The Shifting Boundaries

The marvel is that in a case like that of a scarcity of one raw material, without an order being issued, without more than perhaps a handful of people knowing the cause, tens of thousands of people whose identity could not be ascertained by months of investigation, are made to use the material or its products more sparingly.

An old asset does not become useless when a new productive regime appears. It changes office. Canals fed railways after railways displaced them on many routes. Telegraph networks survived the telephone where written transmission retained an advantage. The interesting question is not which technology dies. It is which one sets the terms of combination.

Alfred Chandler located one such change in the rise of the large managerial corporation.¹ Railroads, steel, and oil demanded coordination across distances and throughputs that older commercial forms handled poorly. Hierarchy became productive because it could schedule flows, standardize operations, and keep expensive plant in use. This was not capital obediently following a production function. Finance, management, transport, and law developed together, each altering what the others could do.

Joel Mokyr's account of useful knowledge makes the complementarity clearer.² Coal does not design a steam engine, and a drawing does not supply heat. Productive capacity joins materials and energy to the techniques by which they can be used. A change in one can raise the return to another. It can also strand knowledge and assets organized around a former combination.

Mancur Olson supplies a warning against treating institutions as friction left over from the physical story.³ Organized interests can make arrangements durable after their original circumstances have changed. Institutions respond to productive opportunities, but they also filter those opportunities through inherited coalitions, rights, and vetoes. A resource base does not write its own constitution.

These accounts do not converge on one law. Together they show why economic boundaries move unevenly. New capacity has to pass through organizations. Organizations have histories. Finance anticipates some uses and misses others; incumbents convert old advantages into new ones when they can; public authority decides which conversions are permitted, subsidized, delayed, or forbidden.

Computation enters this field with an unusual versatility. The same broad infrastructure can support code, design, prediction, research, entertainment, surveillance, and administration. That reach increases the value of software, data, and distribution, while also increasing demand for chips, power, cooling, sites, and network capacity. No single bottleneck follows. A model provider may dominate one market through distribution, a chipmaker another through scarce equipment, and a regulated intermediary a third through permission to act.

The lesson of earlier transitions is not that margin always migrates to a newly scarce physical input. Sometimes an incumbent controls the complement and survives. Sometimes a standard opens entry. Sometimes regulation creates a protected franchise, and sometimes it prevents a private chokepoint from becoming sovereign. Asset prices can anticipate the change long before physical capital turns over, or mistake a temporary shortage for a permanent rent.

This matters for claims about an impending shift from software to plant. Software cannot clear an interconnection queue, but a power contract cannot create a useful model or reach a customer. Network effects do not generate electricity, yet they may decide which inference supplier can pay most for it. The new capital structure will be composed from both light and heavy assets. Which one commands the arrangement must be discovered in the market and in law.

Finance participates twice. It funds the infrastructure through which new capability is produced, and it prices claims upon the income expected from that capability. Those claims can outrun the plant, demand, or institutional permission that must eventually sustain them. They can also accelerate construction that would otherwise arrive too late. Financial excess and productive anticipation are entangled well before anyone can separate them in retrospect.

A further boundary concerns the actor itself. Capital has historically been owned and directed by persons and institutions with durable standing. A computational runtime can now search a design space, recommend an allocation, or execute a bounded transaction without becoming the owner or principal. Its authority comes from accounts, credentials, and institutions that survive it. The practical act may be computational even while legal responsibility remains human and organizational.

That distinction prepares the rest of the book. Work can change state, from service consumed in the moment to capability preserved and repeatedly deployed. Capital can enter the loop by owning that capability and allowing it to participate in allocation. Neither change abolishes physical limits or institutional choice. It makes their intersection harder to see.

Before asking where returns collect, then, the argument needs a physical account. Computation is embodied, but the fact of embodiment does not tell us its current cost, its value, or who will control it.