SYSTEMS / POWER
When the Bet Fails, You Pay.
Private genius.
Public risk.
The billionaire stands at the top. Beneath him are the scientists, workers, universities, infrastructure and taxpayers who made the fortune possible. If artificial intelligence succeeds, who owns it? If it fails, who receives the bill?
The billionaire is real. The self-made billionaire is the fiction.
We are encouraged to imagine technological progress through a small collection of exceptional men.
Steve Jobs gave us the smartphone. Bill Gates brought computers into our homes. Jeff Bezos reinvented retail. Elon Musk electrified the car and conquered space. A new group of artificial intelligence executives is now presented as the author of our collective future.
The story is attractive because it is simple. One brilliant individual sees what everybody else has missed, risks everything and creates an industry through force of intellect.
But look underneath the individual and an entire supporting structure appears.
Scientists conducted publicly funded research. Universities trained the engineers. Governments constructed roads, electricity networks, communications systems and laboratories. Workers built the factories, coded the products and maintained the machinery. Public agencies supplied loans, contracts, tax incentives and early customers.
The billionaire did not create that structure.
He acquired a particularly powerful position within it.
Billionaires do not stand above society because they built everything beneath them. They stand above society because everything beneath them has been edited out of the picture.
That distinction matters as governments commit public money, land, energy and infrastructure to the artificial intelligence boom.
The question is not whether the state should support innovation. Modern economies could not function without public investment.
The question is what the public receives in return.
What Holds Up the Billionaire?
Steve Jobs Did Not Invent the iPhone Alone
Steve Jobs is a useful place to begin because Apple’s achievement was real.
The company combined technology, software, industrial design and consumer psychology exceptionally well. It created a product people could understand and an ecosystem they found difficult to leave.
Jobs possessed an unusual ability to recognise which technologies could become a desirable commercial product.
But that is not the same as inventing every technology inside it.
The internet emerged from decades of government and university research. GPS was developed through the American military. Public research contributed to touchscreens, lithium ion batteries, semiconductors, networking and cybersecurity.
Siri developed from technology associated with a programme funded by the US Defense Advanced Research Projects Agency.
The state funded uncertain research over long periods when no finished consumer product existed and no guaranteed commercial return could be promised.
Apple then performed the commercially valuable work of integration, design, manufacture and distribution.
Both contributions matter.
The distortion begins when collective invention disappears and the person controlling the final company becomes the sole author of the product.
Consumers could purchase an iPhone. They did not receive shares in the company whose value was built partly upon technologies their taxes had helped develop.
This is the first version of the model. The public finances knowledge. Private enterprise packages it. Private ownership captures the commercial return.
Musk Did Not Conquer Space Alone
Elon Musk offers a more direct example of the relationship between private enterprise and the state.
In 2010, the US Department of Energy issued Tesla a $465 million loan to develop and manufacture electric vehicles and associated components.
Tesla repaid that loan early, with interest.
That fact should remain in the story. Public investment is not automatically a loss, and government support is not inherently corporate welfare. The state accepted early risk, the company succeeded and the loan was returned.
SpaceX is also not simply a recipient of free government money. NASA purchases development work, launches, cargo deliveries and crew transportation. SpaceX provides valuable services in return.
But the company’s rise cannot be separated from NASA’s scientific inheritance, facilities, expertise, contracts and willingness to act as an early customer.
NASA’s Commercial Crew contract with SpaceX eventually reached approximately $4.93 billion.
The private company innovated within a market whose most important customer, technical partner and institutional foundation was the American state.
The state did not merely rescue the market after failure. It helped create the market in which private success became possible.
Musk deserves credit for decisions made, risks accepted and companies developed.
He does not deserve the erasure of the thousands of people and public institutions without which those achievements could not have occurred.
Investment, ownership and publicity have concentrated the story around one individual. The work remained collective.
What Does the Public Receive?
| Public role | What is provided | Possible public return |
|---|---|---|
| Research funder | Science, knowledge and early technology | Public knowledge, but often no ownership |
| Infrastructure provider | Power, water, roads, land and connections | Economic activity and tax receipts |
| Subsidy provider | Lower costs and reduced commercial risk | Investment and promised employment |
| Customer | Government contracts and early demand | A product or public service |
| Lender | Early capital | Repayment and interest |
| Equity investor | Capital and strategic support | Ownership and potential profit |
| Rescuer | Emergency finance after failure | Frequently the remaining liabilities |
Artificial Intelligence Raises the Stakes
Artificial intelligence is often described as if it exists somewhere beyond the physical economy.
