Earlier this week, while editorializing around minutes from 2026’s first FOMC meeting, I said we’re at risk of overlooking the circular nature of the disinflation argument as it relates to a prospective AI-enabled productivity boom.
According to the account of last month’s policy gathering, some Fed officials heard from their regional business contacts that more companies are “automating operations,” an effort that’s showing promise in terms of offsetting cost pressures, thereby “reduc[ing] the need to pass those increases on to consumer prices.”
Ostensibly, that’s good news: It’s evidence that AI is in fact starting to deliver on the disinflation promise ascribed to the technology by its proponents. But it doesn’t take a leap of logic — no Olympic feat of second-order thinking — to see how it might be self-defeating.
If “automating operations” means replacing humans with robots (e.g., if agentic AI replaces too many accountants, coders and so on), those lost paychecks will make the disinflationary impact of AI a moot point.
Sure, it’d be nice if services prices stopped rising (or even started to fall) because many tasks are automated by technology which doesn’t ask for a paycheck, let alone a raise. But if people aren’t gainfully employed, they can’t spend into the economy even if goods and services are eminently affordable. Simply put: Nothing’s affordable if you don’t have a paycheck.
It’s with that in mind that we should warily eye charts like the one shown below, which plots the YoY change in real US GDP with the same for payrolls.
The red annotation shows that although growth’s going strong, jobs growth has flatlined. Indeed, the two series are diverging in real time.
In a feature piece published Wednesday, Bloomberg called that a “jobless boom” in the course of drawing some historical parallels. The linked article notes, correctly, that last year’s expansion came courtesy of infrastructure investment tied to the AI frenzy and “the resilience in consumer spending.”
That consumer spending was concentrated in the upper-half of the “K-shaped” economy, which is to say the “haves” spent and the “have-nots” didn’t. Spending among the “haves” was in part a function of elevated stock prices, which owed their buoyancy to the same AI boom.
Recall the chart above, which shows you the disparity in consumer sentiment between those with large stock portfolios and those with no stocks. It actually narrowed in 2025 before widening back out more recently, but you get the idea: If your on-paper net worth’s growing, you’ll be more comfortable spending than someone for whom the term “on-paper net worth” isn’t relevant for lack of financial assets.
This all seems a bit… well, artificial to me. America’s creating an economy that depends on hundreds of billions in data center spending and ever higher prices for stocks associated with the technology those data centers are built to power. Those data centers are pushing up electricity prices for everyday people, who don’t generally own a lot of stock in the companies developing the technology, and as that technology advances, the threat of an employment apocalypse for the same everyday people grows.
Long story short, America’s on the way to building a techno Potemkin village where AI does to white collar jobs what globalization did to the country’s manufacturing base.
Remember: Prices for some goods actually exhibited outright deflation in globalization’s wake. The same could happen for services prices if AI delivers on its promise.
But cheaper goods didn’t translate into a better life for displaced American factory workers. Rather, blue collar America succumbed to an epidemic of depression. The fact that 60-inch flatscreens could be had for $299 at Walmart was small comfort.
The same could happen to white collar workers in the years ahead. Maybe a lot of services will become vastly cheaper, but if the cost is jobs, it’s a Faustian bargain.
And what’s the end game here, by the way? More poignantly: If you outsource all the factory jobs and then you automate all the white collar work, what’s left?
Landscaping, construction and hotel room cleaning, I suppose. And given America’s current immigration policies, there should be plenty of open jobs in those industries going forward.
Who says the Trump administration doesn’t have a plan?
Or here comes UBI, Universal Basic Income, again? This time from the right and oligarchs?
Here’s the problem. UBI is paid by tax revenues, which won’t exist in a jobless economy.
“…what’s left? Landscaping, construction and hotel room cleaning, I suppose.”
And if you believe Elon, by 2030 those jobs will also be disappearing.
If you believe Elon Musk, soon there will be plenty of jobs as long as you don’t mind moving to Mars.
I made a similar argument to two of the LLMs many people are using and asked them to tell me how I was wrong. They essentially said the same thing: First, historically, new technologies create more jobs than they destroy (i.e., “creative destruction”). Second, even if there is a short-term spike in unemployment with adverse cascading impacts on consumer spending, debt servicing, overall demand, etc., the Federal Reserve would take actions necessary to save the economy. These LLMs do a great job regurgitating the Fed’s past actions w/o a ‘thought’ for moral hazard or the possibility it does not work next time.
The trick with AI’s in my experience, is to ask the right questions from the perspective of a firm but evolving context and thesis. They’ve helped me more than any human mentors and books and have dramatically helped augment my execution while updating my thesis. I use multiple models and have them red team me and each other. They agree as often as they challenge each other. My pnl has vastly improved. That said, they do get a lot wrong often. It’s my job to keep the thread. With their help I’m handily destroying the benchmark.
Congratulations
If you are using Gemini, you can employ the NotebookLM framework to constrain the reasoning to only sources you supply which will short circuit the models tendency to ‘consensus’ assumptions in it’s training data. https://notebooklm.google/
“Augment my execution”–of course your are (using) AI.
I think you may have inadvertently hit the nail on the head. LLM’s do a great job of regurgitating any past actions without a thought of moral hazard or possibility it will not work next time. LLM’s are great with a human in the loop. I think we will see many enterprises fail after people place too much stock in running on LLM.
LLMs are great at pontification but not recognizing where it has glossed over or has made a mistake.
In my view AI is a rote machine. Where you need rote, it excells where you need critical thinking it is berfit. That goes for correcting errors or generating original content.
With Powell gone the Fed won’t be able to save anything.
“technology which doesn’t ask for a paycheck, let alone a raise.”
Except it does ask for a raise..every quarter while stripling real people of money and resources. Tech and its billionaire owners dont negotiate a fair price and they dont share.
“technology which doesn’t ask for a paycheck, let alone a raise.”
Except it does ask for a raise..every quarter while stripling real people of money and resources. Tech and its billionaire owners dont negotiate a fair price and they dont share.
A reminder that for AI to start to pay its own way, it will have to produce tangible cost savings for PAYING subscribers. Not anecdotal reports from individual coders and John Q. Public but large-scale cost savings. Which means layoffs.
In our early days we had an agrarian economy and no industry, so we taxed the land. Then we went to an industrial economy and we began to tax those assets, the machines. Finally, we began see we had more laborers than anything else so we created income tax with the 16th Amendment to the Constitution in 1913. That tax and Treasury borrowing have paid our fiscal needs ever since. Now what? If Musk is right (isn’t he just so clever) AI is going to have to pay and that will be difficult since so far it creates no steady source of revenue. All jobs will be gone, no income for anyone (even Musk will lose all his dough, people with no jobs can’t buy his silly stock or the cars he makes.)