AI Data Center "Greening" Operation: Hydrogen, Solar, Offshore Platforms

Silicon Valley — How much electricity do AI data centers consume?

A large AI data center's electricity usage is equivalent to a small-to-medium city. With explosive AI growth, this number is still soaring.

What to do? Tech giants start "greening."

Hydrogen Power: From Concept to Implementation

March 3, Singapore companies announced exploring using floating hydrogen platforms to power AI data centers.

"Hydrogen is perfect clean energy," project lead said. "Only produces water, zero carbon emissions."

Specific plan: build floating hydrogen power platforms at sea, then pipeline electricity to data centers.

"Singapore has scarce land, offshore platforms are the best choice," he said. "Plus hydrogen can store, solving solar and wind power instability problems."

Green Energy Parks: New Site Selection Thinking

Another trend is "green energy parks."

Traditional data center site selection mainly considers network latency and land prices. But now, tech companies start considering "energy."

"Best to build data centers next to solar and wind farms," an engineer said. "Can get electricity nearby, reduce transmission loss."

Reportedly, Microsoft, Google are negotiating building "green data centers" in solar-rich areas.

Nuclear: New "Hot Commodity"

But the hottest is nuclear energy.

"Nuclear is the most stable clean energy," an analyst said. "Available 24/7, near-zero carbon emissions."

US recently approved multiple small nuclear power projects, specifically to power AI data centers.

"Nuclear's problem isn't technology, it's public acceptance," the analyst said. "But AI companies don't care — they only care about stable power supply."

China's Situation

China's situation is slightly different.

"China leads globally in new energy (solar/wind)," an industry person said. "So we prefer integrated solar-wind-storage solutions."

Specifically, build data centers in new energy-rich areas (like Inner Mongolia, Qinghai), paired with solar, wind, and storage equipment.

"This achieves 'green power direct supply,'" he said. "True zero-carbon data centers."

Cost Account

But "greening" comes at a cost.

"Hydrogen power currently costs 2-3 times of thermal power," the person said. "Nuclear also has long investment return periods."

However, they also believe costs will drop with scale: "Like solar power back in the day, now cheaper than thermal."

Who's Leading?

From public information:

Microsoft: Nuclear data center, operational 2028

Google: Green energy park, first projects 2027

Singapore: Offshore hydrogen platform, trial operation 2029

Microsoft: Nuclear data center, operational 2028

Google: Green energy park, first projects 2027

Singapore: Offshore hydrogen platform, trial operation 2029

"Currently still 'testing waters' stage," an investor analyzed. "Large-scale commercial use wait until 2028-2030."

Epilogue

At a climate forum, I met an environmental activist.

"AI companies were 'carbon emitters' before," she said. "Now willing to invest in clean energy is progress."

But she also reminded: "Can't just look at propaganda, need to see actual action."

Perhaps this is a long battle: AI needs to develop, environment needs protection. Can they balance? Time will tell.

Reference: Climate Change News, SolarQuarter, Drexel News Blog