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 $15 billion share sale. Its stated reason was 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 flat, AMD down 1%.

The real news arrived overnight. At 2:36 in the morning ET on Tuesday (9:36 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 on 31,578,947 more. Net proceeds come to roughly $19.7 billion, and the deal closes on August 12.

A company upsizes an offering for one reason: the order book is larger than the deal. Press reports put demand above $100 billion — five times the $15 billion door Intel walked up to.

The Company Asking for This Money Just Had a Good Quarter

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 the growth is real; the dispute is about cash.

In the same quarter, adjusted free cash flow was negative $8.4 billion — investment exceeded what operations produced by that much.

The 2026 capital expenditure forecast was lifted from $18 billion to $20 billion, and management signaled a "meaningful increase" for 2027. (What Is Cash Flow?)

Intel's manufacturing arm 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 — genuine progress, still roughly 36 cents of loss on every dollar of revenue.

Why Equity and Not Debt

A company has three ways to fund itself: earn the money, borrow it, or issue shares.

The first is running negative this year. The second has a ceiling: the release noted Intel's commitment to an investment-grade rating, and if the rating slips, borrowing gets more expensive.

That leaves the third route: used while the stock is expensive, it raises the same money for fewer shares. Intel is up roughly fourfold over twelve months (165% to 175% year-to-date, by source), so it sold at the richest level in years.

Market capitalization before the sale was roughly $489 billion, so the $20 billion raised is about 4.1% of it — against 4.2% dilution. Unless the new money earns more than Intel's existing capital, incumbent shareholders do not break even.

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

This $20 billion is going into leading-edge capacity, where nearly all the cost is fixed. Once the tools are installed, the number of wafers does not change what the fab costs to own.

A leading-edge fab is either full or it is losing money.

You do not build the plant before you have the customer, but the customer will not sign before seeing the capacity. Each waits on the other.

Several commentators read the phrase "external wafer production" in Intel's release as a tell; Dan Niles said the company appears "close to signing up one or more major foundry customers."

An inference, not a company disclosure.

The $293 Million External Business

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 another.

That $293 million is thirteen times the $22 million booked a year earlier, so the direction is right. The scale is not.

Foundry break-even in 2027 requires "only a few billion dollars of additional external revenue," CFO David Zinsner has said. That is several times an external base running near $1.2 billion a year.

Intel's 14A node targets risk production in 2027 and volume output in 2028. Tesla was reported in April as the first major 14A customer, though that engagement is still at R&D stage, a few thousand wafers a month.

Across the 2010s Intel spent roughly $82 billion buying back its own stock — more than four times today's annual capital budget.

Two Readings

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

The difference is which side is willing to wait until 2028.

What's Left

The ledger as of Tuesday morning: $19.7 billion of net cash arriving, a stake 4.2% smaller, a $20 billion capital budget for 2026. Also a "meaningful increase" promised for 2027, a manufacturing arm losing $2.1 billion a quarter, 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.

Correction (September 13, 2026): prose tightened for rhythm; figures and sources unchanged.