Competition, Concentration & the Ownership Question

Concentration is the default. Not a moral failing of any firm, not a prophecy, just the ordinary gravity of platform economics. Where network effects and data flywheels compound, and a bottleneck cannot be routed around, value pools toward whoever sits at the bottleneck, and keeps pooling. Left alone, the agentic economy will follow that gravity the way every platform era did, only faster and at greater scale, because its bottlenecks are denser and its returns to scale steeper. Equalization is the alternative. It never arrives on its own. It is a buildable choice, made or forgone in the design of standards, the structure of ownership, and the governance of the chokepoints. One outcome is automatic. The other has to be engineered.

Two fusions form the spine of what follows. First, the labor question is the ownership question. The fear that agentic labor immiserates working people is not really about how many people have jobs. It is about who owns the capital that accretes the value. The same automation is catastrophe under concentrated ownership and shared abundance under broad ownership. Nothing decides between them except how the capital stock is distributed. Second, concentration is geopolitics. The chokepoints that pool rents inside an economy are the same ones weaponized across borders. A settlement rail, an identity root, or a reserve asset that lets its controller extract is the same asset that lets it exclude, freeze, and surveil. The competition problem and the national-security problem are one problem seen from two angles. That is why governing the chokepoints, rather than letting them be captured, is the through-line of this domain and the hinge into the global-coordination domain that follows.

Competition, antitrust, and power at non-forkable chokepoints

The claim is structural, not rhetorical. The history of open protocols and commoditization is full of cases where power dispersed instead of pooling. So state the disciplined version. Concentration wins at layers that combine two things: strong increasing returns, where each new user makes the product better for the next, and a non-forkable bottleneck. Both conditions matter. Increasing returns without a bottleneck produce a leader who can still be unseated. A bottleneck without increasing returns produces a toll that competition erodes. Only where they fuse does a moat form that ordinary competition cannot cross. The forkability test is concrete. You can fork open-source code. You cannot fork the dominant currency, a regulatory license, a deep liquidity pool, an override key, or a reused identity root. So the question is never whether the agentic economy concentrates. It is which layers sit on non-forkable bottlenecks, answered layer by layer, because conflating them is how competition policy aims at the wrong target.

Begin where the popular intuition is most confidently wrong. The foundation-model layer, imagined as the great toll booth of the era, is in fact among the most likely to commoditize. Frontier capability has repeatedly been matched by open weights within a year or two of release, and inference prices have fallen by orders of magnitude and keep falling. Where capability is cheap and replicable, durable rent migrates off the raw model and onto its complements: proprietary data, distribution, and the lock-in of an agent that knows you and would be costly to leave. The intelligence becomes a falling-cost input. The moat moves elsewhere. Bridges, the connective tissue between chains, are a security story, not a rent story: historically the most-exploited surface in the space, to be hardened rather than broken up. Oracles, the feeds that bring outside data onchain, concentrate around whatever source everyone already trusts, but they stay forkable, and a corrupted one can be replaced. None of these is where durable power sits.

Durable power sits at the identity layer and the override key. Identity concentrates because one verified identity, once established, is reused everywhere. Reuse produces winner-take-most dynamics that no decline in inference cost will erode. The verification is the moat, not the compute. The override is deeper still. Whoever can reverse the deterministic core of a system, freeze it, or reach through it, in the limit controls the firm that runs on it. And an override is not forkable, because forking the code does not fork the key. Token governance, sold as democratization, is plutocracy by the plain construction of one-token-one-vote, unless it is deliberately engineered against that gravity with mechanisms not yet proven at scale. The naive default is that the largest holders govern. These are the layers where exit is hardest and capture most permanent, and where competition policy should be pointed.

The sharpest concern sits above all of these: the actor that holds both the economic and the intelligence infrastructure at once. An actor that supplies the agents and sits at the settlement or identity layer through which they transact does not just hold two moats. It can make each reinforce the other, steering economic activity toward its own intelligence, and feeding that intelligence with the data of the activity it intermediates. That dual position is the genuine frontier of concentration risk. Authorities that fight only the visible toll booths while that fusion forms will have aimed at the wrong target.

