Commerce Sec. Lutnick and OpenAI CEO Sam Altman participate in a fireside chat — 9/2/2026
Speakers
Sam Altman
— CEO, OpenAI
Summary
Sam Altman discussed OpenAI's history and upcoming model during a G20 Innovation Ministerial fireside chat. He framed AI as an electricity-like general-purpose commodity that countries must adopt, predicting a boom in entrepreneurship and small business creation. He projected massive growth in AI usage and token demand, requiring much more compute infrastructure and better chips. He also flagged cybersecurity and biosecurity as key risks that could slow AI adoption.
- Altman says AI is non-negotiable and comparable to electricity.
- He expects a coming OpenAI model to be another step forward.
- AI should accelerate entrepreneurship and small business creation.
- Token usage is projected to rise from near zero per capita to enormous levels.
- Even with efficiency gains, AI requires much more infrastructure, data centers, and chips.
- Cybersecurity and biosecurity are significant near-term risks.
- He stresses making AI abundant and inexpensive to avoid power concentration.
Ideas
AI adoption is non-negotiable like electricity.
Sam Altman argues that AI is a general-purpose technology comparable to electricity and says it would be approximately as bad for a country to reject AI as it was to reject electricity 100 years ago. He believes every person, business, and society will need this 'new commodity,' and countries that do not embrace AI will constrain their capacity to think, grow, and build.
Cybersecurity risks require urgent action.
Altman warns cybersecurity is a top-of-mind issue and that some things will go very wrong unless governments and companies act urgently; this creates a risk-driven setup for cybersecurity attention and spending.
AI compute demand requires huge infrastructure.
Altman projects token usage is scaling from roughly 100,000 tokens per capita per month to hundreds of billions or more, with the heaviest users reaching 100 quadrillion tokens per month. Even with major efficiency gains, the world will need to build much more AI infrastructure, data centers, and more and better chips to make AI abundant and inexpensive rather than a scarce, highly priced commodity.
