AI's Addiction Problem: Chips, Power, and Water

When OpenAI dropped GPT-6 earlier this year, everyone was blown away by what it could do. But pretty quickly, the conversation shifted to something less sexy: infrastructure. Turns out, teaching AI to be smarter requires massive amounts of computing power — and that means we need more chips, more electricity, and yes, even more water than anyone anticipated.

AI isn't just a software thing anymore. It's become a full-blown resource grab, and countries around the world are scrambling to get what they need. We're talking about everything from chip factories in Arizona to hydroelectric plants in Quebec to nuclear facilities in France. This is getting real.

Let's start with the obvious one: computer chips. The U.S. passed something called the AI National Security Act last year, which basically says "we're not selling our best chips to China or Russia anymore." Beijing wasn't thrilled, so they dumped over $80 billion into their own semiconductor companies — Huawei, SMIC, and others you've probably never heard of.

But here's the thing: Taiwan is still king. TSMC (Taiwan Semiconductor Manufacturing Company) makes more than 60% of the world's advanced chips. Think about that for a second. Taiwan has become more critical to the global economy than Middle Eastern oil was in the 1970s. If something happens there, the whole world feels it.

The U.S., Japan, and the Netherlands are working together to control the technology that makes these chips possible. Europe realized they're way too dependent on Asia for semiconductors, so they threw €43 billion at building their own chip factories. Everyone wants a piece of the action.

But even if you have all the chips you need, there's another problem: power. These AI data centers eat electricity like you wouldn't believe. We're talking about a single large data center using as much power as a small city. And we're building them everywhere, doubling the number every year and a half.

The International Energy Agency did the math: by 2027, AI could be consuming more electricity than the entire country of Japan. Let that sink in.

Different countries are handling this in different ways. The U.S. is bringing back nuclear reactors near data centers in Texas and Virginia. Middle Eastern countries are going all-in on solar power for their AI facilities. China's redirecting coal power to what they call "strategic AI workloads" — basically, AI gets priority when the grid is tight. Europe's debating whether to put carbon limits on AI, which is... complicated.

Investors are catching on too. Some analysts are saying if you want to bet on AI, forget buying Nvidia — buy stock in energy companies instead. The real money might be in the power grid, not the algorithms.

Here's something most people don't realize: AI needs water. A lot of it. Data centers use millions of gallons every day just for cooling.

In places like Arizona — where Intel and TSMC are building huge chip factories — this is becoming a serious issue. Intel's facility there uses about 5 million gallons of water daily. Meanwhile, the Colorado River is literally drying up. Do the math on that one.

So what's happening? Cold countries are suddenly looking really attractive. Canada, Finland, Norway — they've got plenty of water, cool temperatures (which means less cooling needed), and clean hydroelectric power. We're seeing a shift where big AI companies are moving operations north, where it's easier and cheaper to keep the machines running without draining rivers dry.

What makes this moment different is that AI has become tied to national security. Countries treat computing power, data, and energy like they used to treat oil or gold. It's that important.

The world is basically splitting into three camps:

The U.S. and friends — focused on keeping supply chains secure and making sure democracies control the data.

China's bloc — all about self-reliance and not depending on anyone else for technology.

The resource-rich players — countries like Saudi Arabia and Canada that have the energy, land, and water to host everyone's AI infrastructure, and they're making bank off it.

Bottom line? The AI revolution isn't happening in the cloud anymore. It's happening in chip factories, power plants, and water treatment facilities. And the countries that control those resources are going to have a lot more power in the years ahead.