Will AI Crowd Out Jobs?

Luddites old and new

According to a new paper by Stijn Van Nieuwerburgh, “a representative 200 MW AI training campus” — that is, a center that processes huge amounts of data to power models like ChatGPT and Claude — “costs roughly $8.2 billion.” Building such a center will temporarily employ a significant number of construction workers and indirectly employ a number of people producing the hardware and software that account for most of a data center’s expense. Once such a data center is up and running, however, typical estimates suggest that it will employ only between 50 and 150 workers.

If this sounds like an extremely small number, that’s because it is. U.S. workers are, on average, equipped with a lot more capital than their counterparts in most of the world. Even so, total fixed assets per worker — a concept I’ll explain later in this primer — are “only” around $700,000. So deploying $8 billion in capital, the cost of a data center, in a non-AI sector will, on average, support more than 10,000 jobs, while investing that money in a data center will support only around 100 jobs. Or to put it another way, money spent building data centers will, on average, create only around one percent as many jobs as investing the same sum in other industries would have created.

There has been very widespread discussion of the possibility that AI will worsen America’s already severe problem of income and wealth inequality. In today’s primer I will offer what I believe is a somewhat novel take on that issue — although, as I’ll explain, I’m following in the footsteps of David Ricardo, one of the founders of economics as a discipline.

My argument will be that so far all the evidence suggests that investment in data centers comes at the expense of other investments. As I argued last week, AI “crowds out” other potential uses of capital. Moreover, because data centers employ so few workers, the vast spending on AI currently taking place will also significantly reduce the demand for labor.

We are not talking about a marginal effect. Van Nieuwerbergh estimates that through 2032 hyperscalers will spend more than $10 trillion on data centers — capital expenditure that, all the evidence suggests, will come at the expense of investment elsewhere in the economy. And because data centers employ so few people, a back-of-the-envelope estimate yields a number that made even my eyes bug out: the capital invested in data centers will employ roughly ten million fewer workers than that capital would have employed if invested elsewhere.

This doesn’t mean that unemployment will rise by 10 million. It means, instead, that reduced demand for labor will put downward pressure on wages and shift the balance of economic power, which has already been moving against labor, even more in capital’s favor. Even before the rise of AI, we were experiencing a rapid decline in the share of national income going to wages relative to profits, along with rising oligarchy. The anti-labor effects of AI will accelerate that trend.

Back in June, while acknowledging that the rise of AI is a form of “capital-biased” technological change, I expressed some skepticism about arguments that AI will have a large negative effect on labor. Since then, however, the sheer scale of capital expenditure on AI has become more obvious, along with the incredibly small number of jobs that spending will create. I’ve also revised how I think about interpreting the data. As a result, today’s primer offers a much more pessimistic take.

Beyond the paywall I will address the following:

1. Capital and labor in AI versus the rest of the economy

2. AI and crowding out

3. A blast from the past: 19th-century parallels

4. The scale of the issue