Intel spent $82bn buying its own shares. Now it wants $15bn back

Intel spent $82bn retiring its own shares in the 2010s. On Monday it started selling new ones, $15bn worth, at roughly five times last summer's price. The filing barely says what the money is for.


Intel spent $82bn buying its own shares. Now it wants $15bn back
Image Credits Credit: Khlnmusa

Intel announced a proposed $15 billion underwritten public offering of common stock before the market opened on Monday. It filed a registration statement on Form S-3 with the SEC. Underwriters hold a 30-day option on up to $2.25bn of additional shares. J.P. Morgan, Goldman Sachs, Morgan Stanley and Citigroup are joint book-running managers.

The use of proceeds runs to a single sentence. The money goes to general corporate purposes, “which may include, but are not limited to, capital expenditures and working capital.”

The release sets no price, discloses no share count and names no project. It does frame the raise defensively. The offering lets Intel pursue growth “while maintaining a strong balance sheet and its commitment to an investment-grade rating.”

The buyback goes into reverse

Russ Mould, investment director at AJ Bell, supplied the arithmetic to Reuters. Intel, he said, went “a long way to wrecking its own balance sheet and prospects by focusing on financial engineering rather than physical engineering, courtesy of $82 billion of share buybacks in the 2010s.”

Raising money now “makes perfect sense”, Mould added, “especially after a five-fold increase in the stock price since last August.”

Set the two halves side by side. Intel retired $82bn of stock through the decade when it still owned the server and PC processor market. It is printing stock now, at five times the level of a year ago. The market has repriced it as a turnaround. The capital policy has flipped, and so has the price.

Four words matter more than the number

One section of the release carries real content, headed “Why Now”. Customers keep signalling strong and sustainable demand, Intel says, driven by unprecedented investment in AI compute. Then it names four growth areas: physical AI, purpose-built silicon, advanced packaging and external wafers.

External wafers is the tell. Intel committed in July to high-volume production on its 14A process in 2028. It had warned the node could go unbuilt without a major external customer. Reuters reports that Tesla has signed on for 14A. Trump said Apple would make processors with Intel, though neither company confirmed it.

The gap sits in what Intel has not said. Its first publicly named foundry customer under chief executive Lip-Bu Tan was Fortinet, on the older Intel 4 node. That deal buys a slice of a roughly $2bn hardware business, not the $10bn-a-year anchor Intel wants. When it committed €5bn to its Irish fab last month, it still had no external customer taking meaningful volume.

Analysts filled the silence. Ben Bajarin wrote: “Tell me you have an external wafer customer without telling me you have an external wafer customer.”

Patrick Moorhead read the filing the same way, arguing that Tan “doesn’t deploy capital without a customer.”

Both readings are inference. Intel has named no new customer, and the offering documents name none either.

What $15bn buys

Intel raised its 2026 capital expenditure forecast from $18bn to $20bn in July. Chief financial officer David Zinsner told CNBC that most of the increase would go on factory tooling. The company is bracing for a “meaningful increase” in 2027, he added.

Tooling swallows real money at this scale. Intel became the first chipmaker to ship a volume logic product patterned on ASML’s High NA EUV machines. Each scanner costs around $400m. The Irish expansion alone accounts for more than 25% of planned 2026 capital spending, Reuters reports.

Everyone else is spending too. Goldman Sachs estimates technology capital expenditure will reach $765 billion this year and $1.2 trillion in 2027. Bloomberg Intelligence analyst Robert Schiffman said the raise lets Intel fund AI and foundry work without adding leverage. He put it in sector terms. Rising investment, he said, “doesn’t have to fall entirely on bondholders.”

The market stepped back

Shares fell on the news. Bloomberg recorded a drop of as much as 5.3%, to $96.30. Reuters and CNBC both logged declines of about 4% in early trading. Dilution does that, and the stock had a long way to fall from.

Intel has roughly quintupled over the past twelve months, and its market value sits close to $500bn. Estimates of the rise so far in 2026 vary by outlet. They run from more than 160% to nearly triple, depending on the measurement date.

The company reported its fastest revenue growth in nearly 15 years in July. Data centre revenue rose 59% year on year. Its shares also jumped nearly 8% on job cuts in the data centre and AI group. That says something about what investors have rewarded.

One problem sits underneath all of it. Intel still lacks an AI accelerator competing head to head with Nvidia or AMD, as Bloomberg notes. The US government holds a 10% equity stake taken last August. Intel’s own risk factors name that stake, the 14A programme and its hunt for significant external customers.

Bloomberg described the raise as what may be Intel’s first public share sale since it listed in 1971. The wire hedged the claim. The hedge earns its keep. Nothing in the filing confirms the thing everyone wants confirmed, which is who agreed to buy the wafers. Until Intel names that customer, $15bn rides on a demand signal the company will describe but not identify.

Get the TNW newsletter

Get the most important tech news in your inbox each week.

Also tagged with