Central banks are good at fighting inflation they can recognize.
The trouble starts when the price pressure arrives from somewhere the playbook never anticipated.
For most of the past four years, the Federal Reserve has sorted rising prices into two bins.
Demand inflation happens when people have too much money chasing too few things, and the fix is higher interest rates.
Supply inflation happens when something breaks in the physical world, a war, a shipping lane, a harvest, and the fix is mostly patience.
Oil is the textbook case of the second kind.
When a barrel gets more expensive because tankers cannot move safely, raising the cost of borrowing in Ohio does nothing to reopen a strait in the Persian Gulf.
Fed officials understand that, which is why they spent most of this year insisting the energy shock would pass through and fade.
Then something turned up in the June and July commentary that does not fit either bin.
Policymakers have started naming artificial intelligence infrastructure as its own source of price pressure, which quietly converts Nvidia (NVDA) from a stock story into a monetary policy variable.
ANGELA WEISS / Getty Images
How the Fed usually reads an oil shock
The energy math this month has been brutal.
Brent crude settled at $100.69 a barrel on July 23, its first close above that level since May 26, while West Texas Intermediate finished at $92.19, according to CNBC.
Crude has climbed roughly 40% this month as disruption spread from the Strait of Hormuz to the Red Sea, according to Trading Economics.
More Fed:
- Fed’s Waller issues stark warning on inflation, interest rates
- Fed’s Warsh faces tough interest-rate smackdown in Congress
- Top economist delivers blunt Fed rate warning for 2026
The Iran war that began in late February has kept a permanent risk premium in the barrel price, and every attempt at a truce this year has collapsed within weeks.
The classical response to that is to wait. Energy shocks are supposed to be one-time level shifts that drop out of the annual comparison twelve months later.
Waiting only works if the shock stays contained, though.
Headline inflation is now running at 3.7%, well above the central bank’s 2% target, according to Forbes.
That is the fifth consecutive year prices have overshot, and patience starts to look like surrender when a target has been missed for that long.
Markets noticed. Odds of a rate increase at this week’s meeting climbed from 10.7% on July 15 to 34.7% by July 22, according to The Motley Fool, with live pricing tracked on CME Group’s futures-based tool.
Why AI data centers now move inflation
Here is the part almost nobody priced in.
Cleveland Fed President Beth Hammack wrote in a July post that business leaders cite “pressures from insurance and the AI data center build up,” according to CNBC.
Read that again, because it is a genuine break from precedent. A regional Fed president put data center construction in the same sentence as energy costs and supply chains as a driver of business input prices.
Related: Nvidia stock is doing something it hasn’t done in years
Governor Christopher Waller made a similar point in a July 13 speech warning that inflation is up this year, adding that tightening may need consideration.
The mechanism is not mysterious once you look at it. Building AI capacity consumes electricity, transformers, turbines, copper, concrete, skilled electricians and land, and it consumes them in the same regional markets where households and ordinary businesses buy the same things.
When I lined up the Fed commentary against the price data, the pattern that struck me was the ordering. Officials mentioned energy first for four straight months, then started listing the buildout alongside it.
Here is what the committee is actually weighing:
- Brent settled above $100 on July 23 for the first time since May 26, according to CNBC.
- Crude is up about 40% on the month as disruption widened past Hormuz, according to Trading Economics.
- Headline inflation sits at 3.7% against a 2% target, according to Forbes.
- Nearly half of policymakers signaled support for a hike later this year, according to CBS News.
What a rate hike would cost Nvidia shareholders
The awkward part is what this does to the AI trade itself.
Nvidia is worth roughly $5 trillion and remains the most valuable company in the world, but the stock has gained only about 9% in 2026 while Apple surged more than 20%, according to CNBC.
The chipmaker also sits well below its all-time high near $5.73 trillion, which means the AI leader has spent this year going sideways while the story around it got louder.
Now the loop closes. Capital spending on AI hardware helps push input prices higher, higher prices raise the odds of tighter policy, and tighter policy compresses the multiple investors are willing to pay for long-duration growth stocks.
Nvidia is, in effect, helping to build the case for the rate environment that would hurt Nvidia most.
For a retail investor this is less abstract than it sounds. If you own an S&P 500index fund, a meaningful slice of your money already sits in a handful of AI infrastructure names, and their valuations are built on the assumption that the discount rate falls from here.
A hike does not just trim those multiples. It raises the cost of financing the data centers that generate the revenue underneath them.
Chief Executive Jensen Huang has said computing costs are heading from roughly $50 billion toward $100 billion per gigawatt, according to CNBC.
That figure is a boast about demand. It is also, read from a central banker’s chair, a forecast of sustained pressure on industrial inputs.
My read is that this is the most underappreciated risk in the AI complex right now, and it has nothing to do with chip competition or export rules. It is that the buildout has grown large enough to influence the discount rate applied to it.
What to watch after the July rate decision
The decision lands July 29 at 2 p.m. ET, followed by Chair Kevin Warsh‘s press conference thirty minutes later.
Most economists still expect no change, and there is no dot plot at this meeting, so the statement and the press conference carry the entire signal.
Warsh has stripped out forward guidance since taking over in May, which leaves markets reading tone rather than projections.
Listen for whether he separates energy from the buildout when he describes price pressures.
If he treats both as temporary, the AI trade gets room to run into Nvidia’s Aug. 26 earnings report.
If he groups artificial intelligence spending with the structural pressures the Fed intends to lean against, the sector’s cost of capital just changed, and no chip cycle fixes that.
For a portfolio, the practical takeaway is that AI exposure and rate exposure stopped being separate bets somewhere this summer.
Most investors are still holding them as though they are.
Related: Fed’s Warsh drops fresh clues on interest-rate path