone or join an organisation of size s. Three forces are at work:
Market transaction costs fall with scale. Large mining operations negotiate industrial electricity rates of 2–4 cents per kilowatt-hour; a solo miner pays residential rates of 10–20 cents. Bulk hardware procurement yields discounts. Shared network infrastructure spreads fixed costs. These aren’t exotic blockchain phenomena—they’re the same economies of scale that drive consolidation in manufacturing, agriculture, and services.
Organisational costs rise with scale (and accelerate). Managing a larger operation requires hierarchy, monitoring, and coordination. Communication overhead grows. Decision-making slows. This is the standard Coasian countervailing force: at some point, the cost of organising one more transaction internally exceeds the cost of doing it through the market.
Revenue variance falls with scale. This one is specific to mining’s probabilistic structure. A solo miner expecting to find one block per week might go months without income—the reward follows a Poisson process with brutal variance. Pooling hash power across many miners reduces per-member variance dramatically. This lowers the risk premium on capital and shrinks the cash buffer needed to survive dry spells. It’s the same logic that drives agricultural cooperatives and insurance pools.
A fourth factor reinforces all three: mining hardware exhibits extreme asset specificity. A SHA-256 ASIC—the specialised chip used for Bitcoin mining—has essentially zero value outside Bitcoin mining. It’s the purest possible case of what Oliver Williamson called “physical asset specificity.” An independent operator with millions invested in single-purpose hardware is dangerously exposed to hold-up risk from counterparties. Vertical integration—bringing suppliers and operations under one roof—is the textbook response.
Put these four forces together and you get a determinate equilibrium. The optimal firm size sits where the marginal savings from scale economies and variance reduction exactly equal the marginal cost of increased bureaucracy. Firms smaller than this optimum leave money on the table; firms larger than it drown in coordination overhead.
The resulting market structure is a long-tail distribution: a handful of large firms operating near the efficiency frontier, alongside a tail of smaller operations with thinner margins. This isn’t monopoly, and it isn’t atomistic competition. It’s oligopolistic competition among firms of varying sizes—exactly what the Coasian framework predicts for an industry with these cost characteristics.
And here’s what makes it a compelling case study: because the Bitcoin mining ecosystem grew from scratch, without pre-existing legal frameworks, inherited corporate structures, or cultural path dependencies, we can watch institutional evolution in real time. CPU mining gave way to GPUs, then to FPGAs, then to ASICs—each wave a Schumpeterian episode of creative destruction that reshuffled the industry while preserving the underlying distributional form. It’s the cleanest natural experiment in firm formation that institutional economists have ever had.
Comparative Advantage Inside the Firm
There’s a nice illustration of why specialisation emerges even among miners with unequal abilities. Imagine Alice generates £500 per hour from mining and Bob generates £120 per hour. Alice also happens to be better than Bob at administration. Working separately, each spending 10 of their 50 hours on admin, their combined output is £24,800.
Now suppose they form a firm. Alice mines full-time. Bob handles 20 hours of joint administration and spends 30 hours on support tasks. Combined output jumps to £28,600. The Ricardian principle of comparative advantage holds even when one party has absolute advantage at everything: any marginal skill differential creates gains from trade. This is why firms develop internal specialisation, and why mining operations develop dedicated roles for procurement, facility management, and operations alongside the core mining function.
Dynamic Legitimacy: A New Kind of Governance Authority
The consolidation story is descriptive. But it raises a normative question: is the resulting governance arrangement legitimate? In conventional settings, legitimacy comes from external sources—democratic mandates, constitutional authority, legal frameworks. Permissionless systems have none of these. So where does legitimacy come from?
The paper proposes five axioms for governance legitimacy, each derived from the structural constraints of the environment:
Sybil-proof weighting. Governance influence must be proportional to a signal that can’t be faked cheaply. (Otherwise, create a million identities and dominate.)
Rivalrous commitment. The signal must be rivalrous—it can’t support contradictory positions simultaneously. (Otherwise, endorse every outcome and the signal is meaningless.)
Temporal non-persistence. Your governance authority today must depend on what you’re committing today, not what you committed last year. (Otherwise, incumbents entrench and the system ossifies.)
Open contestability. Anyone must be able to acquire governance authority by making the required commitment at any time. No licences, credentials, or prior relationships as barriers. (Otherwise, permissionless becomes permissioned.)
Incentive alignment. The cost of participation must be borne at the same time governance authority is exercised, so participants internalise the consequences of their decisions.
These five axioms, taken together, uniquely require what the paper calls dynamic legitimacy: governance authority that is proportional to costly, verifiable, rivalrous resource commitment right now, where past commitments confer no current authority, and the system is open to any entrant willing to pay the cost.
This is what proof of work does. Every block mined requires fresh expenditure of electricity and computation. Yesterday’s mining buys you nothing today. The moment you stop spending, your governance influence drops to zero. Authority is continuously earned through ongoing costly commitment.
And this concept—authority through continuous expenditure rather than accumulated position—is genuinely novel in governance theory. It’s distinct from property-rights authority (where power derives from ownership), constitutional authority (where power derives from agreed rules), and evolutionary authority (where power derives from routinised competence). Dynamic legitimacy posits that authority can derive from ongoing costly action, with its legitimating force extinguishing the instant the action ceases.
The Proof of Stake Comparison
The contrast with proof of stake systems illuminates what’s distinctive about dynamic legitimacy.
In proof of stake, governance weight derives from staked capital rather than ongoing expenditure. On the surface, this looks similar—you need resources to participate. But the institutional dynamics diverge in important ways.
Differential reinvestment. Staking rewards are proportional to stake. When reinvestment rates differ—and they will, because smaller participants have higher consumption needs relative to their holdings—concentration increases over time. This isn’t a bug in any particular implementation; it’s a general property of compounding under wealth heterogeneity.
