Washington Bet Billions on Intel's Comeback. Its Customers Haven't.

---

To value Intel today, you have to separate two things the rally has fused together: the strategic validation of
Intel and the commercial validation of Intel. They are not the same, and the gap between them is the entire
investment question. The strategic validation is overwhelming and real. The commercial validation, where it
counts — paying customers ordering chips from Intel's foundry at scale — is, by the company's own admission,
almost entirely absent. The market has priced the first as though it guarantees the second. It does not.

Start with the strategy, because Intel's plan is genuinely one of the most consequential corporate bets in the
world. For decades Intel both designed and manufactured its own chips, a model that worked until it didn't — the
company fell behind TSMC in manufacturing technology, lost its lead in the products that mattered, and watched
its most important customers and even its own designers turn to TSMC's superior factories. Intel's answer, under
the leadership of CEO Lip-Bu Tan and his predecessor, is to become a foundry — a contract manufacturer that
makes chips for other companies, the way TSMC does — and thereby to give the West a leading-edge chip factory
that does not sit ninety miles from China. It is a vision of immense strategic importance, and it is why the bet
attracted the backers it did.

This essay is about the chasm between that vision's backing and its results — about why a stock can triple on
the enthusiasm of governments and giants while the actual business those backers are betting on remains, in
commercial terms, unproven and deeply unprofitable, and about the specific reason a customer who buys your stock
for strategic reasons is not the same as a customer who buys your chips because they are the best.

The number that measures whether the foundry has customers

Here is the figure that the rally is built to make you overlook, and it is the most important number in the
entire Intel story. Intel Foundry — the contract-manufacturing business that the whole turnaround thesis depends
on — reported quarterly revenue of around $5.4 billion. That sounds substantial until you learn where it comes
from: the overwhelming majority is internal, Intel's foundry making chips for Intel's own product divisions,
one part of the company selling to another. The revenue from genuine external customers — other companies
paying Intel to manufacture their chips — is only about $174 million a quarter. And even that figure
flatters the reality: much of it traces to Altera, a programmable-chip business Intel itself used to own and
deconsolidated in 2025, which now counts as an "external" customer. Strip out the former subsidiary and the
revenue from true, won-on-the-merits third-party foundry customers is smaller still.

Sit with that contrast. The business being valued as the next TSMC, the foundry that justified a tripling of the
stock and a $36 billion government stake, sells roughly $174 million of manufacturing to outside customers in a
quarter — a rounding error against TSMC's tens of billions, and a tiny fraction of even Intel's own internal
foundry volume. And the company is candid about it where candor is legally required: Intel's own SEC
disclosures state, in language as blunt as corporate filings ever get, that the company has been "unsuccessful to
date" in attracting significant external customers to its foundry business. That is not a short-seller's
characterization; it is the company's own 10-Q, the place where a company must tell the truth about what it has
not yet achieved.

Now set that $174 million of external revenue against the foundry's losses: an operating loss of roughly $2.4
billion in the quarter, on an operating margin near negative 50%. The foundry is not a thriving business with
a few marquee customers; it is a deeply loss-making manufacturing operation that has, so far, almost no external
commercial base, burning billions while it tries to build one. The strategic case for Intel becoming a foundry is
airtight. The evidence that it has become one, measured in customers who pay it to make their chips, is barely
visible.

Strategic backing is not commercial demand

This is the heart of the matter, and it requires drawing a distinction the rally has erased: the difference
between the kinds of validation Intel has actually received. Consider who the backers are and why they backed
Intel, because the motivation reveals what the backing does and does not prove.

The U.S. government took its roughly 10% stake not because it ran a return-on-investment analysis and concluded
Intel's foundry was the best place to put taxpayer money. It took the stake because the United States has decided,
as a matter of national security, that it cannot depend on Taiwan — a self-governing island China claims and
threatens — for the most advanced chips on Earth, and Intel is the only American company with any chance of
building a domestic leading-edge foundry. Washington's investment is a strategic necessity, not a commercial
endorsement. The government needs Intel to succeed, which is a very different thing from the government
betting it will.

Nvidia's $5 billion investment follows a similar logic. Nvidia, whose chips are made by TSMC, has every reason to
want a credible second source of leading-edge manufacturing — both as a hedge against over-dependence on a single
Taiwanese supplier and as a friendly gesture in a world where Washington is watching semiconductor relationships
closely. A strategic investment that hedges your supply chain and pleases regulators is sound corporate policy.
It is not the same as Nvidia moving its actual chip production to Intel — and tellingly, the $5 billion deal
explicitly did not include Intel's foundry making chips for Nvidia, which continues to rely on TSMC. Nvidia
bought a stake in Intel; it did not become Intel's foundry customer.

