In March 2021, Sam Altman wrote something that deserves more attention than another prediction about when machines will become smarter than us. In Moore's Law for Everything, he argued that artificial intelligence would shift economic power towards capital, and proposed distributing ownership more widely.

“Even more power will shift from labor to capital.”

That sentence gets to the question I find most interesting about AI. If more people can access extraordinary intelligence, who gets to accumulate the wealth it creates?

Most families prepare children for financial security through education. Study, develop expertise, get a good job, save from your salary. Ownership usually comes later, after the qualifications, the rent, and the first few years of earning.

AI could change the economics of that sequence. Useful expertise may become easier to access before a generation has had time to accumulate assets. People could become more capable while remaining financially exposed.

We should take that possibility seriously. We should also examine how AI could expand earnings and make starting a business easier. The outcome depends on how the gains reach households.

Stanford's 2026 AI Index offers a striking comparison: generative AI has spread faster in its early years than either the personal computer or the internet. Tools once confined to specialist teams are becoming part of everyday work, study and experimentation.

Source: Stanford HAI, AI Index 2026, Figure 4.3.9. Adoption measures use, not the quality of work produced. Source.

Adoption measures use, rather than dependable expertise. It does not tell us whether a solicitor's advice, a financial plan or a production software system became equally accessible. Those require context, verification and accountability, whatever technology supports the work.

Stanford HAI, 16 September 2026: access to a model is different from understanding, trusting or changing it. Source.

It does establish how quickly access can change across society.

The workplace evidence is more tangible. In The Cybernetic Teammate, researchers studied 776 professionals at Procter & Gamble undertaking product innovation challenges. Individuals using AI matched the performance of teams working without it in that setting.

For a founder, that raises an immediate possibility: a smaller organisation could access capabilities that previously required more people and money.

Builder perspective: Andrej Karpathy, 24 January 2023. A memorable description of the changing interface to software. Source.

The qualification matters. METR's early 2025 experiment found experienced developers took longer with the tested AI tools on familiar repositories. Its 2026 follow-up explains the difficulty of measuring newer tools. Productivity depends on the task, the user and the workflow.

METR, February 2026: the initial slowdown result alongside follow-up estimates. Negative values mean less time spent. METR cautions that selection effects make the later estimates unreliable measures of current productivity gains. Original chart, CC BY. Source.

Cheap access creates an opportunity. Turning it into dependable output still takes skill.

Imagine a financial analyst who can now complete in an afternoon work that previously took two days. Several outcomes are possible:

These outcomes can coexist. The distribution depends on competition, bargaining power, customer demand, and who owns the business.

Watch: MIT economist David Autor on how AI could help rebuild the middle class (25 February 2025). A counterpoint to the automation-only story.

Anthropic's Economic Index adds another complication. Its March 2026 report found that more experienced Claude users attempted more complex work and had higher measured success rates. Those are associations in one provider's usage data, with selection effects, rather than proof that everyone benefits equally from practice.

The lesson is that giving two people the same subscription does not give them the same opportunity.

Research update: Anthropic, 26 June 2026. This later Economic Index announcement complements the March report discussed above. Source.

Earning income and accumulating wealth also differ. A worker can become more productive without receiving a proportional pay rise. A shareholder can benefit from higher profits without personally becoming better at anything.

That is the distribution question behind the productivity headlines.

An AI assistant also cannot negotiate a better economic settlement simply by making its user faster. If everyone in a market adopts the same tools, yesterday's advantage can become tomorrow's minimum requirement. The employee, freelancer, and business owner each face a different version of that adjustment. Measuring hours saved is a useful beginning; following where the savings go is the harder, more revealing exercise.

A 2025 IMF working paper, by Emma Rockall, Marina Mendes Tavares and Carlo Pizzinelli, examines a counterintuitive possibility: AI could reduce wage inequality while increasing wealth inequality.

In their model, higher-paid workers can face displacement while also benefiting from complementary skills and capital holdings. The same household can lose ground through one channel and gain through another.

This model shows possible outcomes, not an inevitable future. Its usefulness is in separating wages from ownership.

/

Original Figure 4, IMF Working Paper 25/68, printed p. 13 (PDF p. 16). UK household data by income percentile; this population and measure differ from the US Federal Reserve comparison below. Title, axes and notes retained. Source.

The Federal Reserve's Distributional Financial Accounts show how uneven that starting position is. In the published Q2 2026 comparison, the top tenth of households by wealth held approximately $56.9 trillion in corporate equities and mutual fund shares. The bottom half held approximately $0.37 trillion.

Pension entitlements are separate categories in that comparison, so it is not a complete picture of households' investment exposure. It clearly illustrates concentrated ownership.

Suppose AI improves a company's margins. The resulting benefit might reach a family through its pension, investment fund, employee shares or business ownership. A family with little exposure to those assets has fewer routes to that particular gain.

Neither household needs to lose its job for their financial positions to diverge.

This is why I find the familiar argument about whether AI creates or destroys more jobs incomplete. Employment totals matter enormously, but they cannot tell us whether a household is building security. A person can remain employed, work effectively with AI, and still struggle to acquire assets. Another can benefit through an existing portfolio while doing little to change how they work. Both experiences belong in the same conversation.

Federal Reserve, Figure 24: share of US adults who would cover a $400 emergency expense entirely using cash or its equivalent, 2013–2025. In 2025, the share was 63%. Cash equivalents include savings and a credit card paid off at the next statement. Source.

There is a physical economy behind every apparently weightless AI interaction: semiconductors, servers, buildings, cooling systems and electricity connections.

The International Energy Agency's 2026 update projects global data-centre electricity consumption rising from 485 terawatt-hours in 2025 to around 950 in 2030. That includes data centres generally; it is not all AI consumption.