This diagnosis carries its own unlock, and it follows from the forkability test, not from hope. The architecture already contains, in its open and forkable layers, the property that defeats concentration: contestability. Good competition policy, paired with open, forkable, interoperability-mandated standards, keeps the chokepoints technically contestable. The threat of exit then disciplines whoever sits at them, even when they are not displaced. A chokepoint that can be left is a chokepoint that must behave. On that condition, the agentic economy could be the most equalizing order ever built. No prior era combined this scale of productive capacity with this degree of structural contestability. The chokepoints decide it. Governed, they discipline. Captured, they extract.

The work for policymakers follows from aiming at the durable layers, not the commoditizing ones. Antitrust and interoperability mandates should target identity, the override, settlement, and model-access (where forking fails) and leave the rest to the competition already eroding them. Standards at the durable layers should be open, forkable, and interoperability-mandated by rule, so contestability is a legal requirement and not a hope. Systemically critical infrastructure should carry public-interest governance: independent board seats, golden shares, and escrowed, multi-party, on-the-record control over the override and identity layers, so the most powerful switches are not private ones held by a single operator. Algorithmic-collusion risk (agents converging on supracompetitive prices with no explicit agreement) should be addressed before it is endemic, because pricing agents can coordinate at a speed and subtlety conventional antitrust was not built to catch. And the cloud–model "chokepoint deals" that braid compute providers and model labs into mutually entrenching partnerships should be scrutinized for what they can become. The current posture of the major competition authorities supports this direction.

One layer touches the author's own industry directly. The yield earned on the reserves backing payment stablecoins is, in its present form, a policy artifact: a rent that exists largely because statute forbids paying that yield to holders as interest. What statute creates, statute can redistribute. Reasonable policy could compete the rent, regulate it, or require more of it to be shared. And much of it is already shared. Leading regulated issuers return most of their reserve income to the ecosystem through distribution partners and usage-based participation, in pointed contrast to issuers that return nothing. So the sharpest risk is not the issuer layer, which already disperses most of its value. It is the dual-infrastructure position, where one actor holds both economic and intelligence infrastructure and the reinforcing moats compound.

One tension no governance design can dissolve. The monetary layer cannot be made neutral while one state issues the dominant unit and can reach through the rails to freeze, seize, or exclude. A dominant digital dollar is, by that same fact, both the most likely path to the scale that makes any of this real, and the single largest single-state-control risk in the whole architecture. That is not a contradiction to resolve away. It is one fact with two faces, and the credible posture states both at once. The protocol and technical layers can be genuinely neutral and forkable, and that neutrality is worth defending because it is achievable. The monetary layer carries a residue of state leverage that governance can constrain but not erase. That is why this domain cannot end at competition policy. It must continue into who owns the concentrating layers, and how the chokepoints are governed across borders.

The labor-versus-ownership question: the market structure of an agent-run economy

The serious claim about labor survives the oldest rebuttal in economics. It is not that automation destroys jobs on net. That "lump of labor" assumption has been wrong for two centuries. Displaced work has reliably been reabsorbed, and comparative advantage guarantees that even an agent absolutely better at everything is only relatively better at some things, leaving humans something to do. But comparative advantage speaks to relative productivity. It is silent on price. Humans can stay fully employed at the task where machines are comparatively weakest while the wage it commands falls below what sustains a household: formally full employment, socially a catastrophe. So the real claim is about the labor share of output and the wage at which human work clears. Both can fall under three conditions.

First, new tasks keep emerging, but software colonizes them faster than humans can be retrained, collapsing the historical lag between displacement and reabsorption toward zero. The agent is not just a better worker on today's tasks. It is a faster occupier of tomorrow's. Second, the marginal price of agentic labor tracks the falling cost of inference. Across more of the task frontier the machine is simply cheaper at the margin, dragging the human clearing wage down with it. Third (the genuine break from every prior wave), capital can fund its own expansion. A loom never earned the capital to buy the next loom, and never wrote the design for it. Agents can generate the surplus that finances more agents, and can author their successors. That makes the capital share self-reinforcing rather than bounded by the diminishing returns that limited every machine before. Where these three hold, the route by which most people have always claimed a share of output (their labor) narrows toward a trickle. Comparative advantage guarantees a task, not a living.

