AI Data Center "Arms Race": Tech Giants Build Feverishly, Power Companies Become "New Oil Barons

Silicon Valley — When you open ChatGPT, ask Claude, or generate an image with AI, you probably don't realize: behind these "cloud" AI services are countless "electricity-eating beasts" running.

AI data center electricity demand is exploding.

Tech Giants' "Land Grab"

Over the past 18 months, tech giants globally have been "racing for territory" — feverishly building AI data centers.

Microsoft: Announced over $100 billion investment in data centers Amazon: AWS's fastest growing business is AI data centers Google: Data center count doubled in past year Meta: Self-developed AI chips, reducing NVIDIA dependency

"This isn't expansion, it's war," said a veteran data center industry person. "Who has more data centers has more AI compute."

Power Companies: "Lying to Win"

The AI data center explosion's biggest beneficiaries aren't chip manufacturers, but power companies.

"AI data centers require 10x more electricity than traditional data centers," a power industry analyst said. "Before we begged tech companies to build factories, now they beg us for electricity."

Specifically, a large AI data center consumes electricity equivalent to a small-to-medium city daily.

This means: power companies are transforming from "utilities" to "strategic resources."

"Who controls electricity controls AI's lifeblood," the analyst said. "Power companies are becoming 'new oil barons.'"

"New Energy" Fight for Site Selection

But AI data center site selection faces challenges: Where does electricity come from?

Traditional thermal: Pressured by ESG, nowhere to go Hydropower: Cleanest, but site selection limited Nuclear: Most stable, but public acceptance low New energy: Wind/solar unstable

"Optimal solution is nuclear," a data center architect said. "Stable, clean, available 24/7."

That's why the US recently approved multiple small nuclear power projects — specifically to power AI data centers.

"Electricity Futures": New Financial Product?

Interestingly, the electricity demand explosion has even spawned new financial products.

"We're considering 'electricity futures,'" a Wall Street trader said. "Like oil futures."

Reason is simple: AI companies need to lock in future electricity prices, otherwise electricity cost fluctuations will affect AI service costs.

China's Situation

China's situation is slightly different.

On one hand, China leads globally in new energy (wind/solar); on the other hand, data center electricity guzzlers face "dual carbon" pressure.

"Chinese AI companies' challenge is: obtaining sufficient electricity while meeting environmental requirements," an industry person analyzed.

Solutions include: building data centers in areas rich in new energy, adopting energy-saving technologies like liquid cooling, even directly investing in new energy projects.

Epilogue

At an industry conference, I met a traditional energy company CEO. He told me they're massively investing in data centers.

"Before we thought tech companies were 'opponents,'" he said. "Now we found we can cooperate — they need electricity, and they need the fact that they need electricity."

Perhaps this is the era's transformation: energy companies from "sunset industry" to "hot commodity."

Reference: Bloomberg, Financial Times, The Verge