How Fast Will AI Data Centers Become Obsolete?
And why the recent jobs report is positive for gold...
As we enter the new week following the long July 4 holiday weekend, here’s a quick look at a few things that caught our attention last week and that strike us as significant for the economic and market outlook.
We’ll start with the June jobs report, released last Thursday. We’ve seen it characterized as “underwhelming.” We’d call it crummy.
The U.S. economy added just 57,000 jobs, around half of what had been projected and the lowest figure since February. Moreover, job growth for each of the preceding two months was revised downward. (We sometimes wonder why investors always await the end-of-the-month release of job figures with such intense interest, since almost invariably they prove unreliable and are subsequently revised—in recent years, almost always downward.)
There was one supposed bit of “good” news: a slight drop in the unemployment rate to 4.2% from the month-earlier level of 4.3%. But unfortunately, in reality this was anything but positive, because the decrease simply reflected a drop in the labor force participation rate, to 61.5% in June—the lowest level in more than five years—vs. 61.8% in May. In other words, it’s not that more job seekers had found jobs. It was that more people without jobs have become so discouraged that they’ve stopped looking and therefore were no longer counted as part of the labor force.
This discouraging labor report reinforces our assessment that you can forget about interest rate hikes. Instead, the Fed’s next moves will be rate cuts. We say “reinforces” because, as we wrote last week, we felt that Warsh never really intended to raise rates and had been deliberately disingenuous in giving the impression that he would. Under cover of making hawkish noises, which he needed to do to show he wasn’t Trump’s puppet, he was preparing the ground so as to justify an about-face from hawk to dove. The employment news gives strong support to the case for rate cuts, and we expect the rest of the Fed members, or most of them, will be on board. They will set the scene for a highly inflationary economy—and strong gains in gold.
Meta’s AI pullback
Besides the jobs report, a couple of other things are worth noting. One was the news from Meta that it now plans to sell computing power to other companies in order to monetize excess AI infrastructure it has built. It’s likely a lot of it will be sold to Anthropic. Why is this significant? Because if Meta is pulling back on developing its own AI, which was the original
impetus for the infrastructure it has built, it’s a sign that the company doesn’t see much of a future in the LLMs at the heart of the massive data centers AI companies in the U.S. are racing to build. We’ve written a lot about the futility of focusing on LLMs, and this is another bit of evidence that supports our view.
A second AI-related development came from OpenAI, which announced it has decided to delay its IPO from late 2026 to sometime in 2027. The reason was that the company had realized that at the current time, it wouldn’t be able to achieve the $1 trillion valuation it wanted, and it wasn’t willing to settle for anything less. Of course, in a sane world, the very thought of assigning a $1 trillion value to a company like OpenAI—which is estimated to be losing more than $10 billion a year—would be laughable. But then, sanity hasn’t been a salient feature of today’s world.
Meanwhile, the IPO for Anthropic, which is expected to top $1 trillion, is still slated to proceed as planned for later this year. Anthropic also has an awful balance sheet, but somewhat less awful than OpenAI’s. Moreover, it has created an image of itself as the company with the most powerful AI by virtue of government restrictions on two of the company’s most advanced AI models, Claude Mythos 5 and Claude Fable 5, which supposedly are too dangerous to be let loose on the market for all to use.
Overvalued AI companies have been powering the stock market’s gains. They’ve also made up a big share of the economy, without contributing much of real worth. In the last 18 months, the AI data center build-out has accounted for a disproportionately large share of GDP growth—at least 30% and with some estimates putting it as high as 70%—as companies have poured money into them without any commensurate gains in employment, as the jobs report makes plain. But something else that’s important to realize is that in the not-too-distant future, the data centers face a very real risk of obsolescence. The Nvidia chips that power them will be overtaken by a new generation, leaving the companies to deal with these hulking monsters—writing off their past investments and further stressing their finances as they look for ways to finance whatever they need to do to rebuild or retool.
With both the market and the economy having grown so dependent on a technology that could soon become outmoded and on companies that may be facing a reckoning, the outcome could be catastrophic. The potential for chaos is why we remain committed to gold and to other investments that will endure when the AI mania is exposed as hollow.
