OpenAI CEO Sam Altman told the BlackRock Infrastructure Summit in Washington, DC this week that artificial intelligence will be sold to consumers and businesses through a usage meter, the same way utilities deliver electricity and water. The remarks, made on the same day the Vatican released its first major AI manifesto, have ignited a global debate over who should control access to machine cognition.
Key Highlights
- Altman: "We see a future where intelligence is a utility, like electricity or water, and people buy it from us on a meter."
- Statement delivered at the BlackRock Infrastructure Summit in Washington, DC, on May 25, 2026
- OpenAI generates roughly $13 billion in annual revenue and is on pace for up to $14 billion in losses this year
- Critics warn of a "cognitive infrastructure gap" between users who can afford premium reasoning and those who cannot
- The framing arrived hours after Pope Leo's first encyclical centered on AI and human dignity
The Utility Vision
Speaking before infrastructure investors gathered by BlackRock, Altman reached for an old slogan from the nuclear era to describe where he wants AI to go next. Quoting Lewis Strauss's 1950s prediction that nuclear energy would be "too cheap to meter," he argued that intelligence itself should follow the same trajectory — extremely abundant, continuously available, and billed by the unit.
The unit, in OpenAI's case, is the token: the chunk of text models consume on input and produce on output. Every ChatGPT message, every API call, every agent action is already counted in tokens behind the scenes. Altman's pitch is that this metering should be made visible and pervasive, the way kilowatt-hours are on an electricity bill.
Why Altman Said It Now
The metered-utility framing is not a casual aside. OpenAI is in the middle of a historic infrastructure spend, with capital commitments to Microsoft, Oracle, Nvidia, and the BlackRock-led Stargate consortium that already stretch into the hundreds of billions of dollars. Pitching AI as a basic utility helps justify that buildout to the institutional capital sitting in the room.
The financial backdrop matters. OpenAI's annual revenue is around $13 billion, dominated by ChatGPT subscriptions and API fees. The company is projected to lose as much as $14 billion this year as it scales compute, data centers, and frontier model training. Framing the company as a future utility — boring, regulated, indispensable — softens the loss story.
The Vatican Counterpunch
Altman's remarks landed the same day Pope Leo issued his first major AI encyclical, declaring that "in the era of artificial intelligence, when human dignity is threatened by new forms of dehumanization, ours is the pressing duty to remain profoundly human." The two messages now sit in direct tension: one frames AI as cheap, ubiquitous infrastructure, the other warns against outsourcing what it means to be human.
Observers across X were quick to draw the contrast. "AI is not electricity. Electricity powers the machine. AI will decide what the machine sees, remembers, filters, writes, recommends, negotiates and hides," wrote one widely shared thread responding to both statements.
Critics: Cognition Is Not a Commodity
The strongest pushback to Altman's metaphor centers on one point: a kilowatt-hour does not think for you, but a token does. Treating cognition as something you rent from a private company raises questions utilities have never had to answer.
Three concerns dominated the response:
- Concentration risk. If a handful of labs run the meters, they also decide what reasoning is available, what is rate-limited, and what is silently blocked.
- Cognitive infrastructure gap. Premium AI tiers could create a divide between users who can afford the best reasoning and those who cannot, mirroring and amplifying existing inequalities.
- Public-utility status. Real utilities are regulated, audited, and price-controlled. None of that exists for frontier AI today.
What It Means for MENA and Tunisia
For markets like Tunisia and the wider MENA region, the metered-utility framing has practical consequences. If frontier reasoning ends up priced per token and billed in US dollars, local startups, public institutions, and SMEs will face the same dependency curve that already exists for cloud compute — paying foreign meters for a domestic capability.
That is precisely why open-weights models from Mistral, Qwen, DeepSeek, and the Hugging Face ecosystem matter more than ever, alongside regional efforts like Saudi Arabia's HUMAIN. The alternative to a single metered intelligence provider is plural, sovereign, and partly open infrastructure — including some compute and inference run domestically.
What's Next
Altman's framing will now be tested on two fronts. Politically, US and EU regulators will be asked whether a "metered intelligence" provider should be treated like a public utility, with the obligations that brings. Commercially, every OpenAI rival — Anthropic, Google DeepMind, xAI, Meta, the Chinese labs, and the open-source community — will use the moment to argue that the meter should never be controlled by one company.
For users, the most immediate signal is simpler: the per-token line item on your invoice is no longer a billing detail. It is the business model.
Source: Tom's Guide