The Erosion Of The Virtual Moat

In the early days of the boom, you won by having something no one else had—rare talent, a proprietary dataset or a massive pile of GPUs. Today, those walls are getting shorter.

• Silicon Is Diversifying: Nvidia is still the giant in the room, but it isn't the only game in town anymore. Between Apple’s M-series chips, Google’s TPUs and new hardware from Qualcomm and Broadcom, the performance gap is closing. Compute is becoming a broad ecosystem rather than a single proprietary stack.

• Models Are Diffusing: We’re getting much better at training models efficiently. Between synthetic data and the explosive growth of open-source models like Llama, what once cost a billion dollars to build can now be replicated or even improved upon in a few months. "State of the art" is no longer a permanent advantage; it’s a moving target that everyone is hitting.

• Applications Are Compressing: Generative tools have made it incredibly easy to build software. Workflows can be cloned and interfaces can be rebuilt almost overnight. Differentiation is still possible, but it doesn't last nearly as long as it used to.

As the virtual layers become accessible to everyone, the real bottleneck shifts back to the physical world. There are two things you can’t download, shortcut or replicate overnight: energy and trust.

Energy As The New Bottleneck

Every single AI interaction is a physical event. It’s the literal conversion of electricity into computation and heat. When you look at it at scale, AI stops looking like a "tech" business and starts looking like heavy industry.

The demand is staggering. A single AI prompt pulls significantly more power than a traditional Google search. According to the International Energy Agency (IEA), data center electricity consumption is on track to hit 1,000 terawatt-hours by 2026. To put that in perspective, that’s roughly the same amount of power used by the entire country of Japan in a year.

But the real crisis isn’t demand—it’s supply. In major hubs across North America and Europe, data center expansion is already hitting a wall. Grids are maxed out, permitting takes years and cooling these high-density "intelligence factories" is becoming a massive engineering headache. Power isn't just an expense anymore; it’s a hard ceiling.

The Rise Of The Compute Refinery

This shift is creating a new economic model that I'll refer to as the compute refinery.

For decades, energy-rich nations followed a basic extractive model: dig up a resource, export it and let someone else turn it into a high-margin product. AI flips the script. Instead of exporting raw electricity or gas, countries can keep that power at home, run it through a data center and export "intelligence" instead.

Think of it this way: A kilowatt-hour sold to the grid has a fixed, low price. But that same kilowatt-hour, when used to run an inference model for a global company, is worth exponentially more. You aren't just selling power; you’re refining it into a finished digital product.

A New Global Map

We’re already seeing nations reposition themselves around this reality. The UAE is aggressively pairing its energy wealth with sovereign AI capital. Norway and Iceland are using their natural cold and renewable energy to attract massive workloads. Malaysia is rapidly scaling its grid to become a regional hub for hyperscale demand.

These places realize that reliable, scalable power is the "ante" to get into the game. If you can’t power the compute, you’re just a consumer in the new economy, not a producer.

The Case For Guyana

This shift gives smaller, energy-rich economies a rare chance to leapfrog. Guyana is a perfect example. Driven by offshore oil and gas, Guyana is moving toward a sustained energy surplus—a rare "problem" to have.

• Strategic Leverage: Rather than just being an energy exporter, Guyana has the option to pivot. By redirecting surplus power into domestic compute clusters, it moves up the value chain.

• The Cooling Advantage: This is the part people often forget. High-density compute generates an incredible amount of heat. Guyana’s river systems and Atlantic coastline provide a natural, massive heat sink. Managing heat is one of the highest costs in AI. Having a natural advantage there makes the infrastructure much more scalable.

The Trust Dimension

Energy is the engine, but trust is the steering wheel. As AI moves into sensitive areas like healthcare and finance, legal frameworks matter as much as power lines. Capital doesn’t just go where it's cheap; it goes where it's safe. Countries that can provide stable data residency laws, clear AI governance and predictable regulations will see investment move much faster. Trust is the final piece that connects a power plant to a global customer.

The Sovereign Intelligence Flywheel

When a country builds its own compute capacity, it creates a "flywheel" effect. Domestic compute builds local technical talent. Talent drives innovation. Innovation attracts more capital. This is the same pattern we saw with manufacturing in the 20th century, but now it’s being applied to intelligence.

The next decade will be won by the regions that can provide the power and the stability to run the world's thinking. In the industrial era, the winners were those who mastered extraction. In the AI era, the winners will be those who master refinement.

At the end of the day, energy wins.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?