Bro, do you even fab?

Lip-Bu Tan told Cisco's Jeetu Patel this weekend that CPU demand is so high Intel can only supply half of it, as CTech reported. The Fool's Mike Schwenk put it plainly in his risk section: demand is outrunning Intel's manufacturing capacity, and that caps how fast it can grow.

Well, yeah. They can't staff the fabs!

But wait. This price is based on Intel becoming America's onshore fab. Indeed, Seeking Alpha says the future has never looked this bright. Barchart ran the rivalry with TSMC on Monday.

Deep dives following the money.

Well, as my readers know, before you follow the money, you should follow the people.

Could this fab BE any emptier?

Ten seasons of Friends and nobody could say what Chandler did for a living. Investors have the same problem with Fab 52.

Fab 52 in Chandler is the high-volume home of 18A. D1X in Hillsboro is where the process gets developed. If Intel is turning into the onshore foundry, those two zip codes hire first. Here is what they did instead.

The bleeding slowed to single digits a month this year. It never turned. Neither metro has had a month of net growth since August 2024, and our six-month base case has both flat to lower through March. TSMC built its Arizona team and stopped in February. Intel kept shrinking.

Engineers, not VPs

Intel's version of the cuts is overhead. About 450 vice presidents down to 200, twelve management layers down to six. Fine. So we pulled engineering and IT on their own to see if the line got protected.

It did not. Engineers left at the same rate as everyone else. Phoenix engineering last had a month of net growth in July 2024, and the model has it flat at 3,894 ±101 in March.

Everyone else is hiring

A sold-out manufacturer hires. Ask anyone else in the sector.

Or look at the data. Percent change in headcount, January 2024 to September 2026:

Intel is the only name on that chart with a shrinking workforce and a stock that quadrupled. Broadcom and TI are shrinking too. Nobody is calling them the next TSMC.

But 18A output is up

It is. Barchart pulled it from the second-quarter call: 18A output 25% above target and up more than 50% from the prior quarter, on flat people. That is a fab that was built and staffed before it had customers, filling up. Loading tools you already own takes no new people. It's a brilliant strategy right up until Fab 52 is full.

Then what? Ohio was supposed to start in 2025 and is now 2030 or later. 2026 capex is flat. Tan himself named substrates as a bottleneck, with Japanese and Taiwanese suppliers locking up supply through prepayments, and no amount of Arizona hiring fixes that.

The real labor savings come with 14A. That node goes to risk production in 2027 and volume in 2028, and nobody has signed up for the volume yet.

The money says the same thing. Intel Foundry booked $5.8 billion in Q2. $293 million of it came from outside customers. The segment lost $2.1 billion. TSMC holds 72.5% of the world's foundry revenue. Seventy-one times forward earnings is what you pay to bet that gap closes.

So how does Intel get there from here?

Tan is right that demand is outrunning capacity. Capacity is tools plus the people who run them. The tools are spoken for until Fab 52 fills, the next fab is 2030, and the people have been walking out of Chandler for two years.

The stock is up 4x on a fab that cannot hold onto its engineers.

We will run this again in December.

Every series is lag-adjusted for late profile updates; the ± is the nowcast band. Counts are people in the workforce.ai panel, not a census.