It does not.
It requires semiconductor factories, data centres, transmission networks, cooling systems, water, land, construction materials and enormous quantities of electricity.
It also depends upon decades of publicly supported mathematics, computer science, university research and defence programmes.
The companies own the models, platforms, customer relationships, intellectual property and potential profits.
Governments are increasingly being asked to provide the conditions that make those assets commercially viable.
Across the world, states are subsidising semiconductor manufacturing, accelerating planning, financing research, expanding electricity networks and providing public-sector contracts.
The largest technology companies are now committing extraordinary sums. The five largest hyperscalers were expected to spend more than $1 trillion on artificial intelligence infrastructure across 2025 and 2026.
Investment at that scale is not supported entirely by spare corporate cash.
AI-related borrowing has expanded rapidly. Data-centre financing increasingly uses bonds, private credit and complex project structures. Some obligations are held outside the technology companies’ main balance sheets.
The private bet is becoming connected to the wider financial system.
The Ownership Does Not Move with the Risk
Britain Has Entered the Venture Game
In April 2026, the British government launched Sovereign AI, a £500 million state-backed venture fund.
Its stated ambition is to help British artificial intelligence companies begin, expand and remain in Britain.
The fund can invest directly in companies from the earliest stages through to larger growth rounds. Initial equity investments can reach £10 million.
But the offer extends far beyond money.
Supported companies can receive fully funded access to Britain’s largest artificial intelligence supercomputers, with as many as one million GPU hours available to an individual startup.
They can receive expedited visa decisions, reimbursement of certain visa costs, access to national datasets, research support, product validation and assistance navigating regulation.
The government is also making as much as £100 million available through a procurement scheme intended to turn the state into an early customer for British AI companies.
Suppliers can retain ownership of intellectual property developed through those contracts and are encouraged to commercialise it.
Sovereign AI’s own website tells prospective founders:
The United Kingdom is your unfair advantage.
It explains that an ordinary venture-capital fund can write a cheque, while Sovereign AI can provide the unique powers of the British state.
That is an unusually honest description of the system.
The state offers capital, computing power, public data, visas, procurement and political coordination. The company receives an advantage that no ordinary investor could assemble.
This could be intelligent industrial policy.
It could also become another system in which the public finances the fragile beginning and foreign capital purchases the valuable maturity.
The £80 Million Question
Britain’s AI Growth Zones offer another form of public support.
Government policy proposes faster planning, accelerated grid connections, specialist infrastructure and discounted electricity for selected data centres.
The government estimates that these interventions could reduce the wait for electricity connections by as much as five years.
£80 million Possible annual electricity saving for one 500 MW data centre
The objective is to unlock as much as £100 billion in private investment and create more than 10,000 jobs.
But the phrase “private investment” becomes less straightforward when the investment depends upon publicly organised energy discounts, planning privileges and infrastructure.
The building may be privately owned. Its commercial viability may have been collectively constructed.
That does not automatically make the policy wrong. Governments support strategically important industries because the wider benefits can exceed the immediate financial return.
But it makes scrutiny essential.
If government action saves a private data centre £80 million every year, what proportion of its future value belongs to the public that made those savings possible?
Infrastructure Is Not the Same as Employment
Data centres generate considerable construction activity. They require land, concrete, steel, electrical systems, cooling equipment, security and specialist engineering.
Once built, however, they can operate with comparatively small workforces.
Industry modelling has suggested that Britain’s planned developments could create more than 40,000 jobs.
A competing analysis of 20 projects estimated approximately 10,400 permanent roles, roughly one quarter of the industry projection.
The dispute is not settled. Different studies count direct, construction, supply-chain and induced employment differently.
The distinction matters because temporary construction work can be presented as though it were permanent employment sustained throughout the facility’s life.
Britain already had approximately two gigawatts of operational data-centre capacity at the end of 2025. Another 0.87 gigawatts was under construction, with a further 10.88 gigawatts announced or permitted.
If that pipeline were completed, operational capacity would increase almost sixfold.
That is not a collection of anonymous warehouses.
It is a new industrial system competing for electricity, water, land and grid capacity.
Britain must therefore ask what it receives for those resources, not simply how much money developers promise to spend.
Britain Could Become an AI Host Rather Than an AI Owner
The government says Britain must become an artificial intelligence maker, not merely a taker.
