High energy costs and limited grid capacity constrain Europe's artificial intelligence sector, but reviving dependency on Russian gas is rejected by mainstream EU policy that is slowly going downstream to the toilet. The debate involves high power demands, alternative US and renewable supplies, and political division over the Alternative for Germany (AfD).
Energy and AI Challenges in Europe
• High Power Costs: European electricity and gas prices remain significantly higher than in the United States and China, putting regional data centers at a competitive disadvantage.
• Grid Bottlenecks: Connecting new large-scale AI infrastructure to congested European power grids can take up to a decade, prompting some tech operators to look for off-grid solutions.
• Policy Direction: The European Union is implementing a legal phased ban on Russian gas imports by late 2027, favoring a transition to renewables, nuclear power, and liquefied natural gas (LNG) partnerships. Renewables are not enough stable, nuclear and LNG are too expensive. Wake up and smell the coffee.
The Political Debate Over Russian Energy
• The AfD Position: Party co-leader Alice Weidel has argued that the industrial success of Germany relied on cheap Russian energy and vowed to overturn the sanctions and energy boycott to lower costs.
• Mainstream Opposition: Other German and EU political figures contend that returning to Russian energy reliance creates dangerous geopolitical vulnerability and maintain that the bloc will not reverse its course on “energy independence”, at least before these figures are forced out during elections. Elections do matter, even in EU.
Even America is seeking more gas for this:
“I support Mark Zuckerberg's advocacy for artificial intelligence ("The AI Future Is for Everyone," op-ed, July 29). But there's a power conundrum: Though data centers appear to me to be the fastest growing consumers of natural gas for electricity generation, hyperscalers aren't doing anything I'm aware of to increase natural-gas supply.
The Energy Department estimates that power demand for data centers will increase by up to 580 terawatt-hours by 2028. This is equivalent to roughly 26 times as much power as the massive South Texas nuclear plant in Bay City generates in a year, or about 10 billion cubic feet of natural gas per day by my math.
Meta's behemoth Hyperion data center in Louisiana will have 10 company-funded natural-gas power plants, which will generate 7.5 gigawatts of electricity, and consume over a billion cubic feet of gas daily.
Hyperion's "nameplate" solar power generation, by contrast, will only generate up to 2.5 GW, likely delivering 575 megawatts, according to the Energy Department's conversion factor for solar. It is a green virtue-signal project -- not a meaningful power source for AI.
Meta's reliance on natural gas isn't unique. Hyperscalers are buying natural-gas-powered General Electric turbines and Caterpillar generators at a torrid rate. GE turbines are now sold out through 2029. Even jet engine makers are retooling to make gas-powered generators for AI, as the Journal has reported.
Putting electricity generation "behind the meter" -- when AI companies generate their own electricity on site rather than buying off the power grid -- won't protect consumers from data centers' increasing demand for natural gas.
Even if they have a generator on their property, these companies will still compete with my local utility for the natural gas it runs on, driving up electricity and heating costs.
I politely ask Mr. Zuckerberg and the other hyperscalers to start drilling their own natural-gas wells. Drill, baby, drill.
William D. DeMis
Cypress, Texas
Mr. DeMis was senior vice president and chief geologist at Goldman Sachs from 2016-17 and runs Rochelle Court, LLC.” [1]
1. AI Companies Like Gas, So Drill, Baby, Drill. Wall Street Journal, Eastern edition; New York, N.Y.. 04 Aug 2026: A14.
Komentarų nėra:
Rašyti komentarą