International Finance
FeaturedTechnology

IF Insights: Intel’s USD 20 billion bet on becoming America’s foundry

A record share sale, a Tesla contract and a 175% stock run have put Lip-Bu Tan's turnaround beyond doubt. Beating TSMC is a different question

Intel raised USD 20 billion from an upsized share offering on Tuesday, the largest equity raise in its history and one of the biggest of a year already crowded with them.

The chipmaker had told the market on Monday it wanted USD 15 billion. Demand was so heavy, reportedly north of USD 100 billion in orders, that it lifted the size by a third and priced 210.5 million shares at USD 95 apiece, a discount of just 2.6% to the previous close.

Net proceeds come to about USD 19.7 billion once fees are stripped out, with the deal closing on 12 August. Underwriters JPMorgan, Goldman Sachs, Morgan Stanley and Citigroup hold a 30-day option on a further 31.6 million shares.

The money is going into the most capital-hungry business in technology, contract chip manufacturing. Intel wants to build fabs and advanced packaging capacity fast enough to serve customers other than itself, and it is cashing in a share price that has roughly tripled this year to pay for it.

From near-shutdown to the centre of the strategy
Rewind twelve months and the foundry unit was close to being switched off.

In its July 2025 quarterly filing, Intel warned that without a significant external customer for its next-generation 14A process, it might pause or discontinue the pursuit of leading-edge manufacturing altogether, and slow or stop its Ohio build.

Chief executive Lip-Bu Tan, who took over in March 2025, put it plainly in a memo to staff. There would be no more blank cheques, and investment in 14A would follow confirmed customer commitments.
Intel Graph
That was the bottom. Intel had lost close to USD 19 billion in 2024 and a further USD 3.7 billion in the first half of 2025.
Tan cut roughly a quarter of the workforce, aiming for about 75,000 core employees by the end of 2025, scrapped planned plants in Germany and Poland, and sold control of Altera. Headcount at the end of June stood at 77,600 excluding subsidiaries, down from 96,400 a year earlier.

Then the capital arrived. Washington converted CHIPS Act grants into an USD 8.9 billion equity purchase, taking about 10% of the company at USD 20.47 a share. SoftBank put in USD 2 billion.

Nvidia followed with USD 5 billion alongside a joint product roadmap. For a business that had spent USD 82 billion on buybacks in the 2010s while its process technology slipped, the rescue was pointed.

The Tesla contract that changed the argument
The commercial validation came in April 2026, on Tesla’s first-quarter earnings call, when Elon Musk confirmed that Tesla would use Intel’s 14A process for chips destined for Terafab, the vast AI silicon complex he is building in Austin. SpaceX is expected to use the same node.

Terafab is an extraordinary ambition, targeting something in the order of one terawatt of annual compute, roughly double current US output, at a cost Bernstein has put anywhere between USD 5 trillion and USD 13 trillion.

In the near term Tesla is building a roughly USD 3 billion research fab at its Texas gigafactory campus running only a few thousand wafers a month for validation, with volume production the job of a separate high-capacity operation.
The direct revenue is therefore small and years away. The signalling value was enormous. Intel had told the world it needed an anchor 14A customer to justify staying in the leading-edge race, and it got one with a name nobody could ignore.
Tan has since committed to high-volume 14A production in 2028, with risk production in the second half of 2027. Optimism about a second marquee client grew further after US President Donald Trump said Apple would make processors with Intel, though neither company has confirmed it.

Where the profits actually come from
One caveat is worth stating clearly, because the market narrative has run ahead of the accounts. Intel Foundry is not yet a profit engine. It is the strategic centre of the company and its fastest-growing reporting line, but it is still loss-making.

In the June quarter, foundry revenue rose 31% to USD 5.77 billion while the unit lost USD 2.09 billion at the operating level. That is a real improvement on the USD 3.17 billion loss a year earlier, and on the USD 2.51 billion of the final quarter of 2025, but it is a narrowing loss, not a profit.

Intel Graph
Almost all of that revenue is internal, Intel Products buying wafers from Intel Foundry, which is why intersegment eliminations came to USD 5.48 billion.

The profits sit in products. Data Centre and AI, where revenue jumped 59% to USD 6.26 billion, delivered USD 2.47 billion of operating income. The client group added USD 2.34 billion. Group revenue of USD 16.13 billion was up 25%, the strongest growth in more than fifteen years, and the seventh consecutive quarter to beat guidance. Non-GAAP earnings came in at USD 0.42 a share. The headline GAAP loss of USD 2.16 a share reflects a USD 12.5 billion non-cash charge on shares held in escrow for the government under the Secure Enclave agreement, not operating deterioration.

A stock run in a once-in-a-generation sector rally
Intel shares have gained about 175% in 2026 and have roughly quintupled since last August. That is a comfortable win over both AMD and Nvidia, and over the Philadelphia Semiconductor Index, which is up around 75% for the year.

The sector backdrop has been extraordinary. The SOX index rose 101% in the first half of 2026, its best half since 1999, after an 87.8% second quarter that was the largest quarterly gain in records going back to 1994. By comparison the Nasdaq Composite added 12.8% and the S&P 500 9.6% over the same six months.

Memory led, with SanDisk up more than 700% and Micron around 300%. Intel gained 257% by the end of June. Nvidia, oddly, was near the bottom of the index with single-digit gains as money rotated out of the obvious AI winners and into the infrastructure behind them.

That rotation cut both ways. The index fell 19% from its record high during July, Intel gave back close to a third from a peak of USD 142.35, and Marvell dropped 40% from its own high.

Sector forward earnings multiples of around 26 times sit well above the ten-year average. Tuesday’s raise, priced at a modest discount into a hot book, looks very much like a management team taking the money while it is on the table.

Can Intel actually take on TSMC
Not on current form, and the gap is not close.

TSMC held 72.3% of the global foundry market in the first quarter of 2026, up from 70.4% the quarter before, on USD 35.86 billion of revenue. Samsung was second with 6.5% and SMIC third with 5.1%.

Intel does not feature in the top ten by external foundry revenue at all. TSMC made USD 122.4 billion in 2025, serves more than 500 customers across some 12,000 products, and is spending USD 52 billion to USD 56 billion this year. Intel’s 2026 capex guidance, raised in July from USD 18 billion, is USD 20 billion.
Intel Graph
Scale is only part of it. A foundry sells trust, design kits, verified IP libraries and predictable yields as much as it sells wafers, and TSMC has spent nearly forty years building that. Intel is still assembling the 14A design kit, with version 0.9 the release at which customers commit real volume.

There are genuine reasons to take Intel seriously. 18A is in high-volume manufacturing using High-NA EUV lithography, ahead of TSMC on that equipment. Yields have been improving at around 7% to 8% a month.

Intel has ten long-term agreements signed, a named customer in Fortinet, a custom silicon business running at roughly USD 2 billion annually, and it is supply constrained rather than demand constrained.
On timing, 14A volume production in 2028 lines up against TSMC’s A14 in the same year.

The realistic prize is not leadership. It is becoming the credible second source for leading-edge logic, in the United States, for customers who want an alternative to a single Taiwanese supplier.

That is a large and defensible business. Whether USD 20 billion buys it is the question the next three years will answer.

What's New

Deutsche Bank becomes first European bank to clear China’s Renminbi

International Finance Business Desk

Singapore doubles down on AI boom, raises forecast after Q2 GDP growth

International Finance Business Desk

Norway wealth fund posts record USD 184 billion profit as CEO warns of risks

International Finance Business Desk

Leave a Comment

* By using this form you agree with the storage and handling of your data by this website.