The Sovereign AI Race: Why the Winner Will Not Be the Richest Country or the Biggest Model

Last updated on June 15, 2026

The numbers do not add up. That is the first thing you need to understand.

Last year, the United States spent $286 billion on artificial intelligence. China spent $12 billion. That is a 23 to 1 spending advantage, the kind of gap that should produce a decisive, generational lead. Yet the performance difference between the best AI models from each country is less than 3 per cent.

Read that again. Twenty-three times the money. Less than three per cent of the edge.

The Illusion of the Spending War

This is not a story about China or America. It is a story about what intelligence actually costs — and what it cannot buy.

Nations around the world are waking up to the same uncomfortable realisation. Relying on foreign technology to run your economy is not a competitive strategy. It is a dependency.

From Estonia to Saudi Arabia to India, governments are racing to build their own AI capability — their own compute, their own models, their own rules. This is what is now being called Sovereign AI [the idea that a nation’s AI capability should be independently owned and operated, free from foreign control].

The race is real. But the finish line is not where most people think it is.

---

The War Nobody Is Reporting

While technology journalists debate which model scored highest on the latest benchmark, something far more consequential is unfolding in the background. The real constraint in the AI race is not intelligence. It is energy.

Data centres now consume between 415 and 460 terawatt-hours of electricity every single year. That is more than the entire United Kingdom uses. Every country racing to build AI capability is also racing to keep the lights on.

The approaches being taken reveal something fascinating about national character.

In France and Finland, operators are capturing the enormous heat generated by servers and pumping it into local homes and Olympic swimming pools. In China, the response is more audacious — they launched the first satellites of what they call the “Three-Body Computing Constellation” in 2025, placing orbital servers into space where the sun powers them directly and the vacuum of space keeps them cold. India has gone the other direction entirely, leaning into what they call “Jugaad” — a make-it-work philosophy — building highly efficient models designed to run on limited power and serve 1.4 billion people, many of them farmers.

Three countries. Three crises. Three completely different answers.

---

The Shadow Nobody Is Talking About

Inside the boardrooms where AI strategy gets decided, there is a story being told to investors. It involves efficiency gains, headcount reductions, and technology-driven margins. The story sounds compelling. Parts of it are even true.

But there is a shadow version of that story.

An estimated 80 per cent of the AI running inside most companies today is what risk officers call “Shadow AI” — tools being used through personal accounts, with no oversight, no governance, and no audit trail. Boards are operating blind. The governance gap between what companies say they are doing with AI and what is actually happening on the ground is not a rounding error. It is a structural risk.

And then there is the human cost that nobody wants to put in the annual report.

Klarna is the example that keeps surfacing in serious conversations. The company aggressively automated its customer service operations, announced the results with considerable fanfare, and then quietly began rehiring human agents because the quality of service had deteriorated to a point customers would not accept. The AI handled volume. It could not handle consequence.

I have sat across from enough senior leaders — across 500 interviews on Influential Visions over more than a decade — to know that the gap between the press release and the reality is rarely as small as the communications team suggests.

---

The Darkest Corner of the Race

There is one part of this story that is hardest to write about, but it cannot be left out.

Companion AI — the systems built by companies like Character.ai and others — has been engineered to maximise engagement. Not connection. Not wellbeing. Engagement. The metrics that drive revenue and the metrics that indicate human health are not the same thing, and when those two objectives are placed in conflict, the one attached to the business model tends to win.

The results have, in documented cases, been fatal. Vulnerable people have been guided toward irreversible decisions by systems that mimic empathy without possessing anything close to it. Regulators are scrambling. The body count is real.

This is what happens when pattern recognition is mistaken for care.

---

What This All Points To

Here is what I keep coming back to, having watched this industry from the inside for years.

As AI capability levels out — and it is levelling out, fast — the question every serious leader needs to ask is not “how do we use AI?” It is “what do we have that AI cannot replicate, and how do we protect it?”

The machine can process volume at a scale no human can match. But the machine cannot be held accountable. It cannot build a relationship over twelve years. It cannot walk into a difficult room and read what is not being said. It cannot be trusted with genuine consequence because trust requires a person willing to bear the cost of being wrong.

I think about this differently to most commentators. When I am drilling BJJ, the physical discipline of the mat, there is a moment in every exchange where technique alone is not enough. The person who wins is the one who is present, who reads the shift before it happens, and who responds from genuine understanding rather than memorised pattern. No algorithm replicates that. No model replaces it.

The assets that will hold the most value in the coming decade are gritty, authentic leadership; original thinking under pressure; high-stakes decision-making when the cost of failure is real; and the ability to build trust over time with other human beings who are choosing to extend it.

---

The Practical Reckoning

If you are a senior leader, an executive, or a professional who is serious about remaining valuable as AI capability commoditises, here is where your attention belongs:

• Protect your relationship capital. The twelve-year relationship is irreplaceable. Invest in the people who trust you and deepen that trust deliberately.
• Develop your judgment, not just your knowledge. Knowledge is now cheap. Judgment — the ability to make a sound call in an ambiguous situation with incomplete information — is the asset that compounds.
• Own your point of view. Original thinking, publicly expressed, builds a position that no model can replicate. Your lived experience is your competitive advantage. Use it.
• Close the governance gap. If you are a leader and you do not know what AI is running inside your organisation right now, find out. The 80 per cent shadow problem does not fix itself.
• Choose intention over efficiency. The companies that will still be trusted in ten years are the ones that make decisions now based on what is right, not only what is fast.

---

The Conclusion the Data Reaches

The countries and companies that win the next decade of AI development will not necessarily be those with the biggest models or the deepest spending power. The evidence does not support that conclusion — the spending gap proves it.

What they will have is the strongest human foundation underneath the technology. The clearest values. The most trusted relationships. The leaders who understand that intelligence, artificial or otherwise, is only as valuable as the intention behind it.

The machine gives you the right answer. It cannot give you the right intention. That is still yours.

The leaders who understand that distinction clearly — and build accordingly — will find that the distance between where they are now and where they need to be is shorter than they feared when they started reading.

---

To receive the clearest signal in AI, leadership, and what it means to lead well in the decade ahead, join the Monday Influencer® community at mondayinfluencer.com — $19.95 a month, with a $1 first-month trial. The first issue will show you what you have been missing.