And even the customer agreements — Microsoft's Maia accelerator on Intel's 18A node, Amazon's custom silicon,
the reported preliminary deal with Apple — share this character. They are, for now, pilots, derivatives, and
preliminary agreements: toes in the water, strategically valuable to both sides, hedges and signals more than
volume commitments. Intel's total signed lifetime commitments reportedly rose above $15 billion — a real and
encouraging figure — but "lifetime commitments" spread over many years and contingent on Intel's nodes actually
working is a promise of future revenue, not the $174 million of present revenue that is all the external base
currently produces. Every one of these backers has reason to want a healthy Intel foundry to exist. None of
them has yet voted for it with the only ballot that ultimately matters: large, sustained, present-tense orders.

The technology gap is real, and TSMC is not waiting

The commercial validation is thin in part because the product is not yet good enough to win on merit, and this is
the uncomfortable engineering reality beneath the strategic story. A foundry wins customers by offering the best
combination of performance, yield, and cost — and on the metric that most directly determines cost, manufacturing
yield, Intel still trails TSMC by a meaningful margin.

Intel's flagship 18A process — the node on which the entire near-term foundry thesis rests — has seen its yields
improve encouragingly, reportedly climbing from around 50% toward 55% and rising 7–8% per month, with internal
targets of 65–70% by year-end. That progress is genuine and Tan has rightly highlighted it. But TSMC's
competing N2 node was already estimated around 65% and reaching roughly 70% by early 2026 — meaning that even on
Intel's optimistic trajectory, it is racing to reach, by the end of the year, roughly where TSMC already is now.
Yield is not a vanity metric; it is the number of usable chips per wafer, and therefore the cost per chip, and a
10-to-15-point yield deficit is a direct cost disadvantage that a rational customer choosing on merit must weigh
against the strategic appeal of a second source. A customer picks a foundry to save money and get the best chips;
if Intel's node yields worse than TSMC's, the customer pays a "support-the-Western-foundry" premium that only
strategic or politically motivated buyers will willingly pay.

And TSMC is not standing still while Intel catches up. The leader is advancing its own nodes, expanding capacity
(including in the United States, partly defusing the very national-security argument that is Intel's biggest
selling point), and reinvesting its enormous profits into staying ahead. Intel is chasing a moving target with a
business that loses billions, and there are even reports that Intel may de-emphasize 18A for external customers
and pin its hopes on the later 14A node — which, if true, would mean the node that was supposed to win customers
this year has already been partly conceded, with the real battle pushed further into the future. A turnaround
whose decisive product keeps receding into the next node is a turnaround whose proof keeps being postponed.

The hardest business in technology to break into

There is a reason Intel's external foundry revenue is so small, and it is not merely that Intel's technology
trails — it is that the contract-manufacturing business is one of the most difficult to enter in all of
technology, for reasons that have nothing to do with how good your factories are. A foundry's customers are
handing it their most precious intellectual property — the chip designs that are their entire competitive
advantage — and trusting it to manufacture them reliably, on schedule, at promised yields, for years. That
trust is earned over a decade of flawless execution, and it is precisely what TSMC has spent forty years
building: a reputation for being the dependable, neutral manufacturer that does not compete with its own
customers and never lets them down.

Intel has two structural problems here that money and government backing cannot quickly fix. The first is the
trust deficit: Intel's foundry has no track record as a contract manufacturer, and its history of manufacturing
stumbles on its own products is exactly the history a prospective customer scrutinizes. The second, and more
insidious, is the conflict of interest: Intel is not a neutral foundry like TSMC — it designs and sells chips
that compete directly with the very companies it is asking to become foundry customers. A company weighing
whether to have Intel manufacture its flagship processor must reckon with the fact that Intel's own product
division builds a rival to that processor. TSMC competes with none of its customers, which is a core reason they
trust it with their crown jewels. Intel competes with many of them, and no amount of strategic goodwill fully
dissolves the hesitation that creates.

