AI as a Utility: What Sam Altman’s BlackRock Moment Means
AI as a utility is the future Sam Altman publicly described at BlackRock’s U.S. Infrastructure Summit in Washington on 11 March 2026, in a single sentence that captured the commercial logic of the entire AI industry.
‘We see a future where intelligence is a utility, like electricity or water, and people buy it from us on a metre,’ the OpenAI chief executive told the room of infrastructure investors, alongside OpenAI board member Adebayo Ogunlesi.
The remark drew a sharp public backlash within days but the more useful question for SME leaders is not whether Altman should have said it. The useful question is what the statement reveals about where AI is heading and what your business should do about it.
The reaction online was loud and at times angry. Critics traced what they called the three-step playbook of the AI industry. Scrape the open internet without asking, train a model on the collective output of human thought, then position the resulting system as a utility you charge people to access by the token.
One Reddit user captured the mood by writing that companies ‘stole all this data from us, the people, our life’s work, creativity, art, by devouring the internet and blowing through all copyright laws. Now they want to sell it back to us in the form of a utility’. Whether you agree with that framing or not, the commercial direction it points to is real, and SMEs that ignore the strategic implications will pay for it through vendor concentration risk that compounds every quarter.
AI as a Utility: The Quote in Its Proper Context
It helps to read what Altman actually said in full, because the short version that travelled across LinkedIn and X strips out the commercial logic underneath. The longer quote runs: ‘Fundamentally our business, and I think the business of every other model provider, is going to look like selling tokens. We see a future where intelligence is a utility like electricity or water and people buy it from us on a metre and use it for whatever they want to use it for. The demand that we see for that seems like it’s going to continue to just go like this.’
In other words, Altman was making three connected points to an audience of infrastructure investors who fund things like power stations and data centres. First, that the model providers are fundamentally token-selling businesses. Second, that demand is exponential. Third, that the constraint on supply is compute, which is what infrastructure investors can actually build. The Stargate programme came up immediately afterward as the obvious response.
The pitch landed badly with the wider public because the utility framing made the economic relationship between AI companies and their users uncomfortably explicit. When you describe intelligence as something you sell to people on a metre, the question of how you came to own that intelligence in the first place becomes harder to avoid. As Disconnect blog put it, Altman ‘laid out an intensely dystopian future as just another business development’, and that gap between the cold commercial framing and the lived implications is what drove the reaction.
AI as a Utility: What This Reveals About OpenAI’s Economic Position
The metred utility framing is not a philosophical preference, it is a business model search. According to reports surfacing in January 2026, OpenAI generates approximately $13 billion in annual revenue from ChatGPT subscriptions and API access, but is projected to lose roughly $14 billion in 2026 against costs of infrastructure expansion, model training, research hiring and compute. The same reports suggested bankruptcy risk by mid-2027 if the trajectory does not change.
The utility framing solves that problem in theory. A subscription business has a ceiling, capped by the number of people willing to pay $20 a month. A metred utility business does not. If your business is fundamentally selling tokens, and AI becomes embedded in every workflow, every decision and every transaction, revenue scales with usage rather than with subscriber count. The economics get genuinely interesting if you can land the framing.
Whether the framing lands is a different question. Utilities exist because the underlying resources (water, electricity, gas) require enormous physical infrastructure to deliver, and because society has historically chosen to regulate that infrastructure for the public good. The AI industry has the first part of the analogy (compute genuinely is enormous infrastructure) but is moving in the opposite direction on the second. Public oversight of these systems is currently minimal, ownership is concentrated in a handful of US companies and the businesses making the utility argument are also lobbying against regulation that would treat them like utilities.
AI as a Utility: Why Vendor Concentration Risk Matters for UK SMEs
The utility framing matters to SMEs not because of how it reads emotionally but because of what it tells you about the shape of the market. When the dominant provider in any category starts publicly describing itself as the future utility for that category, the strategic implication is clear. They are pitching for the kind of position that creates lock-in, switching cost and pricing power. Businesses that build their AI strategies around a single vendor will pay the consequences of that lock-in for years.
Three specific risks deserve attention from SME leaders.
The first is pricing risk
If intelligence really does become a metred utility, the company controlling the metre sets the rates. Token prices have moved sharply downward in the last two years thanks to competitive pressure from Anthropic, Google, Meta and the open-source ecosystem. That pressure is the only thing keeping costs reasonable. A future in which one or two providers dominate will not preserve those favourable economics.
