During an appearance at BlackRock’s U.S. Infrastructure Summit in Washington, D.C., OpenAI CEO Sam Altman articulated a vision where artificial intelligence matures into a metered utility, fundamentally changing how the world consumes “intelligence.”
Altman argued that the industry is shifting away from selling software products and toward selling compute-on-demand, comparing the future of AI to essential services like electricity and water.
The “Intelligence as a Utility” Model
Altman’s comments suggest a shift in the economic structure of AI from fixed subscriptions to a usage-based commodity.
- Buying by the “Metre”: Altman explicitly stated, “We see a future where intelligence is a utility, like electricity or water, and people buy it from us on a metre.” * The Token Economy: He explained that the business of AI labs will increasingly look like “selling tokens”—the fundamental units of data processing. Just as you don’t pay for the concept of electricity but for the kilowatts you use, you will pay for the exact amount of “reasoning” your task requires.
- On-Demand Cognition: This model is designed to manage finite computational resources. Instead of “always-on” flat rates, users will scale their consumption based on the complexity of the task (e.g., a simple email vs. a complex physics simulation).
“Too Cheap to Meter”
Evoking a famous (and historically controversial) 1954 phrase from the nuclear energy industry, Altman stated that OpenAI’s goal is to make intelligence “too cheap to meter.”
| Goal | Strategic Implication |
| Abundance | Moving from “scarcity of thought” to a world where high-level reasoning costs fractions of a cent. |
| Price Convergence | Altman predicts the cost of AI intelligence will eventually converge toward the marginal cost of the electricity used to generate it. |
| Scaling Laws | To achieve this, OpenAI is pursuing a $1 trillion infrastructure roadmap, including massive data center buildouts to ensure supply outstrips demand. |
Context: The Infrastructure Crisis
The “utility” comments come at a time when OpenAI is under pressure regarding its massive energy and financial requirements:
- Government Backstop: In early 2026, OpenAI leadership reportedly suggested that the U.S. government act as an “insurer of last resort” for AI infrastructure, further leaning into the idea of AI as a public utility.
- Environmental Pushback: Altman has spent early 2026 defending the energy footprint of AI, recently sparking debate by claiming that human energy consumption (20 years of food and life to “train” a smart human) is a fair comparison to the energy used to train a model.
- Financing Hurdles: Reports indicate OpenAI recently scaled back parts of its “Stargate” data center project in Texas due to financing complexities, making the shift to a profitable “metered” revenue model more urgent.
Industry & Public Reaction
The “metered intelligence” vision has faced immediate backlash from two fronts:
- The “Pay to Think” Critique: Social media users and tech ethicists have criticized the idea as “Black Mirror-esque,” arguing it commodifies a fundamental human trait and could lead to a “cognitive divide” between those who can afford high-quality AI reasoning and those who cannot.
- Competitor Dissent: Leaders from Anthropic and Google have reportedly signaled that they do not necessarily agree with the utility-only model, preferring to keep AI integrated into broader, value-added software platforms rather than selling it as a raw commodity.
