Sam Altman and the Economic Case for the Next Stage of Artificial Intelligence

At the G20 Innovation Ministerial, the OpenAI chief presented AI as a general purpose technology that could lower the cost of turning ideas into companies while raising difficult questions about access, infrastructure and safety.

A growth argument built around individual agency

Sam Altman has spent much of the artificial intelligence boom describing increasingly capable models. At the G20 Innovation Ministerial in Chapel Hill, North Carolina, his emphasis shifted toward what those systems could allow ordinary people and small teams to do. In a fireside conversation with US Commerce Secretary Howard Lutnick on September 2, 2026, the OpenAI chief argued that AI could support an unusually broad expansion of commerce by reducing the expertise, time and capital needed to turn an idea into a working product.

The central claim is simple: entrepreneurship is often limited less by imagination than by execution. A founder may understand a customer problem but lack the ability to write software, analyze a market, design a campaign or navigate technical operations. AI tools can already assist with many of those tasks. As models become more capable, Altman expects them to function less like isolated chat tools and more like collaborators that can plan, create and complete longer sequences of work.

That does not guarantee that every idea will become a viable business. Customers still decide what is useful, and capital, distribution and judgment still matter. The economic significance lies in the falling cost of experimentation. More people can test an idea, build an early version and learn from the result before committing large amounts of money. If that pattern spreads, the number of people able to participate in entrepreneurship could grow even if most experiments never become large companies.

The policy setting around the speech

Altman's remarks came during a two-day ministerial attended by representatives of G20 members and major technology companies. The official consensus statement organized cooperation around six areas, including pro-innovation policy, workforce development, intellectual property, technical standards and investment in supply chains. That agenda shows why the debate has moved beyond the capabilities of individual models. Governments are now considering the institutions, skills and physical systems required for AI to affect productivity across an economy.

For OpenAI, this policy environment matters directly. Advanced models require large data centers, reliable electricity, specialized chips and extensive networks. A company may offer powerful software, but adoption depends on whether businesses can access compute, connect models to their operations and trust the results. The next stage of AI therefore involves infrastructure and governance as much as model performance.

Altman's growth case also carries a distribution question. When a new technology becomes essential, uneven access can widen existing differences between large corporations and smaller firms, wealthy regions and poorer ones, or highly trained workers and people with fewer opportunities to reskill. The G20 setting gave his argument a global audience, but it also raised the standard by which the argument will be judged. A technology cannot deliver broad-based entrepreneurship if its most useful capabilities remain expensive, unreliable or concentrated in a small number of markets.

OpenAI's next phase

OpenAI has described frontier AI as moving from research into daily use at global scale. That transition changes the company's job. It must continue developing models while also turning them into dependable products for businesses, public institutions and individuals. The company has recently discussed automated research systems, cybersecurity services and broader government access, illustrating how its ambitions now extend into areas with direct consequences for critical infrastructure and public policy.

The expansion makes safety harder to treat as a separate research topic. A model used for brainstorming creates one kind of risk; a system allowed to write code, operate tools or make decisions over longer periods creates another. OpenAI has argued for testing, monitoring, incident reporting and international coordination. These measures will be judged against the pace at which the company releases new capabilities and the commercial pressure to deploy them widely.

Altman's optimistic message is therefore paired with a demanding practical test. If AI is to help millions of people create businesses, the systems must be affordable enough to use, predictable enough to trust and secure enough to connect to real operations. Governments will also need policies that protect competition rather than locking in today's largest technology companies.

The measure of the promise

The most important part of Altman's G20 argument was not a forecast about a particular model. It was the suggestion that AI could widen the group of people capable of building useful things. That possibility is plausible because the technology can compress tasks that once required several specialists into a workflow managed by one person or a small team.

Whether that becomes a durable wave of entrepreneurship will depend on evidence outside the technology sector: new company formation, productivity in smaller firms, wage growth and access across different regions. The next stage of OpenAI will be measured not only by stronger systems, but by who can use them effectively and who captures the economic value they create.

Sources

• White House statement on the G20 Innovation Ministerial, September 2, 2026
• G20 fireside conversation with Sam Altman and Howard Lutnick
• OpenAI plan for broad public benefit, June 8, 2026
• OpenAI on frontier AI governance, September 9, 2026