You wouldn’t necessarily know it to look at the headline hiring numbers — although payroll growth did slow dramatically last year — but we’ve just witnessed the first annual drop in overall, US big-cap employment in a decade.
I mention this Friday for obvious reasons: Tech layoffs are in the news after back-to-back announcements from Meta and Microsoft who, between them, are cutting jobs and implementing buyouts with the potential to impact nearly 25,000 people.
Before I go any further — and at the risk of coming across as insensitive — a lot of those soon-to-be ex-Meta/ex-Microsofters will have no trouble finding new jobs. And chances are, some of them are well-off enough by now they might not have to, at least initially.
Still, the juxtaposition between staff cuts impacting between 7% and 10% of the workforce and massive outlays for technology with the potential to replace human workers is jarring. Insult to injury: The job cuts are an effort to help offset the cost of building that technology. Insult atop insult: Some of the impacted employees might’ve engineered their own jobs away.
As the figure above, from Goldman, shows, hyper-scaler capex growth is expected to peak in quarterly results due next week at more than 90%.
Although spending growth should decelerate going forward, overall spending on the AI buildout’s likely to keep rising for the foreseeable future — or anyway as long as capital markets will bear some of the cost. That’ll mean ongoing pressure to preserve margins. One way of doing that is cutting headcount.
The figure below, from the latest installment of BofA’s popular weekly “Flow Show” series, illustrates the statistic mentioned here at the outset:Â S&P 500 employment was down 400,000 last year, the first drop since 2016.
Workers are grappling with “chunky job cut announcements,” the bank’s Michael Hartnett said, adding that global youth unemployment rates remain elevated. An “insecure worker,” he went on, makes for a “grumpy consumer.”
Still, and as discussed here earlier this week, there’s scant evidence to suggest US payrolls are on the brink of a major contraction. If anything, the incoming data suggests hiring might be picking back up.
Hartnett said as much and noted that an acceleration in forward EPS estimates could presage “surpris[ing] US payrolls strength” which would be “very positive for consumer cyclical stocks, especially if the oil and gasoline price surge unwinds in May.”
Of course, that doesn’t “sell papers,” so to speak. CNBC’s lead story on Friday afternoon carried this headline: “20,000 job cuts at Meta, Microsoft raise concern that AI-driven labor crisis is here.”
I think some of the story is AI, but I think the main story is that in 2021 everyone thought the world was ending, and tech jobs were everywhere. Meta was banging down my door. Now that the labor market has turned, employers can shed all of these overpaid hires from 2021, and slowly add where needed.
Further, I am also seeing an acceleration in software engineering jobs because everyone realized they need us to babysit AI. Again, less of us for sure. But I think now there are many more companies being started are are able to be “AI lean” so to speak.
I think you are correct. I see ai as doing a very small part yet of what it can accomplish.
I use it every day in a project I am publishing open source. It is improving and the newest models supposedlly are much better than the models of 6 months ago. In any event I do not see ai moving on it’s own even with the advent of agents any time soon. The productivity gains are still sustained but at some point will be plateauing.
Saw a you tube featureing one CEO for one of those small AI companies you referenced. His human resources cost is twice the ai token cost, even with them trying to maximize ai. The ai is accoblishing, by his estimate, 5-6 X the productivity of humans and therefore once ai cost increases to 5X his HR costs he will start to consider his business has plateaued. He did say that AI has changed the equation that execution used to be hard and now it is easy/cheap. WHile ideas are the real limit to data product development.
What seems obvious is the physical world can only undergo so much optimization. Ultimately where the rubber meets the road is the physical world: steel, concrete, travel, food, clothing etc. It give me hope that we might see ultimately a productivity gain. I do not think software engineers who learn ai will find this change to be detremental to their career. I think all engineers need to be conversant, skilled in ai. I do believe AI is creating an intellectual divide. Those who do vs. those who are learning from it and about it. I am waiting for AI to be used for execution of engineering projects: piping, civil, structural drawings and calculations. I believe that will allow us to make very real bankable estimates which have historically been bullshit estimates at best.