Micron Technology, Inc.

MU

Micron Technology, Inc.

@david
• 1 week ago

40 towns said no to data centers. Wall Street panicked about the wrong year.

On Tuesday, data centers were on the ballot in 41 places across 10 states. Voters sided against them in every one but Harwood, North Dakota. Ohio alone had 21 of those votes.

Thursday morning, the market did the math. More than $2 trillion came off US stocks in the first 50 minutes of trading. The chip index fell around 10%, the Nasdaq confirmed a correction, and Nvidia gave back its entire year.

I get it. If you can’t build data centers, you don’t need as many chips to fill them. But I think the market got two things wrong: who voted no, and when it actually matters.


Who voted no

The story on TV today is “Democrats swept, Democrats hate data centers.” Look at the ballots, though. These were mostly small towns, a lot of them in Ohio. And before the election, SemiAnalysis, a research firm that tracks basically every big data center project in the country, looked at the 20 candidates for governor in the big data center states. Nine wanted to pause construction. Those nine split roughly evenly between the parties. In Ohio and Oregon, both candidates wanted a pause.

I don't think it's necessarily a political thing by nature. It’s a power bill issue.

The Energy Information Administration, the government’s energy scorekeeper, expected the average home to pay about 5% more for electricity this year, after a similar jump last year. When a building the size of a shopping mall goes up down the road and your power bill goes up with it, you vote no.


When it actually matters

Here’s what the selloff skipped: data centers take years to build. SemiAnalysis says most of what comes online in 2027 is already under construction. The real fight is 2028, about 62 gigawatts of planned US data centers, most not started yet. (A gigawatt is roughly what one big nuclear reactor puts out)

About 30 gigawatts of that sat in states where a pause candidate was running.

That’s exposure, not cancellations. A campaign promise has to become a law, and the law has to catch a project at the wrong stage.

Meanwhile, the four biggest buyers, Amazon, Microsoft, Alphabet and Meta, are guiding to roughly $730 billion of capital spending this year, meaning money spent on buildings, chips and equipment. That’s my math from each company’s latest guidance.

So is today a fire sale or a fire alarm? Honestly, it can be both. It depends on what the business was worth before the markdown and what it’s worth after. A great company at a silly price can fall 15% and still be at a silly price.


What I’m looking at next: Micron

I scanned the day’s biggest losers, and one stopped me. Micron makes memory chips, the stuff AI chips need to hold data while they work.

Three years ago, in fiscal 2023, Micron did $15.5 billion of revenue and lost $5.8 billion. In fiscal 2026, it did about $133 billion of revenue and roughly $85 billion of profit. In one quarter this year it kept $28.2 billion of profit on $41.5 billion of sales. That’s 68 cents of every dollar.

For comparison, AMD, the AI chip company everyone talks about, kept $2.4 billion on $13.4 billion last quarter. About 18 cents. Right now a memory maker is nearly four times more profitable per dollar than a GPU designer.

But this is not without risks!

Memory is close to a commodity, and margins like that attract competition. China’s CXMT just posted a sevenfold jump in revenue. Peter Lynch warned that cyclical stocks look cheapest right when earnings are about to fall.

Before today’s drop, Micron traded around 12 times last year’s earnings. Cheap, or a trap?

Here’s the tie back to Tuesday. Micron’s CEO said it can only fill half to two-thirds of demand for its AI memory, and its 2026 supply is sold out on multi-year contracts. That carries it through the 2027 build. The question is 2028, the same year the ballots matter.

So my homework is one Buffett question: what does Micron earn in a normal year? Not 2023’s loss, not 2026’s jackpot. Somewhere in between. I’m doing the full research before its Dec 16 earnings. Come do it with me.