For years, I kept hearing the same argument about AI.

It would make companies more productive.

It would reduce costs.

It would help workers get more done.

And eventually, that should help bring prices down.

It sounded pretty convincing.

But now I'm starting to wonder if we're looking at the problem from the wrong side.

Because before AI has had time to deliver all those productivity gains, the money being poured into AI infrastructure is already creating new costs.

And those costs are starting to show up in places you might not expect.

AI Needs a Lot of Stuff

Think about what AI actually needs.

Chips.

Servers.

Data centers.

Buildings.

Electricity.

Cooling systems.

And a lot of people to build all of it.

Companies are spending enormous amounts of money trying to build AI infrastructure as quickly as possible.

Goldman Sachs estimates AI investment could reach around $600 billion in 2026, equal to roughly 2% of U.S. GDP.

That's a huge amount of spending.

And when everyone wants the same things at the same time, prices can go up.

That's just basic supply and demand.

Then There Is Electricity

This is the part I find especially interesting.

AI data centers need an enormous amount of electricity.

And the U.S. is already expecting electricity consumption to hit new records in 2026 and 2027, with AI and crypto data centers among the important drivers.

Now imagine you're living somewhere close to a huge data center.

The data center needs more electricity.

The power grid needs more capacity.

New infrastructure has to be built.

Someone has to pay for it.

And eventually, some of those costs can find their way into electricity bills.

The Dallas Fed has estimated that data-center demand could add to inflation through higher retail electricity prices, although the exact impact depends heavily on how quickly the buildout happens.

So the technology that's supposed to make everything cheaper could actually make some things more expensive first.

That's the strange part.

But AI Could Still Reduce Inflation

I don't think this means AI is automatically inflationary.

That's too simple.

And this is where the story gets interesting.

AI could eventually make businesses much more productive.

A company might be able to do the work of ten people with five.

Software development could become cheaper.

Customer service could become faster.

Manufacturing could become more efficient.

If those productivity gains become large enough, companies could eventually lower prices.

The Richmond Fed has made a similar point: AI and data-center investment appear to be adding some inflation pressure now, but if the investment produces enough productivity gains, the effect could eventually turn disinflationary.

So we're basically dealing with two different stories.

AI could push prices up today and push them down tomorrow.

The problem is that we don't know exactly when the second part will arrive.

And That's a Problem for the Fed

This is where I think things get really interesting.

The Federal Reserve doesn't get to set interest rates based on what AI might do five years from now.

It has to look at what is happening right now.

And right now, AI investment is creating a huge amount of demand.

That means the Fed has another thing to watch while deciding what to do with interest rates.

Reuters recently noted that AI is already influencing investment and markets, although its direct effect on broader U.S. inflation remains relatively small and difficult to isolate.

So I'm not saying AI is suddenly responsible for inflation.

It isn't.

But I think it's becoming harder to ignore.

The Irony Is Hard to Miss

This is the part that made me stop and think.

We started the AI boom believing that better technology would eventually make the economy more efficient.

And maybe it will.

But to get there, we're spending hundreds of billions of dollars building the machines, data centers and power infrastructure needed to make that future possible.

That spending creates demand.

Demand creates pressure.

And pressure can create inflation.

So we might have a strange situation where AI eventually becomes deflationary, but the road to getting there is inflationary.

Honestly, I don't know how this ends.

Maybe productivity wins and AI eventually helps bring costs down.

Maybe the infrastructure boom keeps pushing prices higher for longer than people expect.

Maybe we get a little bit of both.

But I'm watching this closely.

Because if AI keeps increasing demand for chips, electricity, construction and other scarce resources, the Federal Reserve may have to deal with a problem nobody expected from the technology that was supposed to make everything cheaper.

AI might still fight inflation.

It just might have to make inflation worse first.