European politicians are feeling mightily pleased with themselves in 2025.
An EU singular focus upon ESG initiatives, according to environmentalists, produced another incremental victory for pro ESG media outlets when reporting that gross electrical power generation by renewables in 2024, at 44% of aggregate production, surpassed the 40% level for the first time ever. Wind generated electricity production surpassed natural gas.
Media types and the political class alike are now feting themselves for the 19% gross reductions in fossil and CO2 emissions. “If only the rest of the world would adopt the EU model” is the line of thinking, “just how much better off would the public and industry be as a result”.
What is completely overlooked in this frenzy of self-congratulation is one very salient point; aggregate European power generating capacity continues to fall and stands at the same capacity level as 2007. Yes, renewable energy production has risen, but a greater reduction in conventional electrical generating capacity, resultant from lack of grid investment and by political edict, has more than offset the growth in wind and solar output.
The electrical grid in Europe, for all the collective back-slapping, is far too strained to respond to the incremental demand required by artificial intelligence data centers
Even generative AI response, driven by data input designed to trumpet the emissions reduction from conventional electrical capacity (framed from searches designed to promote bias in favor of ESG), misses the point.
It took a quarter century for net power production by renewables, primarily wind and solar, to double in Europe, yet total conventional and renewable electrical generating capacity remains mired in the 2.7 terawatt range for 2025, down from 2.85 terawatt in 2008. Europe has no ability, whatsoever, to supply any electricity for the power intensive data centers that are considered to be key to AI. Conventional base load electrical capacity continues to be removed from the grid, beyond what is added by renewables. Not only is the reduction in power production absolute, the removal of base load capacity makes the grid less dependable.
A lack of electrical capacity and a lack of political interest in growing electrical capacity from European politicians effectively kills the AI model for the entirety of Europe.
European electricity pricing is roughly 2x higher than the United States and China, cementing the lack of viability of major data centers to be located within the EU.
Oracle has added some color to their enterprise data ambitions on the heels of their AI data contract with OpenAI.
All of the initial data centers to be housing the AI cloud in the OpenAI contract are to be based in the United States. Oracle is also building and operating more cloud data centers than ALL other competitors, combined, for data storage.
https://www.datacenterdynamics.com/en/news/openai-confirmed-to-be-behind-30bn-a-year-oracle-cloud-deal-45gw-expected-across-multiple-data-center-sites/
As an aside, the size and open ended storage capacity of the contract, suggests that analyst forecast revenues for OpenAI, as improbable as they may appear to Joe Q Public, are likely to be understated. A single $30 billion per annum contract by 2028 with Oracle by OpenAI, taking into account the point that Oracle is just one of several data storage providers for OpenAI, makes a $90+ billion gross revenue forecast for OpenAI, going into 2029, seem conservative. Data storage rental costs will be a function of revenues, and in a “winner take all” scenario, margins will be far higher than expenses.
European investors, seeking a local champion in the AI race, are highly likely to be disappointed in the coming years.
There are no compelling reasons for even a single enterprise data center to operate in Europe, given the lack of electrical capacity and the absurdly high operating costs that would ensue from European location. No doubt, as lip-service to the effete political aristocracy of Europe, a few token centers will be built, likely funded by European tax-payers. They won’t operate as profit centers for a major enterprise data company. When European political overseers refuse to meaningfully increase power generating capacity during an actual war (Russia/Ukraine), what hope do European investors have for a change in policy during an AI data storage economic war?
Europe’s loss, Oracle’s gain.
One cannot really frame AI enterprise data storage as a “space race” when much of the world removes themselves from competition due to political intransigence. The speed required to build out centers eliminates Europe, Canada and many other locales from contention. Canada is enthralled with their newish Prime Minister, a devout Europhile, who might prefer to direct a nation with abundant untapped and unbuilt potential electrical capacity, down the same path as the EU, towards central planning irrelevance.
This leaves only the US as a nation with the will to build out enough incremental electrical output. It takes roughly 18 months to build a data center and traditionally it takes about 3x as long to develop an electrical power plant and connect it to a grid. Approximately 100 GW of US data center plants are permitted and/or under construction, with surplus electrical supply available, or coming, in time to meet this new demand. What data centers require, above all, is base load capacity; brown outs are intolerable. The middle east also has interest, and Oracle will site out locations there as well.
Oracle understands the benefit of working in “pro-power” regions. The media and investors, as yet, may not. When much of the developed world refuses to compete in the AI data storage build-out, it isn’t a race, it is monopolistic or oligopolistic, with monopoly or oligopolistic margins by default.
Global data maintained in centers run by American companies and domiciled in the United States means America wins and the American business warehousing the data wins, because Europe, Canada and the remainder of the developed world (ex China and Russian vassals) will be forced to rent back their own knowledge, accumulated over many decades. The data rental, like an introductory cell phone contract, initially will be priced attractively, only to get increasingly expensive as time goes on.
Artificial intelligence generative responses, to superficial queries, promote a bias that Europe remains in the hunt.
As yet, AI cannot wrap its head around the lack of logic involved in ESG. For those who doubt the veracity of my assessment, or for those who have abdicated their knowledge base to AI algorithms, do a simple search yourself.
Type in a phrase such as “European electricity production decline in 2024” and what comes spilling out is summary after summary describing, not the reduction in net electrical output, but rather, the reduction in C02 emissions due to a shift away from conventional sources by renewables. The summaries point out the increase in renewable electrical production, but not a peep, not a data nugget whatsoever, on the net decline in output. Then, change the query to “European electrical productive capacity decline”, and what comes out is more of the same. Yet, assuredly, AI has been trained to differentiate the word “emissions” from the phrase “electricity production”.
There is no search that one can conduct, using the AI bots, that will bring up the reduction in grid output, nor the consistent decline in EU total electricity production, going back decades. Because that data is effectively hidden, sequestered from the public. An insufficient volume of queries from intrepid investors exist to make the AI algorithms pay attention and train it accordingly. As a result, the AI output represents a deliberate misdirection. A search for simple data, previously readily available prior to the overlay of AI algorithms, now produces an outpouring of propaganda.
This reinforces the investment value of my blog post, more than a year back, stating that AI doesn’t actually need to be smart, just smarter than us. We make AI seem smarter by ourselves becoming dumber, by delegating our own research, vital to growing our life savings, to biased bots.
Unbiased, unfiltered data still exists; you only find it by circumventing generative bot bias.
Leave a Reply
You must be logged in to post a comment.