This is why the AMD precedent hangs over the whole endeavor. AMD — Intel's archrival — spent years struggling
with exactly the integrated design-and-manufacture model Intel is now doubling down on, and ultimately spun off
its factories entirely, going "fabless" and outsourcing its manufacturing, much of it to TSMC. AMD concluded
that owning leading-edge fabs was a capital-devouring distraction it was better off escaping, and that decision
helped fuel its remarkable resurgence. Intel is betting the opposite — that owning the fabs and turning them into
a contract foundry is its path back — and it is asking customers to embrace a model that its most successful
competitor deliberately abandoned. That does not mean Intel is wrong; strategic necessity and government backing
change the calculus. But it is a reminder that the integrated-foundry path Intel is now selling as its salvation
is the same path the rest of the industry, given the choice, has spent two decades walking away from.

What the bulls genuinely get right

In fairness, the bull case is substantial, and several of its pillars are strong enough that this is a genuine
debate rather than a clear short. First and most important, the government backing and CHIPS Act funding —
roughly $19.5 billion in grants and loans — materially de-risk the single greatest threat Intel faced, which was
that the staggering capital cost of building leading-edge fabs (over $20 billion a year) would bankrupt it before
the foundry could mature. With Washington as a shareholder and a funder, Intel has the financial runway and the
political protection to keep investing, and a company that cannot be allowed to fail for national-security reasons
enjoys a kind of backstop few businesses ever get. Second, the operational progress is real: data-center and AI
revenue grew 22%, foundry revenue grew 16%, 18A yields are genuinely improving and heading to high-volume
manufacturing, and the marquee customer agreements, however preliminary, are real validation that Intel's
technology is good enough to take seriously. Third, the strategic logic is sound — the world genuinely does need
a non-Taiwan leading-edge foundry, and if Intel is the one that provides it, the addressable opportunity is
enormous and the willing, strategically motivated customer base is real.

The honest synthesis is that Intel has a credible, well-funded, strategically vital path to becoming a real
foundry — and that the stock has priced the successful completion of that path as far more certain than the $174
million of external revenue and the company's own "not at scale" disclosure can currently support. The bulls are
right that the backing is real and the progress is real. The question the rally answers too confidently is
whether strategic backing and engineering progress will convert into the commercial reality — paying customers at
scale, profitably, against a TSMC that is pulling away — and that conversion is precisely what has not yet
happened.

The pattern: betting on the validators instead of the validation

Step back and Intel illustrates a subtle and recurring error in how markets price turnarounds: confusing the
prestige of a company's backers for the proof of its business. When the U.S. government, Nvidia, Microsoft, and
Apple are all associated with a company, the human instinct is to reason that they must know something, that such
sophisticated parties would not back a loser, that their involvement is itself the evidence. But each of those
parties has its own motive — geopolitical, strategic, defensive, reputational — and none of those motives is the
same as the simple commercial judgment that Intel's foundry is the best place to make chips. The market has
treated the identity of the backers as a substitute for the commercial results the backers are betting on,
and that substitution is the mispricing.

The tell, as always, is in the financial statements rather than the press releases. The press releases announce
governments and giants and lifetime commitments and improving yields. The financial statements report $174
million of external revenue and a $2.4 billion quarterly loss and a risk disclosure conceding the absence of
customers at scale. When the narrative and the numbers diverge this far, the forensic discipline is to weight the
numbers, because the numbers are what the company is legally required to get right, and the narrative is what it
is incentivized to make exciting. Intel may well close the gap — the backing gives it a real chance that few
turnarounds get. But an investor buying today is paying a tripled price for a commercial outcome that, by the
company's own filing, has not yet arrived.

The kicker

Intel's turnaround has assembled the most impressive guest list in corporate America — the Treasury, the most
valuable chipmaker on Earth, the giants of cloud and consumer technology, all lending their names to the comeback.
And the guest list is real, and it matters, and it gives Intel a fighting chance at a goal of genuine national
importance. But a guest list is not a revenue line. The governments came because they need a domestic foundry;
Nvidia came because it needs a second source; the cloud giants came to hedge and to signal — and all of that
strategic logic, however sound, sits on top of a foundry that still sells almost nothing to outside customers and
loses billions doing it. The stock has tripled on the conviction that the validators must be right. The
validators are betting on a conversion that has not happened, against a competitor that is not waiting, with the
proof postponed to the next node and the one after that.

Washington owns ten percent, Nvidia wrote a check for five billion, and Apple is reportedly on board — and the foundry they are all betting on sold a hundred and seventy-four million dollars of chips to outsiders last quarter and lost the better part of two and a half billion dollars doing it. The backers have voted with their prestige. The customers, who vote with their orders, have not yet shown up in any number that matters — and theirs are the only votes that, in the end, actually get counted.