The second is access risk
Altman’s own quote points to compute as the binding constraint on AI supply, and he explicitly noted that if there is not enough compute, ‘the price gets really high and it kind of goes to rich people, or society makes a bunch of sort of central planning decisions that I think almost always go badly’. The implication for SMEs is uncomfortable. In a compute-constrained world, the businesses willing to pay the most for AI capability get it and the businesses operating on tighter margins do not.
The third is dependency risk
A business that builds critical workflows on a single AI provider, with prompts, contexts, integrations and operational knowledge all tied to that provider’s specific interface, is a business that has handed over significant strategic optionality. The cost of migration to an alternative provider rises every quarter as the integration deepens. Treating AI as a utility means treating your AI provider with the same caution you would apply to any other utility supplier, knowing that switching is rarely as easy as the marketing suggests.
AI as a Utility: The Practical Response for SMEs
The good news is that the response to vendor concentration risk does not require any moral or philosophical position on the utility framing. It just requires structured strategy.
Build multi-vendor capability from the start
Our recent comparison coverage has gone deep on this point. As we explored in our Claude vs ChatGPT for business blog, the most effective SMEs are increasingly running both Claude and ChatGPT with clear workflow rules about which handles which task. As we covered in our Microsoft Copilot vs ChatGPT for business blog, 34% of enterprise AI deployments now include multiple AI platforms by deliberate strategy. As we explored in our Claude vs Gemini for business blog, the right tool decision depends on workflow rather than brand, and multi-vendor setups are increasingly the default rather than the exception.
Design for portability
Workflows that depend on a specific provider’s custom features (custom GPTs, vendor-specific prompt engineering, proprietary connectors) become expensive to migrate. Workflows built around portable patterns (Model Context Protocol, model-agnostic prompts, modular integrations) preserve your optionality as the market shifts. The principle is the same as never building your business critical software on a single proprietary platform without an exit plan.
Pay attention to terms of service and data handling
As we covered in our AI compliance blog, the regulatory and contractual environment around AI is tightening. Where your data goes, how it is used, whether it informs model training and what happens to it when you cancel the subscription are all decisions that affect your business risk profile. The ‘utility’ framing makes some of these questions feel routine but the legal and operational specifics still matter.
Treat compute as a strategic resource, not a procurement line item. This is more relevant for SMEs building AI-powered products than for those just using AI internally, but it matters increasingly for both. As we explored in our AI as Infrastructure blog, the underlying compute layer of AI is the binding constraint on the entire stack, and businesses dependent on that compute should think about it the way they think about energy supply, with hedging, diversification and forward-looking strategy rather than passive consumption.
AI as a Utility and Your AI Confidence Journey
The structured response to vendor concentration risk maps onto the AI Confidence Journey, the five-stage path every SME travels from initial AI uncertainty to genuine operational confidence. Confused businesses cannot make sensible vendor decisions yet, which is why our free AI Readiness Assessment is the right starting point for understanding where your business stands and which workflows matter most.
Curious businesses use an AI Workshop to test multiple vendors against real work scenarios, surfacing the practical evidence for a multi-vendor strategy rather than committing prematurely to one provider. Committed businesses use an AI Roadmap to formalise which vendor handles which category of work, with portability designed in from the start.
Capable businesses have deployed AI across multiple vendors with clear governance, and Confident businesses revisit the vendor decisions as the market continues to shift, which it does constantly. The journey is what protects you from the kind of vendor concentration risk that Altman’s BlackRock comments make uncomfortable to ignore.
AI as a Utility: The Strategic Question Every SME Should Be Asking
AI as a utility is the future the dominant provider has publicly described, and whether or not that future arrives in exactly the form Altman pitched to BlackRock, the strategic implication for SMEs is clear. The businesses that survive market concentration are the ones that prepared for it. The businesses that ignored the early warnings will pay through pricing pressure, access constraints and the slow accumulation of dependency that becomes painful to unwind.
The pragmatic response is not to refuse to use AI on principle, nor to commit blindly to any single vendor’s vision. The response is to treat AI vendor strategy with the same care you would apply to any other foundational supplier decision, with diversification, portability and continuous strategic review built into the approach. SMEs that do this will keep their optionality as the market continues to shift through 2026 and beyond.
Complete our free AI Readiness Assessment to understand where your business currently sits, which vendors fit which workflows and how to build a multi-vendor AI strategy that protects your business from the concentration risk the industry is now publicly signalling.