The phrase reveals the danger it is attempting to prevent.
Britain possesses strong universities, valuable scientific research, skilled workers, public datasets, the NHS, land for data centres and a government willing to purchase technology.
The United States possesses Microsoft, Amazon, Google, Meta, Nvidia, OpenAI, enormous capital markets and the dominant global cloud platforms.
Britain can therefore provide the inputs while American companies retain the highest-value assets.
- Research
- Land
- Electricity
- Water
- Public data
- Government contracts
- Customers
- Models
- Intellectual property
- Subscriptions
- Customer relationships
- Global infrastructure
- Market power
- Profit
Brexit deepened this problem.
Britain once offered companies an English-speaking base with frictionless access to the European Single Market. It is now a medium-sized market beside a much larger commercial and regulatory bloc.
Having weakened its structural offer, Britain increasingly attempts to manufacture attractiveness through faster planning, lighter regulation, cheaper infrastructure and direct state support.
Artificial intelligence has also been given an extraordinary political task.
It is expected to raise productivity, compensate for labour shortages, transform public services, create high-value employment, attract investment and demonstrate that post-Brexit Britain still possesses an economic future.
That is a great deal of national expectation to place upon one technology.
If Britain provides the land, power, research and customers while somebody else owns the platform, it has not built an AI economy. It has built an extraction site for somebody else’s AI economy.
Hosting Is Not Owning
Who Pays if Artificial Intelligence Crashes?
There is no automatic government guarantee that failing artificial intelligence companies will be rescued.
An AI crash would not necessarily resemble the banking crisis of 2008. A failed model developer does not perform the same systemic function as a major clearing bank.
But the absence of a formal guarantee does not leave the public untouched.
A speculative boom can transfer costs without the government purchasing a failed company directly.
Public subsidies may already have been spent. Electricity networks may have been expanded around demand that does not materialise. Local authorities may have granted tax concessions for employment that never arrives.
Suppliers can fail. Workers can lose their jobs. Pension and insurance funds can hold exposed debt. Communities can inherit enormous, specialised buildings with little alternative use.
Public services may become dependent upon proprietary systems that are withdrawn, repriced or acquired.
Governments can also face pressure to support semiconductor, infrastructure or defence-related companies considered strategically too important to disappear.
The taxpayer may never see a bill labelled “AI bailout”.
The bill can arrive through higher network costs, lost tax revenue, unemployment support, retraining programmes, public procurement failures and diminished pension values.
The loss becomes public by travelling through the system.
Public Investment Is Not the Problem
The answer is not to demand that government withdraw from innovation.
Private markets often avoid technologies whose development is expensive, uncertain and slow. Governments can invest across longer periods and accept risks that conventional finance will not.
Public investment can produce knowledge, employment, national capability and industries that would otherwise never exist.
The problem is an agreement in which the public repeatedly accepts the uncertainty while surrendering the most valuable return.
Loans can be repaid with interest. Government procurement can purchase useful services. Equity investment can give the state a share of commercial success.
Public funding can also attach conditions concerning employment, domestic investment, affordable access, environmental performance, intellectual property and profit sharing.
Britain’s Sovereign AI fund takes equity in selected companies. That is a meaningful improvement over support that simply reduces private costs.
But the principle must extend across the wider structure.
When public investment helps create a successful company, does the public receive a proportionate return, or merely the opportunity to buy the finished product?
If taxpayers provide capital, infrastructure and strategic advantage, they should not be treated as spectators waiting outside the company.
They are investors, whether the accounting recognises them as such or not.
THE PEOPLE UNDERNEATH
The Table
Has No Legs.
The billionaire appears self-made only when the foundation is removed from the frame.
Restore the universities, laboratories, workers, taxpayers, contracts, power stations, roads, data and public institutions, and the solitary genius becomes something more recognisable.
He becomes one participant in a collective system, albeit the participant granted the greatest ownership, visibility and reward.
Artificial intelligence may justify substantial public investment. Britain cannot create new industries by refusing to take risks.
But taking risk without retaining power is not an industrial strategy. It is a subsidy for somebody else’s future.
Britain must decide whether it wants to own the technologies of the next economy or merely provide the land, electricity, data and customers upon which foreign platforms become richer.
It must also decide whether public investment makes the public a genuine shareholder in success or simply the final guarantor against failure.
Until those two answers involve the same people, the system is not rewarding genius. It is concentrating ownership.