Intel Raised Its $15 Billion Share Sale to $20 Billion

The Second Press Release, Filed at 2:36 a.m.

On Monday morning in New York, Intel announced a share sale. The company wanted $15 billion, and the stated reason ran to a single sentence: "general corporate purposes, which may include, but are not limited to, capital expenditures and working capital." The market read that as dilution. The stock fell 5% intraday to $96.97. The rest of the semiconductor complex barely moved — Nvidia and Broadcom finished flat, AMD down 1%. The selling was specific to Intel, not a verdict on the supply chain.

The real news arrived overnight. At 2:36 a.m. ET on Tuesday (9:36 a.m. in Istanbul), a second release went out: the offering had been upsized from $15 billion to $20 billion and priced at $95 a share. That is 210,526,315 new shares, with a 30-day option for the underwriters to buy 31,578,947 more. Net proceeds to the company, after fees, come to roughly $19.7 billion. The deal is expected to close on August 12.

The gap between the two numbers is not a bookkeeping detail. A company raises the size of an offering for one reason: the order book is larger than the deal. According to press reports, demand exceeded $100 billion. Intel walked up to a $15 billion door and found more than five times that standing behind it.

This piece is not about a one-day move in a stock price. It is about when a company issues new equity, why it does that instead of borrowing, and why the economics of the factory this money will build force the capacity to be committed before the customer is.

The Company Asking for This Money Just Had a Good Quarter

Equity raises usually travel with bad news. This one did not. Intel's second-quarter results showed the fastest growth in fifteen years: revenue of $16.1 billion, up 25% year over year. The data center and AI segment reached $6.3 billion, grew 59%, and ran at a 40% operating margin. Adjusted earnings came in at $0.42 a share, nearly double consensus.

Nobody disputes that the growth is real. The dispute is about cash. In the same quarter, Intel's adjusted free cash flow was negative $8.4 billion — the company spent $8.4 billion more on investment than its operations produced. The 2026 capital expenditure forecast was lifted from $18 billion to $20 billion, and management signaled a "meaningful increase" for 2027. Why free cash flow diverges from reported profit is its own subject: What Is Cash Flow?

And then there is the foundry. Intel's manufacturing arm, the one that makes wafers for outside customers, booked $5.8 billion of revenue in the quarter and lost $2.1 billion at the operating line. The loss narrowed by $1.1 billion from a year earlier, which is genuine progress. It still amounts to roughly 36 cents of loss on every dollar of revenue.

The chart below tracks Intel over the past three months — useful for seeing the price level against which this week's sale was struck.

Why Equity and Not Debt

A company has three ways to fund itself: earn the money, borrow it, or issue shares. The first route is running negative this year. The second has a ceiling — the press release went out of its way to note Intel's commitment to an investment-grade credit rating. The rating is what sets the coupon on debt; if it slips, new borrowing gets more expensive and covenants on existing debt start to bite.

That leaves the third route, which has one advantage: used while the stock is expensive, it raises the same money for fewer shares. Intel is up roughly fourfold over the past twelve months. Year-to-date figures vary by source, between 165% and 175%. The company sold this stock at the richest level it has seen in years.

There is a symmetry here worth pausing on. Intel's market capitalization before the sale was roughly $489 billion. The $20 billion raised is about 4.1% of that. The dilution is 4.2%. The two numbers are essentially identical, and the implication is blunt: unless the new money earns more than Intel's existing capital does, incumbent shareholders do not even break even. An equity raise is not free money. It is an exam the company has just enrolled in.

The Arithmetic of a Fab: What an Empty Line Costs Per Wafer

Now to the mechanism. This $20 billion is not going into a software company's marketing budget; it is going into leading-edge manufacturing capacity. The economics of that kind of plant differ fundamentally from a retailer's: nearly all of the cost is fixed. Once the tools are installed, how many wafers pass through them does not change what the fab costs to own.

The conclusion compresses into one sentence: a leading-edge fab is either full or it is losing money, with very little in between. From that follows a rule of behavior. You do not build the plant before you have the customer. But the customer will not sign before seeing the capacity — nobody commits a chip two years out to a line that does not yet exist. Each waits on the other.

That is where the news value of this raise sits. The party that breaks the deadlock is the one that finds the money first and commits the capacity. Which is why several market commentators read the phrase "external wafer production" in Intel's release as a tell: if the money has been raised, perhaps a customer has been signed but not yet disclosed. Dan Niles, founder of Niles Investment Management, said the company appears "close to signing up one or more major foundry customers." That is an inference, not a company disclosure, and it should be read as one.

The $293 Million External Business

One figure keeps the signal in proportion. Of Intel Foundry's $5.8 billion of second-quarter revenue, only $293 million came from genuine outside customers. Everything else is Intel manufacturing Intel's own chips — money moving from one pocket to the other.

That $293 million is thirteen times the $22 million booked a year earlier, so the direction is right. The scale is not. Chief financial officer David Zinsner has said foundry break-even in 2027 requires "only a few billion dollars of additional external revenue." Measured against an external base running near $1.2 billion a year, "only a few billion dollars" means several times the entire current business.

The calendar is not close either. Intel's 14A node — the one this bet rests on — targets risk production in 2027 and high-volume customer output in 2028. In April, Tesla was reported to be the first major customer for 14A; even that engagement is at a research-and-development stage today, running a few thousand wafers a month.

The Late Bill for $82 Billion of Buybacks

This story has a long tail. Across the 2010s, when it dominated the processor market, Intel spent roughly $82 billion buying back its own stock. Repurchases shrink the share count, lift earnings per share, and support the price. In the short run, shareholders enjoy them.

The stock traded far below today's level in those years, so on price alone the company bought back cheaply what it has now sold at $95. The problem is not the price. It is the opportunity cost. That $82 billion is more than four times Intel's entire annual capital budget today. Had it gone into fabs over a decade, the capacity Intel is now waiting on would already exist.

In the words of investment director Russ Mould, the company "went a long way to wrecking its own balance sheet and prospects by focusing on financial engineering rather than physical engineering." That is a judgment, but the sequence supports it: a company that once retired its shares is issuing them a decade later to buy capacity.

Two Readings

The same event supports two coherent interpretations, and today's information refutes neither.

No line in that table contradicts another. The difference is which side is willing to wait until 2028.

The chart below shows QQQ, which tracks the broad technology index, over the same three-month window — the backdrop against which Intel's story is playing out.

What Is Left

As of Tuesday morning, here is the ledger: $19.7 billion of net cash arriving, a stake 4.2% smaller, a $20 billion capital budget for 2026, a promise of a "meaningful increase" for 2027, a manufacturing arm losing $2.1 billion a quarter at the operating line, and a process node meant to reach volume in 2028.

There is also an unverified assumption: that raising the money means the customer has been found. The company has said no such thing. The market has priced it anyway — and pricing a sentence nobody said is the most expensive habit in valuation.

This article draws on public news sources, company press releases, and analyst commentary reported in the press as of the date of publication. The size of the order book and the customer claims have not been confirmed by the company. This is not investment advice.