The AI Power Surge Could End Badly

• StockStory Top Pick NVDA -3.15%
• 9984.T -10.90% KLAR -7.62%

Is AI in a bubble? When will the bubble burst? Are there signs? Well, we don’t have a crystal ball, but there are signs. On the markets, several prominent investors have spoken. Peter Thiel, one of the best-known tech investors, has sold his entire position in Nvidia. SoftBank, a huge Japanese tech investor is unloading its shares, too. Meanwhile, Michael Burry, of Big Short fame, warns of an impending bubble burst. And the head of Swedish fintech Klarna, a man with major AI holdings, said that he’s thinking of hedging his positions in companies building AI infrastructure. Not that he thinks AI has no value, but rather that it can operate with a lot less computer power than implied by all the construction.

We’re never sure about the motives of high profile investors. They may want to influence the market to serve their portfolio purposes, to send messages, put their estates in order, or take advantage of tax quirks, or maybe they are following the advice of that old Wall Street adage: “Bulls make money and bears make money but pigs never make money.” However, there are a few other signs.

The electric utilities that will sell to AI centers have started to make demands, such as long term (15 year) contracts, payments for incremental costs of serving AI centers and even deposits to be put in the line for service, because they suspect that many of these applications are no more than place savers in the queue, that is, as one report put it, “phantom” data centers. Then there is the problem of fast changing technology. The Chinese have developed simpler, more focused, less energy intensive AI, we believe. Will that endanger the economics of our AI industry if applied here? And what about changes in computer technology, such as quantum computing?

Related: U.S. Backs Nuclear Plant Restart as AI Power Demand Soars

Then, the CEO of BlueOrigin, Jeff Bezos’ rocket company, made his prediction, that data centers were better located in space, where power would be abundant, and the data could be beamed down to earth. He said that the technology to do so was available and the move would happen “within our lifetimes” but he then narrowed that down to within the “next five to 10 years.” Before you brush this off, consider that NASA was considering power in space and on the moon more than two decades ago, but decided to do other things.

Now, let’s return to the decision making process in the electricity business. If investors (especially lenders) get cold feet, many of these centers won’t get financed or might not be completed.  As for the ones that are completed, do you want to recover your investment over a 15-year contract (or worse, the usual utility depreciation life of 30 plus years ) when your customer might be filing for bankruptcy in the next tech wreck? Serving data centers on a regulated basis will create a concentrated risk for the utilities whose cost will eventually be passed on to customers, who are already suffering from price increases unrelated to AI. Which is why we believe that AI should fund its own electricity system. The utilities can always manage it for them.

Are we daft, thinking that rocketeers or Chinese enterprises or computer geeks developing stuff that barely works now will affect the future of the biggest boom to the economy that we’ve seen in decades or that a few financial naysayers will stem the flow of money to the AI centers? (Next year, the top AI firms are predicted to spend $500 billion, which is twice as much as the electric industry spends in a year. The ratio of sales to capital expenditure for the big AI builds has halved from 2023 to 2025.)

On the other hand, isn’t it daft to believe that all this money pouring into fixed assets for a fast changing technological service will produce adequate remuneration? Technology firms traditionally do not do steel and concrete. The product changes too fast. What might explain what is going on?  How about FOMO for lemmings? But back to the electric companies.  Let’s say the demand does not materialize as much as expected (our favored scenario), or worse, it materializes, and then fades away. Those contracts become hard to enforce. That nuclear plant ordered to meet the new demand turns into a white elephant, the most expensive unit on the system. Prices go up to support unneeded infrastructure. As we said before, serving AI on a regulated basis increases the risk level and probably increases prices as well, and the other customers don’t need that.

To sum up, there are the developments to watch:

• Easy access to equity capital (everyone wants to invest).

• Overbuilding (everyone wants to build a data center ).

• Underutilized data centers (what happens when there are more data centers than needed).

• Shifting the risk to the utilities (they put up the infrastructure to serve, and if demand does not materialize, they foot the bill which they pass on to their customers).

• Rising prices (overall prices up and electricity prices up more).

• Political backlash (sense that the regulator failed, as evidenced by the Georgia electorate demanding change).

Here’s the odd thing. If investors get antsy and cut off easy access to capital, the bubble will burst, then there won’t be overbuilding or underutilized data centers, and that will reduce the risk to utilities and the pressure on prices, thereby lessening pressure on regulators. Of course a lot of people will lose a lot of money all at once, and the biggest industrial boom the country has seen in ages will end, and there goes the growth story for utilities.

Well, as we said before, we are not making predictions, just pointing out that AI entails risks for the utilities and their customers without clear offsetting benefits.

By Leonard Hyman and William Tilles for Oilprice.com

More Top Reads From Oilprice.com

• Sanctions Slow Russian and Iranian Crude Flows to China

• Russian Crude Piles Up but Oil Prices Refuse to Move

• EU's Fossil Fuel Phaseout Stance

Oilprice Intelligence brings you the signals before they become front-page news. This is the same expert analysis read by veteran traders and political advisors. Get it free, twice a week, and you'll always know why the market is moving before everyone else.

You get the geopolitical intelligence, the hidden inventory data, and the market whispers that move billions - and we'll send you $389 in premium energy intelligence, on us, just for subscribing. Join 400,000+ readers today. Get access immediately by clicking here.