Original Figure 1.4, IEA, Key Questions on Energy and AI (2026), p. 24. All data centres, not AI alone. Scenarios are projections. IEA, CC BY 4.0; extracted without altering the figure. Source.

Falling costs per task can coexist with rising total infrastructure demand as usage expands. The companies supplying that capacity need substantial capital, and somebody owns the claims on their future earnings.

This adds a geopolitical dimension. A country can become an enthusiastic user of AI while much of the associated investment income accrues elsewhere. Hosting infrastructure, using software, and owning productive assets are different forms of participation.

There is no automatic investment windfall here. Competition, overbuilding, energy constraints, and the price investors pay can erode returns. Technological importance does not make an asset attractively priced.

For families, the relevant question is how to participate in productive growth without making their future depend on correctly picking the winning AI company.

The strongest optimistic argument deserves space. If software development, research and administration become more accessible, more people may be able to build viable businesses.

Source: Anthropic, Cadences, 26 June 2026, Figure 2.1. Outputs observed in Claude conversations; not a survey of the whole AI market. Source.

Andrej Karpathy's Software Is Changing (Again) captures the technical shift: natural language becomes another way to instruct software. That expands who can attempt to build, even though reliable products still require engineering and judgement.

Watch: Andrej Karpathy, Software Is Changing (Again), Y Combinator AI Startup School, 2025. Natural-language programming and its practical limits.

MIT economist David Autor makes a related argument about employment. AI could extend expertise to workers who previously lacked access to it, enabling them to undertake more valuable work.

That possibility is essential. A future with broader ownership can include more founders, more capable small businesses and better-paid workers.

Wharton professor Ethan Mollick, 11 November 2025, on small teams using AI. An observation about organisational design, rather than a controlled productivity result. Source.

But the ability to experiment has a balance sheet. Someone with savings can spend months testing an idea. Someone already struggling with rent may have the same tools and none of the room to fail.

Financial resilience therefore matters twice: it helps people weather disruption and gives them freedom to pursue opportunities. Lowering the price of software addresses only part of the cost of becoming an owner.

The founder still needs a customer willing to pay. They need to understand a problem, earn trust, and take responsibility when the product fails. Those constraints are particularly visible in finance, where a convincing demonstration is a long way from a service that people can depend on. AI may reduce some costs dramatically while leaving these obligations firmly in human hands.

Source: Stanford HAI, AI Index 2026, Figure 4.4.29. Employment trends differ by age; this chart alone does not establish that AI caused the changes. Source.

This is where the question becomes personal for me as a founder building NestiFi. We are trying to help families participate in a child's financial future. AI gives that work a wider context.

Education remains fundamental. Judgement, curiosity, relationships and the ability to learn will matter in a changing economy. Children also deserve to understand how ownership works long before they receive their first pension statement.

Consider two hypothetical young adults with comparable qualifications. One has a modest financial cushion and experience managing investments. The other starts with no assets and immediate pressure to earn. The difference affects which opportunities they can afford to explore.

Family contributions can help build that foundation. They can also reproduce inequality when some families have much more to give. A serious ownership agenda has to consider children whose relatives cannot contribute, including public seed funding and appropriately designed matching programmes.

World Bank, 19 March 2026, using Global Findex 2025 data. Mongolia illustrates widespread account use alongside a long-running Child Money Programme. These comparisons do not establish that the programme caused the outcomes. Source.

The question is whether financial institutions make participation straightforward for households with ordinary incomes. Products should help families build consistent saving habits while explaining costs, access and risk from the beginning.

AI's benefits can reach a household through several routes. An ownership argument should recognise all of them.

Lower prices improve purchasing power. Better work can increase earnings. Entrepreneurship creates opportunities. Investments and pensions can provide claims on business growth. Public policy affects the distribution across these routes.

For fintech builders, three practical priorities follow:

AI can help explain financial choices and reduce administrative work. Blockchain infrastructure may improve how some assets are recorded or transferred. Each needs to demonstrate a practical benefit.

A token is a technical representation. The rights attached to it determine what someone owns. Fractional access is useful only when the underlying investment, costs and protections make sense for the person buying it.

Original BIS diagram, Annual Economic Report 2025, Graph 4: a domestic payment using tokenised commercial bank deposits. A conceptual illustration of the claims and settlement process. Source.

That distinction connects with my earlier writing on the wealth transfer and Ireland's investment accounts. Better financial infrastructure should make meaningful participation easier.

For a credit union or community bank, that means looking beyond whether a family opened an account. Can relatives contribute without confusion? Can the family see charges clearly? Does the child gradually learn what the account holds? Can someone explain what happens when circumstances change? These are ordinary product questions, but their answers determine whether access becomes a lasting daily habit. They make the ownership argument concrete enough to build and measure.

Watch: BIS Project Agorá (27 May 2026; 3:28). A wholesale cross-border payments prototype illustrating better financial infrastructure, not a retail investment product.

I am optimistic about what people will build with increasingly capable AI. More accessible expertise could improve businesses, public services and everyday decisions. It could give many more people the confidence to attempt something ambitious.

But financial security will still depend on the institutions and arrangements around that technology. Wages, competition, taxation, education and social protection all matter. An investment account cannot substitute for a functioning economy or an adequate income.

What finance can do is widen the routes through which people participate. That includes helping families accumulate assets, understand them, and maintain ownership over time.

We are preparing a generation to work with machines that may create extraordinary wealth. We should also be preparing them to own a share of what those machines help build.

Created with AI assistance for research and drafting, under my editorial direction. I take responsibility for the final content.

If you found this analysis valuable, subscribe to Synaptic Finance for deep dives at the intersection of AI, blockchain, and the future of finance.