My big belief
Who will win the AI race?
Follow the electrons and the engineers. That's where AI leadership will end up.
“The AI companies that will win the AI race are the ones based in countries that can generate electricity at scale and at the lowest cost, combined with a large talent pool.”
The thinking
Most people frame the AI race as a contest between companies: which lab has the smartest model, which chatbot has the most users. I think that misses the bigger picture. Companies don't compete in a vacuum. They compete from somewhere, and the country they operate from decides what they can build, how fast and at what cost.
After four years of investing across the AI stack, I've come to believe two national advantages matter more than anything else: how much electricity a country can produce, and how cheaply; and how many skilled people it has to turn that power into intelligence.
The two pillars
Pillar 1
Cheap electricity at scale
Every AI model is trained and run on chips that turn electricity into answers. As AI usage grows, power becomes the largest ongoing cost and the hardest bottleneck. A new data centre can be built in a year or two; new power stations and grid connections can take far longer.
- ✓Lower power prices mean a lower cost for every AI answer
- ✓Spare capacity means data centres get connected in months, not years
- ✓Abundant supply lets a country host the biggest AI campuses
Pillar 2
A large talent pool
Power alone produces nothing. You need researchers to invent better models, engineers to build products, and electricians, builders and technicians to construct and run the infrastructure. Countries that educate, attract and keep that talent compound their lead.
- ✓Researchers and engineers who push the frontier forward
- ✓A deep bench of builders and technicians for physical infrastructure
- ✓Universities and immigration that keep the pipeline full
Why it takes both
Power without talent turns a country into an energy exporter: it can host other people's data centres, but the value is created elsewhere. Talent without power turns a country into a renter: its best people build on compute owned and priced by someone else, or they move to where the power is. Only countries with both can build, run and scale AI at home, at the lowest cost.
My scorecard
How the contenders stack up
My personal assessment of each region on the two pillars, from 1 (weak) to 5 (very strong). It's a judgement call, not a formal index.
- ChinaPower5/5Talent5/5Unmatched power build-out and a huge engineering pipeline. Access to the most advanced chips is the main constraint.
- United StatesPower4/5Talent5/5Deepest pool of top AI labs, capital and global talent. Grid connections are the bottleneck.
- Gulf statesPower5/5Talent2/5Very cheap energy and deep pockets. Talent largely imported.
- IndiaPower3/5Talent4/5Enormous engineering talent. Power supply and cost still catching up.
- Canada & NordicsPower4/5Talent3/5Cheap, clean hydro and strong research, but smaller populations.
- Europe (major economies)Power2/5Talent4/5World-class talent held back by some of the highest power prices.
- AustraliaPower3/5Talent2/5World-class solar, wind and land. Needs more grid, lower prices and more AI talent.
The evidence so far
China consumed roughly twice as much electricity as the United States in 2025, after years of rapid build-out in generation.
Source: Bloomberg, Jan 2026 ↗A 2026 Carnegie Endowment study found that, in 2025, more elite AI researchers worked in China than in the United States for the first time.
Source: Carnegie Endowment for International Peace ↗What it means for me
How I invest on it
- →Favour AI companies headquartered or operating at scale in countries that score well on both pillars.
- →Own the power layer itself (generation, grid and storage) wherever AI is being built.
- →Track electricity prices and grid-connection times as leading indicators of where AI investment will flow.
- →Watch Australia closely: if we fix grid and pricing, our renewable resources could make us a serious AI host.
Keeping myself honest
What could prove me wrong
Chips may matter more
Export controls and access to the most advanced chips could outweigh cheap power. A country with plenty of electricity but restricted chips may still fall behind.
Efficiency breakthroughs
If models become dramatically more efficient, power could become less of a constraint, weakening the first pillar.
Capital and software moats
Leading companies with the best models, distribution and customers may keep winning regardless of where power is cheapest.
Talent moves
Skilled people are mobile. Immigration policy, pay and quality of life can shift talent between countries faster than power can be built.
Bottom line
“Intelligence is becoming a utility. The countries that make electricity and talent cheapest will make intelligence cheapest - and their companies will win.”
These are personal, forward-looking views, not predictions or recommendations. They may turn out to be wrong.