Delegation economies. Smaller holders delegate to large validators for a fee. Larger validators can offer lower fees because of economies of scale, attracting more delegation, which further increases their governance weight. This is a classic increasing-returns dynamic that produces winner-take-most outcomes.
Endogenous governance capture. Here’s the most subtle problem. Large validators have disproportionate influence over protocol parameters: reward structures, minimum stake thresholds, slashing conditions, delegation rules. A dominant coalition can adjust these parameters to disadvantage smaller validators and new entrants—not by breaking the rules, but by writing them. This is George Stigler’s regulatory capture theory, extended to settings where there’s no external regulator. The governance system captures itself.
The critical violation is temporal non-persistence. In proof of stake, capital staked in the past continues to generate governance weight in the present with low marginal cost. A stakeholder who acquired their position years ago retains authority simply by holding, not by actively committing resources. Past accumulation translates directly into present power.
You might think this could be fixed by introducing stake decay—automatically reducing governance weight over time unless actively renewed. But who sets the decay rate? The existing large stakeholders, through the very governance process that’s supposed to be reformed. The fix reintroduces the problem at the meta-governance level.
The comparison isn’t between proof of work and perfection. It’s between two actually available institutional alternatives under a specific set of constraints. Neither achieves an ideal. But proof of work satisfies the governance axioms in ways that proof of stake structurally cannot, within the permissionless design space.
Mining Pools: Exit as Governance
One of the more interesting institutional forms in the mining ecosystem is the pool. Mining pools are cooperatives where miners contribute hash power and share rewards, smoothing the brutal variance of solo mining. Critics often point to large pools as evidence of dangerous centralisation.
But this conflates the pool operator with the pool membership. Pool members retain the ability to exit—to redirect their hash power to a competing pool—at minimal switching cost. This creates governance through what Albert Hirschman called exit rather than voice. An operator who pursues unpopular policies—attempting to censor transactions, for instance—faces rapid membership departure. The operator’s authority is delegated, continuously revocable, and disciplined by competitive pressure.
The analogy with political parties is instructive. Citizens delegate political voice to parties but retain the right to change affiliation. Calling a mining pool a “centralised actor” is like equating a political party’s leadership with its voter base. The pool’s behaviour reflects aggregate member preferences, disciplined by the ever-present threat of defection.
What This Means Beyond Blockchain
The real payoff of this analysis lies outside blockchain.
A new governance category. Dynamic legitimacy identifies a type of governance authority—derived from ongoing costly commitment rather than historical accumulation, democratic mandate, or expert credential—that doesn’t fit neatly into existing frameworks. It may have analogues in conventional settings: the authority of a continuously investing corporate parent over subsidiaries, or the legitimacy of a state’s governance conditional on ongoing service provision. The concept gives institutional economists a new lens.
Institutional emergence under extreme constraints. Proof of work systems offer something rare: a natural experiment in institution-building from scratch, without pre-existing authority structures, legal scaffolding, or cultural inheritance. The rapid emergence of pools, firms, and hierarchies from an initial state of atomistic participation confirms Coasian predictions with unusual clarity. The compressed timescale—a single decade from individual hobbyists to vertically integrated corporations—makes institutional evolution directly observable.
The problem of endogenous governance capture. The mechanism identified in the proof of stake analysis—incumbents entrenching their position by shaping governance rules rather than violating them—applies far beyond blockchain. Cooperatives, mutual organisations, professional associations, and any member-governed institution face the same structural risk. The concept of endogenous governance capture, where the governance system captures itself without any external regulator involved, deserves wider attention.
Comparative institutional method. Perhaps the most general lesson is methodological. The paper illustrates Williamson’s principle that institutional analysis should compare discrete feasible alternatives rather than measuring each against an unattainable ideal. Specifying the feasibility constraints first, then comparing what’s actually available within those constraints, is a discipline that applies to any institutional design question—not just blockchain governance.
The Energy Question
No discussion of proof of work is complete without addressing energy consumption. The environmental critique is real and deserves serious engagement. But the institutional analysis reframes it: energy expenditure in proof of work isn’t waste—it’s the cost of producing governance legitimacy under the constraints of a permissionless system. Just as democratic elections have administrative costs and judicial systems require expenditure on courts and personnel, proof of work’s energy consumption is the price of satisfying the governance axioms in an environment without trusted authorities.
Eliminating energy expenditure without substituting another costly signal doesn’t save resources—it reintroduces the Sybil vulnerability that the entire system is designed to prevent. The question isn’t whether the energy cost is zero, but whether what it buys—a governance mechanism satisfying dynamic legitimacy—is worth the price.
Conclusion
Bitcoin mining consolidation isn’t a story of decentralisation’s failure. It’s a story of institutional economics’ predictive power. Coase told us in 1937 that firms form where market transaction costs exceed the costs of internal organisation. Williamson told us in 1985 that asset specificity, transaction frequency, and uncertainty drive vertical integration. Mining exhibits all of these, and the organisational structures that emerged are exactly what the theory predicts.
The deeper contribution is the concept of dynamic legitimacy—governance authority that must be continuously earned through costly commitment, that vanishes the moment commitment ceases, and that remains perpetually open to challenge by any willing participant. It’s a governance type we haven’t had a good name for, operating in an environment that strips away the institutional scaffolding we usually take for granted. Understanding it doesn’t just illuminate blockchain. It gives us new tools for thinking about authority, legitimacy, and institutional design wherever governance must operate without the safety net of external enforcement.
The firms that emerge in proof of work systems aren’t a betrayal of decentralisation. They’re what decentralisation looks like when it meets the real world of transaction costs, risk management, and comparative advantage. Coase would have recognised them immediately.