Is AI really fueling inflation? It’s not that simple
The main economic impact of artificial intelligence today is on interest rates.
This article originally appeared on Morningstar’s US website. We believe the insights are valuable for our Australian audience.
It might seem obvious that the stupendous sums of money being spent on artificial intelligence must be fueling faster economic growth, and that a hotter economy is pushing up inflation.
This kind of thinking is flawed, however, because it fails to consider the economy as an integrated system. In particular, we have to reckon with the Federal Reserve. Once we do, we see that the main effect of AI today is lifting up interest rates.
We believe this because the Fed is offsetting the demand shock from AI, leading to higher interest rates but keeping gross domestic product in line with potential. If the latter is true, then AI isn’t likely having a large impact on inflation.
That means that if the AI boom loses steam, it will likely lead to lower interest rates in coming years, as we forecast in our latest US economic outlook.
Here’s why I think this.
Demand shocks (like AI) are offset by the Fed
The key argument is this: AI is (currently) a demand shock, and the Fed generally offsets demand shocks.
A demand shock is just some increase in desired spending in some sector of the economy.
It’s fairly uncontested that AI is currently providing a large demand shock. Not only are vast amounts of money being spent on data centers, but the boost to stock prices is supporting consumer spending.
Let’s break out a diagram. The real (inflation-adjusted) interest rate and the level of GDP are jointly determined by the intersection of two lines. First, the “investment-saving,” or IS curve, which shows GDP as a negative function of interest rates. The key intuition is that lower interest rates encourage more spending by the private sector.
Second, the “monetary policy” curve illustrates the Fed’s goal of pushing GDP in line with its potential. Potential GDP is the goldilocks zone for the economy: not too hot, not too cold. In fact, we could define potential GDP as full employment without inflationary pressure, which is exactly both prongs of the Fed’s dual mandate.
So, the Fed seeks to set interest rates at whatever level is required to set GDP equal to its potential. Hence the vertical line.
Demand shocks cause the IS curve to shift. Given our framework, we’re saying that the Fed seeks (and generally is able) to offset this demand shock by the amount needed to keep GDP in line with potential. So, the IS curve shifts along a vertical line. That means higher interest rates, while GDP (and, for the most part, inflation) is left approximately unchanged.
In a counterfactual of no AI boom, interest rates would have receded much closer to the lower levels before the pandemic, in our view. The 10-year Treasury averaged 2.5% over 2017-19; it currently stands at 5.0%.
Inflation is mainly a product of GDP above potential
Skipping over some nuance, our basic framework is that inflation is mainly a product of GDP being above its potential. Now there could be many reasons why GDP increases above its potential: fiscal stimulus, asset bubbles, or excess money supply growth. The common thread is that you have too much spending chasing too few goods.
But if the Fed is keeping GDP about in line with potential by offsetting AI’s demand shock, then AI isn’t driving much inflation.
We say “much” because AI is driving some industry-specific price increases, such as the soaring price of memory chips. But these are a small component of consumer spending compared with, say, oil. Also, it’s not a given that industry-specific price increases are driving higher overall inflation, unless accommodated by the Fed. (This is a tricky topic to explain, so I defer to this article by the great John Cochrane if you’d like to read more details.)
The key point is that sustained high inflation requires GDP running well ahead of potential, producing an economy that is overheated across the board. But that’s not really the case currently. As we discuss in our latest economic outlook, broader inflationary pressures are rather subdued; if not for the tariffs and US-Iran war, inflation would be pretty close to the Fed’s 2% target currently.
What happens if AI becomes a supply shock?
A corollary to the conventional narrative is that AI may be causing inflation today because it’s a demand shock, but it will shift to being deflationary once AI becomes a supply shock.
When we talk about AI becoming a supply shock, we mean specifically that AI will ultimately boost potential GDP. The component of potential GDP that AI would affect is productivity: output per worker.
However, we’ve argued that AI isn’t yet boosting productivity in a major way. (We’ll have more to say about this in the future, along with the prospects for this changing.) For now, we’d say that AI is just a demand shock, not a supply shock.
But even if AI starts boosting productivity and thereby becomes a supply shock, we disagree with the idea that this is inherently deflationary. The result would be faster potential GDP growth, and as long as the Fed is able to stimulate the economy sufficiently to accelerate GDP in line with this faster potential, there’s no inflationary impact.
AI is crowding out other spending
Ok, enough of theory. What does the data say?
If the Fed was offsetting the demand shock from AI by keeping interest rates higher than they would be otherwise, then we’d expect to see those high interest rates weigh on non-AI types of spending. And this is exactly what we see.
All of the growth in US private fixed investment in 2025 and 2026 has been coming from tech-related categories, driven by AI. All other private fixed investment, such as housing, commercial real estate, and other categories, is contracting.
Some readers may recall the notion from introductory economics of high government spending “crowding out” private investment. This is similar, except we have one type of private spending, AI, that’s crowding out all others.
If the AI boom fades, then interest rates will fall
So, what’s the key takeaway?
If AI is playing a major role today in lifting interest rates, then any cooling of the AI boom should act to lower interest rates. Interest rates are still too high for much of the economy, but the AI boom has concealed this.
There are a lot of forces acting on the economy right now; the oil price shock from the Iran war looms especially large. But once the dust settles on that, the AI boom will have a major role to play in the future course of interest rates.
In our latest US economic outlook, we expect interest rates to eventually fall, with the 10-year Treasury yield dropping to 3.5% by 2029, versus 5.0% currently. Slowing AI spending growth is a major reason for our forecast.
If the boom inflates into a bubble, however, that could fuel higher interest rates in the near term.
