Intel's Supply-Constrained Turnaround

Programming note: The next issue lands 30 January, where we'll take a look at the state of the AI economy.

Hand-to-Mouth at the Foundry

The central wager at Intel is that the world really needs a high-end alternative to TSMC, and that being a U.S.-based company is enough of a competitive advantage to justify losing billions of dollars in the short term. This foundry strategy was the signature initiative of previous CEO Pat Gelsinger, and it has since been handed over to Lip-Bu Tan, who is tasked with turning those industrial ambitions into a functional business.

Intel's fourth-quarter results, released yesterday, show that while the strategy is clear, the execution is currently running into the stubborn reality of manufacturing lead times. The company swung to a net loss of $600 million and provided a first-quarter forecast that missed even the modest expectations of analysts.

The awkward reality of owning your own factory is that you are solely responsible when you guess the market wrong. Intel was caught off guard by the sudden surge in demand for the server central processors that run alongside AI chips. Despite running its plants at capacity, Intel is leaving profitable sales on the table because its machines were busy making the wrong things. As CFO David Zinsner put it during the earnings call, it is currently a "hand to mouth" operation:

[Both DCAI and Client segments] will be down as a function of supply. Obviously, we're shifting as much as we can over the data center to meet the high demand. But we can't completely vacate the client market. So we're trying to support both as best we can and obviously work our way out of this supply issue...it's just literally hand to mouth, what we can get out of the fab and what we can get to customers is how we're managing it.

...we do have our own fab so we can squeeze out supply as much as possible, which is what we're working on, but we directionally weren't managing the supply to an expectation that there would be unit increased that significantly in data center.

The 14A Standoff

The real tension, however, is probably in the "Foundry of the Future." Intel is actively pitching its next-generation 14A manufacturing process to external customers. This is the technology that is supposed to finally allow Intel to compete directly with TSMC.

Usually, to win a customer for a factory, you have to show them the factory. But under Tan, Intel has adopted a policy of strict capital discipline. Tan made it clear on the earnings call that they will not invest in actual factory capacity for the 14A process until they have a firm commitment from a customer:

My discipline is until they have a commitment to the volume, then starting to really build and the foundry expansion so that we can meet their requirement... engagements with potential external customers on Intel 14A are active. We believe customers will begin to make firm supplier decisions starting in the second half of this year.

This creates a game of chicken:

• The Customer's View: Why would I bet my multi-billion dollar product roadmap on a factory that doesn't exist yet, from a company whose current yields are "still below" where the CEO wants them to be?
• Intel's View: Why would I spend $20 billion on a factory for a customer who might go back to TSMC if their own roadmap changes?

Intel is trying to de-risk a business that is fundamentally defined by massive, upfront risk. The upshot is a strategy that treats a cutting-edge semiconductor fab more like a build-to-suit real estate project, provided you can find a tenant willing to pre-lease a building that exists only on paper

The Endorsement Premium

There is now a widening gap between the financial narrative surrounding Intel and the operational reality in its fabs. In 2025, Intel's stock gained 84%, far outperforming the broader semiconductor index—not because Intel suddenly made more money, but because it accumulated a kind of endorsement premium. Nvidia invested $5 billion, SoftBank added $2 billion, and the government arrived with CHIPS Act support.

Put differently: the endorsements are doing the heavy lifting. They've pushed investors to price Intel's future as if it's increasingly certain, even while the present is still underwater. The result is a valuation that's hard to justify using traditional math: Intel trades at 126x expected earnings, versus 45x for AMD and 26x for Nvidia. Normally, if your NTM P/E is trading at a multiple five times higher than the industry leader, the market expects you to be the one doing the disrupting. In Intel's case, the multiple is high primarily because the earnings have almost entirely eroded while the stock price has stayed aloft on the strength of its backers.[1]

But endorsements don't change the physics of fabs. Intel is still supply-constrained in the products it needs and overstocked in the ones it doesn't. It has $11.6 billion in inventory, and it's the wrong inventory at the wrong time: too many PC chips and not enough of the server chips the AI boom actually wants.

Fixing that mismatch means re-tooling factories, which is why Intel is reporting losses today in hopes of profits in 2027 or 2028. The endorsement premium buys time. The question is whether 14A can convert that time into something more durable: customers willing to sign, and capacity worth building.

[1] A more grounded NTM Price-to-Sales ratio makes Intel look cheap (around 5.4x) compared to its peers. But that also highlights the execution risks in converting revenue to profits.

More on Semiconductor:

• A timeline of the US semiconductor market in 2025 (TechCrunch)
• AI Boom Reshapes Global Semiconductor Supply Chains (WSJ)

On Our Radar

Our Intelligence Desk connects the dots across functions—from GTM to Operations—and delivers intelligence tailored for specific roles. Learn more about our bespoke streams.

Jensen Rejects AI Bubble Narrative

• The Story: At Davos, Nvidia CEO Jensen Huang pushed back against AI bubble fears, arguing that current high investment levels are necessary to build foundational infrastructure and that the economic benefits are now spreading to various industries, justifying further spending. (Dow Jones)
• The Investment Implication: Huang is actively managing market expectations by framing the immense AI capital expenditure not as a speculative bubble, but as a rational, multi-layer infrastructure build-out with a clear path to ROI at the application level. For investors, this narrative is a direct attempt to justify sustained high valuations and defend against cyclical concerns by positioning the current spending as a long-term, foundational investment necessary to unlock future economic growth across all industries.

Nvidia Fortifies Inference Ecosystem

• The Story: Nvidia invested $150 million in Baseten, a startup that helps companies deploy AI models, signaling a strategic move to strengthen its position in the growing AI inference market where it faces more competition. (Barron's)
• The Investment Implication: This investment in an MLOps platform is a strategic move to defend Nvidia's market share as the AI workload shifts from training to the more competitive inference market. By investing in the software layer that manages model deployment, Nvidia aims to create ecosystem lock-in, ensuring that its hardware remains the preferred choice for running inference workloads, thereby protecting its future revenue streams against a growing number of hardware competitors.

P.S. Tracking these kinds of complex, cross-functional signals is what we do. If you have a specific intelligence challenge that goes beyond the headlines, get in touch to design your custom intelligence.

You received this message because you are subscribed to ARPU newsletter. If a friend forwarded you this message, sign up here to get it in your inbox.