The decisive move is to see this as a distributional problem, not an output problem. The three conditions do not produce immiseration on their own. They produce a question of who owns the capital stock that is generating, by every indication, an enormous output. The same machinery is the catastrophe case and the abundance case. The only thing between them is who holds the capital it enriches. A person who owns a share of the machines shares in the abundance. A person who owns nothing is merely a more capable worker whose wage is still being dragged toward the cost of the agent beside them.

Because the problem is distributional, it is solvable by design, and for the first time the material exists to act at scale. The onchain substrate can distribute capital, governance, and upside to vast numbers of people at near-zero administrative cost, with digital tokens as the building block: the read, write, own ethos made operational rather than slogan. And the case is evidenced, not just aspirational. Leading regulated issuers already distribute most of their reserve value to ecosystem participants. Stakeholder-based network tokens are designed so utility, governance, and economic value flow across all participants (validators, application developers, and end users alike), embedding broad participation in the infrastructure itself rather than bolting it on afterward. These are living proof that broad participation can be built in from the start.

The work is design and fiscal, not one instead of the other. Reserve-yield sharing comes first and bites hardest, because it is the remedy that costs the author's own industry. The yield on stablecoin reserves should be competed toward, shared with, or returned to the people who hold the money, including by incumbents, the issuers among them, and the rule that manufactures the rent by forbidding holder interest should itself be revisited. Around it sit the rest: open, forkable, interoperability-mandated standards at the identity, settlement, and model-access layers, so chokepoints stay contestable; public-interest governance and escrowed, multi-party override over critical infrastructure; contribution- and usage-based ownership distribution (so the people who use and build a system come to own it) with vesting and transferability constraints and progressive caps to resist secondary-market re-concentration; and a fiscal instrument, a tax on automation or capital that funds a broad onchain ownership dividend, distributing a stake and not merely a transfer. The governance primitives should be non-plutocratic by construction (proof-of-personhood one-person-one-vote, quadratic and conviction voting, contribution-weighted rights) and pointed at the override and identity chokepoints, where durable power sits. Design distributes what the market will bear. Fiscal policy distributes what it will not.

Two facts sharpen this program, and both strengthen rather than soften the case for it. Distributing ownership is not the same as distributing power; the ordinary retail shareholder proves it, and onchain the gap can be worse through founder super-voting, foundation control, low-turnout delegation, and whale-dominated voting. Durable power sits not in the cash-flow claim but in the control chokepoints, so distributing governance is a separate and harder task than distributing upside, and it must be aimed squarely at the override and identity layers. And liquidity is the enemy of broad ownership: transferable tokens re-concentrate toward the largest holders the moment they have value, so broad ownership must accept transferability constraints. Both facts point the same way. They do not weaken the program. They specify it, telling a serious design exactly where to aim and what to constrain.

History sharpens the point rather than blunting it. The historic broad-ownership movements (syndicalism, the cooperative movement, guild socialism) were not defeated by a coordination deficit that better technology now repairs. They were defeated by power: by capital, employers, and the state, through hostile law, suppression, and force. The movements that succeeded prove the same point from the other side. Large worker cooperatives, credit unions, mutual insurers, and consumer cooperatives all reached distributed ownership at real scale, for decades, with sound legal form and durable institutions, and no blockchain at all. Onchain rails lower the cost of broad ownership and strip out gatekeepers, a real contribution, but they do nothing about the power asymmetry that actually killed the historic movements. The technology makes distribution cheaper. It does not make it happen. The path wins only through contested countervailing power: forkability and open standards that make capture defeatable by exit, antitrust and public mandates aimed at the control layers, public ownership of the most critical rails, and a broad class of owners with a real stake to defend, itself the constituency that makes the politics durable. The framing names the adversary plainly, and the adversary includes the author's own industry. A shared set of ideas that leaders gather around is not a transition plan. It is the thing that must be fought for against those who profit from the alternative.

This domain's diagnosis is concentration. Its resolution is broad ownership by design, together with chokepoints that are governed rather than captured. But the chokepoints that concentrate rents inside an economy are the same ones weaponized across borders. The settlement rail, the reserve asset, the identity root, and the override are each, at once, a place where value pools and a switch a state can throw. That is why neither remedy can be completed within a single jurisdiction, and why global coordination is the decisive arena where it is finally settled whether these chokepoints discipline or extract.