OpenAI CEO Sam Altman recently proposed that artificial intelligence will eventually be sold like electricity or water, metered by usage and paid as a utility. But a new analysis from KPMG Canada warns that this vision, while compelling, poses a significant strategic risk for businesses that adopt it without building internal understanding of how these systems work.

Why Is Renting AI Intelligence Risky for Your Business?

The appeal of Altman's utility model is straightforward: companies purchase "tokens" of intelligence the way they pay their power bill, consuming what they need when they need it. The problem, according to Dr. Andrew Forde, Head of AI Research at KPMG Canada, is that intelligence is fundamentally different from electricity. Electricity powers a factory the same way for every competitor. Intelligence shapes pricing models, risk decisions, capital allocation, product design, and customer strategy. It is not a neutral input; it is the logic by which organizations decide.

Dr. Andrew Forde, Head of AI Research at KPMG Canada

"If intelligence becomes a metered utility controlled by a handful of providers, then decision making becomes capacity-constrained infrastructure," explained Dr. Andrew Forde, Head of AI Research at KPMG Canada.

Dr. Andrew Forde, Head of AI Research, KPMG Canada

When companies rely on AI platforms built and controlled elsewhere, they face several interconnected vulnerabilities. Access depends on compute availability, pricing power, and platform terms. When supply tightens, prices rise. When models evolve, dependencies deepen. In an extreme scenario, American tech companies could theoretically restrict intelligence access for Canadian firms during geopolitical tensions.

What Are the Hidden Costs of Moving Too Fast Without Understanding?

Canadian companies are racing to deploy AI pilots, copilots, and predictive systems, often through platforms built elsewhere. The competitive pressure is unmistakable, but the rush to keep pace creates a quiet tradeoff: gaining speed today by surrendering understanding tomorrow. Many organizations are making core business decisions using systems that fewer leaders can explain or control.

The real risk is not moving too slowly on AI; it is moving superficially. Deploying tools without building internal understanding means hardwiring dependency into enterprises. Companies outsource core decision-making logic to foreign companies, allowing their proprietary data to strengthen ecosystems they do not control. They may discover, too late, that they no longer understand the engine driving their own margins, risk models, or customer strategies.

This concern is already materializing in practice. Software-as-a-Service (SaaS) companies are changing terms to restrict customers from freely using data stored in their systems to train AI models. Organizations can put their data into the platforms, but they may not be able to use it to build intelligence outside of it.

How to Build Strategic Control Over Your AI Capability

• Invest in Research Literacy: Organizations that take AI seriously invest in building institutional depth to rigorously study how these systems work, how platform terms evolve, how models are governed, and how data rights shift. This is not about becoming technologists; it is about becoming informed stewards of technological risk and opportunity.
• Approach AI From First Principles: When organizations approach AI from first principles, including mathematics, systems theory, and empirical validation, they gain something far more valuable than efficiency: they gain agency. Their AI capability becomes a strategic asset rather than a vendor-managed feature.
• Build Partnerships With Research Institutions: Tighter partnerships between universities, governments, and industry are essential. This means converting research strength into applied enterprise capability, shaping, governing, and scaling the systems that will define your economy, rather than only adopting tools built elsewhere.
• Engage With Frontier Models While They Evolve: Creating value and managing in fast-moving environments requires more than implementing emerging technologies once they are mature. It requires engaging with them while they are still evolving and collaborating more directly with the companies and labs developing frontier models.

For Canada specifically, the stakes are especially high. Canadian researchers helped pioneer the scientific foundations of modern AI, yet scientific leadership does not automatically translate into economic leadership. Canada has one of the lowest levels of business-driven research and development investment in the G7, with the private sector accounting for roughly 47 percent of total R&D spending. The country produces breakthroughs but too rarely turns them into lasting commercial advantage.

"In a market moving at this speed, surface-level adoption is easy. Serious understanding is harder, but it is the difference between leading transformation and being shaped by it," noted Dr. Andrew Forde.

Dr. Andrew Forde, Head of AI Research, KPMG Canada

Financial institutions managing risk in markets that move at machine speed, health systems serving aging populations with constrained workforces, and energy companies making decades-long infrastructure decisions in an era of geopolitical instability all require more than implementing emerging technologies once they are mature. They require leaders to move upstream: to invest in research literacy, build internal technical depth, collaborate more directly with frontier model developers, and ensure their own data, talent, and governance structures are positioned to shape outcomes rather than simply absorb them.

The fundamental question facing business leaders today is not whether to adopt AI, but whether to understand it. If businesses simply consume the systems shaping their future, they will not be building that future; they will merely be living in it, renting the intelligence necessary to build it along the way. And renting intelligence is a risky way to run a business.