Three Bull Cases vs Three Bear Cases for Intel — 95% of Foundry Revenue Was Intel Billing Itself

The debate around Intel right now isn't whether the business got better. The earnings settled that. The real debate is whether "better" justifies a stock that's up more than 250% this year.

So I built out three bull cases and three bear cases, as fairly as I could, and put them side by side. Here's the pattern that emerged: the bull cases rest on things that already showed up in the results, and the bear cases rest on bills that haven't arrived yet. That asymmetry explains most of what makes this stock hard.

Bull case 1: Intel can ride AI without ever beating Nvidia

Intel's AI exposure doesn't come from winning the accelerator war. It comes from selling the CPU that every AI server needs regardless of who wins that war.

Nvidia takes all the headlines. But every AI server also needs a general-purpose brain to run everything, and that's Intel's home turf. As AI infrastructure expands, CPU demand expands with it.

This quarter, data center and AI revenue grew 59% and profit in that segment nearly quadrupled. Intel went as far as saying it couldn't make these chips fast enough to meet demand. That's a good problem to have.

There's a second-order effect worth noting. As AI shifts toward agents that carry out tasks on their own, they constantly call up data and programs — and coordinating all of that leans harder on CPUs. The smarter AI gets, the more CPU load it generates. Intel can quietly ride the whole wave without ever going head to head with Nvidia.

Bull case 2: the profits actually came back

A company that was losing money on operations a year ago just posted a 17% company-wide operating margin. Of the three bull cases, I think this is the strongest.

The products business ran at a 32% margin and the data center piece hit 40%. And the swing was produced by cutting costs while revenue grew. That's the classic recipe for profit explosions. Grow revenue alone and margins stay flat. Cut costs alone and growth dies. You only get numbers like these when both move together.

This is the first quarter where the turnaround shows up in actual dollars instead of press releases.

Bull case 3: Intel still owns the PC world

Intel sold fewer chips this quarter and still grew PC revenue, because it raised prices by 27%.

That came from shifting the mix toward pricier, more capable AI PC processors — which is evidence that pricing power is intact. A company whose competitive position has genuinely collapsed cannot raise prices 27%.

Intel is also woven into the procurement stack of essentially every computer maker and every large enterprise IT department on the planet. That kind of distribution, certification, and compatibility asset doesn't get displaced overnight because a rival shipped one good chip. And it throws off a lot of cash while it lasts.

Bear case 1: 95% of foundry revenue was Intel billing itself

The "we'll build chips for other companies" pillar of the Intel story barely exists yet. I consider this the most serious weakness.

On paper the foundry segment showed almost $6 billion in revenue. Only about $290 million of that came from actual outside customers — roughly 5% of the total. The rest was Intel's factories charging Intel's own divisions. Money moving from one pocket to the other.

And that segment still lost more than $2 billion in a single quarter.

Until a name like Apple or Nvidia actually signs, the foundry is a story, not a business. And the market is currently assigning real value to that story.

Bear case 2: the company can win while shareholders lose

Intel plans to spend over $20 billion this year building factories, and more next year. It already carries a fair amount of debt.

When asked directly, Intel's finance chief would not rule out selling more shares to raise money. A share issuance slices the pie into more pieces. The company can get bigger while your slice gets thinner.

This is the most nuanced trap in the whole thesis. Even if Intel successfully fixes its factories, the debt, capital spending, and new shares required to get there can eat the value created before it reaches you. The enterprise wins; the shareholder gets a disappointing return.

Turnaround investors miss this constantly. "The company recovered, so the stock should go up" and "how much of the recovery actually reaches shareholders" are completely different questions.

Bear case 3: Intel isn't in the most profitable seat in AI

The real gold mine in AI is the ultra-powerful accelerator chips, and Nvidia flat out owns that market.

What Intel gets is the less glamorous CPU side. That's genuinely good — but the danger is what happens when the market stops pricing Intel as a CPU supporting player and starts pricing it as an AI headliner.

The real Intel is fighting AMD on one side and Nvidia on the other. Winning a slice of AI is a good thing. Being priced as if you've already won the whole race is a dangerous one. I'd argue some of that danger is already embedded in today's price.

The two cases side by side

Issue	Bull case	Bear case
AI exposure	Data center/AI revenue +59%, segment profit ~4x, agents add further CPU demand	Nvidia owns the highest-margin accelerator market; Intel is a supporting player
Profitability	17% company operating margin, 32% product margin, 40% data center margin	$20B+ annual capex plus existing debt eats the improvement
Foundry	Structural re-rating possible if outside customers commit	Only $290M external revenue (~5%), losing $2B+ per quarter
PC	Revenue grew on a 27% ASP increase despite lower unit volume — pricing power intact	Mature market, not a structural growth engine
Shareholder take	Brand and customer network can't be replaced quickly	Equity raise not ruled out — dilution risk is live

Where I land

The three bull cases already happened. The three bear cases are still ahead. That makes Intel a stock that looks good now and can get expensive later.

The bear case I weigh most heavily is the second one. Even assuming the foundry eventually works and AI CPU demand holds, spending $20 billion a year, carrying debt, and printing new shares along the way leaves a much thinner slice for shareholders than the headline growth suggests. That's exactly where business success and investment success separate.

Once both cases are on the table, one question remains: does today's price make sense given all of it? I ran that math in Is Intel cheap at $84? What a 26-year round trip teaches.

FAQ

Q: Why does the $290 million in external foundry revenue matter so much? A: Because a large part of Intel's valuation story rests on the premise that it becomes a credible alternative to TSMC. External revenue at roughly 5% of segment revenue means that premise is still unproven. Internal transfers are accounting revenue, not new cash coming in the door.

Q: Is calling Intel an AI winner an exaggeration? A: Not an exaggeration, but a different category. Intel benefits through the CPUs every AI server requires, not through accelerators. The 59% data center and AI growth this quarter is the evidence. It's just not the highest-margin seat, so applying an Nvidia-style premium is a separate argument.

Q: How realistic is a share issuance? A: There's no confirmed plan, but the finance chief declined to rule it out, and with $20B+ in annual capex on top of existing debt it's hard to dismiss. Dilution cuts per-share value directly, so it belongs in any turnaround scenario you build.