What three people can afford to begin

The three founders can afford to investigate one field. Their first contracts secure access to the land and equipment for the wells. Money leaves the company long before anyone knows how much electricity the rock will yield. A nearby industrial district wants a dependable source of power, but it will pay for energy delivered, not heat somewhere underground.

In the future in which they begin, advanced drilling reaches temperatures that earlier equipment could not tolerate. Robotic construction and capable scientific models are widely available. The founders can establish an AI organisation with the expertise to investigate the field, design its plant and develop missing equipment. They still have to finance the wells. Their advantage is that the scientific and engineering capacity around those wells no longer requires building a large company before the work can start.

The first question is whether the field can deliver enough heat, for long enough, to justify everything above it. Water must pass through useful parts of the rock and return through the production wells. A route that carries plenty of water may exchange too little heat. Conditions change as the reservoir cools and the rock responds to its operation.

Before the first electricity sale

The company's researchers treat the reservoir and the plant as one undertaking. Geologists investigate where fluid can travel. Equipment designers need to know which temperatures and chemical conditions their components will face. The plant's designers need a forecast of what will arrive at the surface. Each team works with uncertainties that another team may be able to reduce, at a cost.

A measurement deep in a well may cost more than a long programme of computation. Before committing to it, the AI teams compare competing accounts of the reservoir, find where their predictions diverge and design an intervention that makes the next field measurement informative. The company has to learn which uncertainty is worth spending its limited capital to reduce.

Months into the investigation, an unresolved geological possibility still affects which plant the company is prepared to build. New models join with the field's developed understanding and the methods used to challenge it; they can apply stronger reasoning to a live scientific problem. The founders' capital pays for that further inquiry without repeatedly paying to reconstruct the work that preceded it.

This changes what their limited capital can support. They can maintain a serious investigation into an uncertain field while preparing equipment manufacture and a commercial proposal. The work remains risky, but its intellectual capacity is no longer confined to whichever specialists a small company can keep occupied and employed throughout every delay.

The reservoir inside the forecast

An early circulation test produces an encouraging flow, followed by a temperature decline the preferred reservoir model did not predict. One explanation is that the water is finding a short route between wells. Another is that it reaches a broad region of rock but exchanges heat less effectively than expected. The distinction determines whether the company should change its operating pattern, develop another connection into the reservoir or abandon that part of the field.

The researchers combine a tracing experiment with changes in injection. They compare the water's travel time with its temperature and the response to different operating conditions. The resulting evidence cannot be explained by the company's original picture of one connected reservoir. Some paths dominate early flow; other regions contribute slowly and respond differently when the operating pattern changes.

The company needs a usable prediction of how these familiar processes interact in this particular, changing volume of rock. It builds models that represent several plausible flow networks and identify operations that distinguish them. Its control system learns to choose among those operations while accounting for the heat needed by the plant.

That ability changes the equipment the company asks for. Its engineers design tools and well components around the measurements and interventions the research now requires. Field experience changes the instruments; better instruments change the quality of the reservoir model. The company acquires a way to develop a field through controlled learning, with the plant and the investigation advancing together.

An expensive lesson travels

At the next site, the inherited organisation moves quickly. It recognises familiar rock characteristics and finds an operating pattern resembling the successful one at the first field. The plant forecast looks sound. But the apparent resemblance hides a different arrangement of connected fractures. The team's most experienced model gives too much weight to the pattern it already knows how to exploit.

The mistake becomes visible when a deliberately altered circulation pattern produces a response outside the forecast. Following the old plan would send further investment towards the wrong part of the reservoir. The company delays the expansion and repeats an investigation its experienced team had considered largely settled.

The correction changes what later sites inherit. The company separates broadly useful investigative methods from conclusions tied to a particular field. Its successor models learn to recognise which measurements would justify transferring a successful operating policy, and which similarities are insufficient. They keep the ability to conduct the investigation while revising how confidently that ability should be applied.

The founders have to carry the delay. They give up an opportunity to reserve equipment for another site and fund the additional work. The distinction between transferable investigative methods and conclusions tied to one field will now influence where they commit the company's next capital.

Why own the factory

As the first plants settle into operation, the company gains observations that were unavailable when it began. Components encounter repeated heating and cooling, abrasive flows and local water chemistry. Its engineers follow how parts perform after manufacture and repair. They can connect a field problem to choices made much earlier in a design, then change the process that will produce the next part.

The organisation develops specialised drilling and well equipment. Outside manufacturers can make the first designs, and licensing them would give the founders an attractive source of income. Instead, the founders commit capital to manufacturing capability of their own. They want their researchers to be able to change a process, build the necessary tooling and follow its consequences into an operating well.

Owning a factory introduces more problems than owning a drawing. The company must learn how to make parts consistently, measure whether they meet their purpose and maintain the machinery that produces them. Its AI organisation acquires that competence through production. Manufacturing specialists become part of the research system, able to tell a designer which proposed improvement would require an entirely different process and which experiment could establish whether the process is worth building.

By the time later plants are commissioned, their engineering teams begin with an experienced capacity for reservoir development and equipment manufacture. Stronger models join with that practical knowledge already usable. The company can undertake work that its original organisation could not have specified: designing the tools, factories and scientific methods needed to reach a new class of resource.

The cost of the next company

The fleet expands across continents. Each region still requires investigation, construction and agreements with the people who live and work there. What travels is an organisation that knows how to acquire the missing knowledge and turn it into a working industrial system. Local experience changes that organisation in return. The intellectual starting point of each expansion rises while the amount of human coordination required from the founders grows much more slowly.

Some gains reinforce the research that produced them. More dependable energy supports the company's factories and experimental facilities. Those facilities make difficult equipment programmes affordable, and the resulting machinery opens further resources. The company still accounts for materials, capital and the cost of inference.

Its plants supply industries whose location and scale depend on reliable power. A district can consider desalination or energy-intensive manufacture with a different set of expectations. The company's equipment business equips other operators. Experienced AI organisations make it possible for new founding teams to develop projects of their own, beginning with scientific and engineering capacities that once had to be assembled over decades.

The three founders' role changes with the company. Their early decision concerned one uncertain field. Later, they decide which industrial programmes deserve the capital generated by a mature fleet. Their organisation can investigate those choices across disciplines, invent what is missing and execute a selected programme. A small human team has acquired the means to originate work at the scale of an industry.

A different first contract

At a new site, the proposal includes a power plant and a factory to make equipment for the region. The factory would once have required a second company, with its own long search for experienced engineers and a separate accumulation of manufacturing knowledge. Now it begins as a programme within an organisation that has learned how to build both.

The founders still have to choose. The project will tie up capital, and the field may teach the company something expensive. They commission the first investigation. Before the drill reaches the rock, its research team is already designing the measurements that will determine which version of the industry to build.