Earnings Calls Tout AI Growth, but Jobs Don’t Follow

A recent Barron’s commentary begins with a joke that isn’t really a joke. It’s like a knock-knock gag, but it isn’t funny. (Well, maybe it’s a little funny, depending on your sense of humor). It’s not a riddle, either, because the question is answered, the puzzle apparently solved. Here’s how the Barron’s piece began:

When does $360 billion buy you fewer jobs? When it is spent on artificial intelligence.

The world’s largest tech companies, all of them based in the world’s biggest economy, say they will spend that much this year alone to build the data centers that will house their AI technologies. Planned investment to build, run, and power AI infrastructure could rise to around $7 trillion by the end of the decade, according to a study from McKinsey.
At present, however, the staggering sums are simply not having a positive impact on the job market. And AI development is likely to keep a lid on hiring over the mid and longer term.

We can already see this trend, can’t we? Spending on AI datacenters and infrastructure has reached staggering numbers, yet we’re not seeing commensurate job growth. Enviable, sometimes unprecedented, financial results are recorded by companies such as Nvidia, Oracle (!), Microsoft, and Google on a quarterly basis. At the same time, purveyors of chatbots and other AI services are raking in investment funding, spending a large portion of that the money on AI infrastructure.

Yet, even as money is splashed around like beer at a frat party, even as extravagant spending inflates the revenues and profits of vendors providing AI “picks and shovels (and earthmovers, don’t forget those),” we find that the companies recording skyrocketing revenues and distended earnings, boasting impressive shareholder returns, are barely adding to their employee payrolls.

The Barron’s piece goes on to note that the U.S. domestic economy has lost approximately 78,000 manufacturing jobs to date in 2025, according to data published recently by the Bureau of Labor Statistics. Oddly, areas where you’d expect expansion, such as heavy construction and civil engineering — involved in the construction of datacenters and the expansion of power grids — have added just 8,000 jobs this year. In Virginia, a perennial datacenter hotspot, where about 330 datacenters have been built or are in the planning stages, the count of construction jobs is down by 10% from last year’s numbers.

Those numbers, or lack of numbers, seem at variance with the prodigious sums that Amazon, Microsoft, Alphabet, and Meta Platforms (among others) are spending on datacenters larger than football fields. Meta has indicated that it intends to build datacenters as big as Manhattan, a proposition as monstrous as anything you’d find in the grimmest dystopian science fiction. How will these new-age leviathan’s be powered, how will they be cooled? What sort of opportunity cost — for land, water, energy, alternative development initiatives — will accrue to these city-sized datacenters?

A New Kind of Growth?

Economist Paul Kedrosky, quoted in the article, observes that AI-related capex accounted for slightly less than half of GDP growth in this year’s second quarter. Given its relative dearth of job creation, it’s a strangely disembodied, impersonal growth, spawning massive buildings with few people working within them.

There are obvious concerns about economic growth that doesn’t produce many jobs; or, to be more precise, growth that doesn’t produce jobs on the same scale as prior generations of conventional or tech-driven growth.

We could have political and geopolitical arguments about the provenance of the infrastructure (servers, switches, routers, GPUs, CPUs, etc.) that fills those AI datacenters. We could argue about where it should be manufactured or assembled, about the imperative to relocate supply chains, and so on and so forth. Go ahead and have those arguments, but not here and not now. That’s not on the agenda today.

We’ve had waves of technological growth where jobs were readily and commensurately spawned, even as many of the tools and infrastructure we’ve used — from smartphones to desktop and portable computers, to cell towers and cabling, and switches and routers and servers — were manufactured and assembled globally, mostly in Asia.

The salient issue on the docket today is not about where things are made. In the past, growth has come from economies and industrial sectors that generated jobs at approximately the same pace — though not at the same levels of reward — as the gains that accrued to shareholders. A basic (though by no means perfect) reciprocity of prosperity existed, a functionally sustainable sharing of the wealth. Now? Maybe not. This could be a turning point.

If Consumer Spending Weakens, Nearly Everybody Suffers

That concerns me, and it might also concern you, because so much of the overall economy is dependent on consumer spending. If job growth slows or flattens, the consumer-driven core of the economy will sputter and stall. If that happens, problems will ramify throughout the rest of the economy, from real estate to service industries (food, beverages, entertainment), to retail, to wholesale, to transportation . . . well, to nearly everything that consumer spending touches, directly and indirectly. It will even — horror of horrors — have a detrimental effect on spending on cloud and AI services.

Aside from the incommensurability of spending on AI datacenters and infrastructure to job creation, is AI at a point where it is directly eliminating jobs or precluding job creation? There’s no incontrovertible evidence to suggest that is happening. The Barron’s piece mentions that Challenger Gray & Christmas (a Dickensian company name if ever there was one) reported recently that AI was linked to about 31,000 job losses this year, but the article doesn’t explain AI’s connection to the job losses. It’s hard to comment intelligently on something that isn’t clearly stated. I could try, firing conjecture and speculation from the Pez dispenser of my mind, but what’s the point? It would be all heat and no light, like most of social media.

The Council on Foreign Relations and the New York Fed, we learn, concur that there’s little evidence to prove conclusively that AI has a causal relationship to the cooling job market. That situation might change. We’re still in AI’s early days. Once thing that seems probable, if not certain, is that we’ll get some surprises along the way.

AI Job Displacement Lurks

What is true today might not be true, or equally true, in the future. A consensus is forming that AI will displace jobs, and doubtless preclude the creation of others, in the months and years to come. Nobody seems to know, reasonably enough, how many jobs will be affected, or whether and how soon AI will create other types of jobs to replace the ones it renders superfluous.

We can, I think, reasonably assume that there will be a temporal lag between when existing jobs are displaced and when new ones are created. Will that lag be months, or, more likely, years? How many years? We do not and cannot know — not yet, anyway.

What we do know is that the stocks of the biggest companies in the technology firmament have risen in value even as job creation has advanced more slowly, sometimes imperceptibly. Is that sustainable?

Here’s how the Barron’s commentary concludes:

That could leave investors to wonder whether the productivity gains promised by AI can offset the fact that companies will have fewer people to whom they can ultimately sell their products. Greater efficiency enhances earnings, but revenue generates it.

Perhaps there’s a way to pivot from an economy that is mostly driven by consumer spending to one that is increasingly influenced by AI and autonomous processes that obviate or mitigate the need for human labor. If there is such a way, it will require considerable forethought now, well in advance of what looks, from our vantage point today, to be an abrupt turn into the unknown.