Ousted by the Blueprint

The Five-Node Bet

Eli: February 15th, 2021. Pat Gelsinger walks outside the Robert Noyce Building on Intel's headquarters campus in Santa Clara, California, as the company's new chief executive. That's the picture Intel itself put out of the day. And it's a homecoming. He first joined Intel in 1979, at eighteen, as a quality-control technician. He spent the first thirty years of his career there before leaving. Now he's back to run it.

Mara: And the listener's question starts from the far end of that story: why did Pat Gelsinger step down from Intel in December 2024, after championing its turnaround? It's Friday, October 9th, and today's angle is the architect ousted by his own blueprint.

Eli: His road back ends in a boardroom in late November 2024. Intel has not published the verbatim discussion or the vote breakdown from that meeting.

Mara: The plan got concrete fast. On March 23rd, 2021, Arizona's governor and the U.S. commerce secretary joined Intel executives to announce a fab expansion in Arizona. The price tag: $20 billion.

Eli: Gelsinger's line that day: 'We are excited to be partnering with the state of Arizona and the Biden administration on incentives that spur this type of domestic investment.' The factories served a strategy he called IDM 2.0. Intel would keep designing chips and making them in its own plants. It would also make chips for other companies, as a foundry. And it came with a roadmap: five nodes in four years. A node is one generation of manufacturing technology. Five of them, squeezed into four years, to leapfrog rivals. The last stop was called 18A.

Mara: So: Intel-brand chips, plus a factory for hire, all under one roof.

Eli: One roof, two businesses, and the aim was world-class at both. 18A was meant to be ready for production around 2024 into 2025. Frank Yeary, the board chair, later said Pat 'returned at a critical time for the company in 2021.'

Mara: Critical, and expensive. Chipmaking at the cutting edge is capital intensive: cleanrooms and lithography machines swallow tens of billions before a single chip is sold. By December 2021 Gelsinger was saying so out loud. His capex, the spending on plants and equipment, would climb: 'we’re going to lower our margins, increase our CAPEX significantly, to build a manufacturing technology powerhouse. And on a quarterly basis, that’s going to suck.'

Eli: He was blunt about the reaction, too. The markets, he said, rewarded Intel 'with about a 10 percent reduction of our stock price.' And he'd set himself a clock. When he took the job, he said, he saw it as 'a five-year assignment to get it done.'

Mara: Five years from 2021 runs into 2026. Wall Street grades Intel every three months.

Eli: A plan built to pay off over years, graded every quarter. By 2024, Mara, the quarterly grades were piling up.

Fracture in August

Mara: Part of the bill was supposed to be shared with Washington. Gelsinger lobbied for the CHIPS Act, the federal law paying to bring chipmaking back onto U.S. soil. Intel then negotiated its own award: up to nearly eight billion dollars. That's well under half the price of the Arizona expansion alone. And it wasn't a check up front. The money was tied to terms, and to Intel actually executing.

Eli: And execution is where 2024 got rough. Take AI. Intel's answer to the boom was Gaudi, an AI accelerator, meaning a chip built specifically to train and run AI models rather than a general-purpose processor. Intel set Gaudi a $500 million sales target for 2024. It missed. In October, Gelsinger's own explanation: 'the overall uptake of Gaudi has been slower than we anticipated.' He pointed to the switch from one Gaudi generation to the next, and to software that wasn't easy enough to use.

Mara: So, a product transition. A slow stretch, the kind any chipmaker has.

Eli: Partly. But The Daily Herald reported an internal side too. Gelsinger had made optimistic claims about prospective AI-chip deals, beyond Intel's own estimates. That led Intel to scrap a recent revenue forecast.

Mara: In a plan graded quarterly, a scrapped forecast lands on the next report card.

Eli: The worst report card came in August. Intel Foundry, the factory business, lost about $2.8 billion in the second quarter of 2024. That's one division, in three months. So Intel cut deep. It suspended its dividend, the regular cash payout to shareholders, stopped to keep the money inside the company.

Mara: About 15,000 jobs went, roughly one in seven, from the biggest workforce Intel had ever had.

Eli: Then a departure upstairs. On August 19th, Lip-Bu Tan resigned from Intel's board, over disagreements with Gelsinger's strategy and spending. Tan had turned around a faltering chip company himself. His public statement called it a 'personal decision to reprioritize various commitments.' But Reuters, citing unnamed sources, reported frustration with Intel's 'bloated workforce, risk-averse culture, and lagging strategy.' Reporting read Tan's exit as a sign of deep discontent with the pace of the turnaround, and of Gelsinger's standing among the directors weakening.

Mara: So Washington's money depended on Intel executing, and one of the board's own chip veterans had just walked out over how it was executing. He went into the autumn with his plan intact and one fewer chip operator at the board table.

The Sixteen-Billion Shock

Eli: By September, Gelsinger was reshaping the company under that pressure. He and the board went through the strategy, the portfolio and the plan from August. Out of it came a wall: Intel Foundry would become an independent subsidiary. It would get its own board, with independent directors who have deep semiconductor expertise, and its own financial reporting. The point was to court outside customers. The ambition behind it was the one he'd voiced earlier that year: 'Through our foundry I want to manufacture every AI chip in the industry.'

Mara: So Intel still owns the factories.

Eli: And he had a number for 18A. At Deutsche Bank's technology conference that September, Gelsinger reported a defect density below 0.4 per square centimeter. Defect density counts the flaws on a patch of silicon. Fewer flaws, more chips that actually work.

Mara: Which brings back the man walking out of the Noyce Building in 2021. Three and a half years in, the last of his five nodes had a number to show. What's still unknown is whether enough 18A chips were coming out working for real volume production.

Eli: And then came October 31st, and the third-quarter results.

Mara: Intel reported a net loss of about $16.6 billion for that one quarter. A year earlier, the same quarter had shown a small profit. Most of that loss wasn't cash walking out the door. It was impairment charges: the company admitting, on paper, that things it owns are worth far less than its books said. Part of it was manufacturing, mostly the older Intel 7 process. That's the same side of the business those Arizona billions were meant to build up.

Eli: And the market had been voting all year. Intel's stock lost more than half its value in 2024. In November, Nvidia took Intel's place on the Dow Jones Industrial Average.

Mara: By late November, Eli, the directors were weighing a plan that still asked for years against a share price that had halved in one.

The Ultimatum

Eli: Late November 2024. Intel's board meets. According to a person familiar with the matter, directors felt Gelsinger's costly plan to turn Intel around wasn't working, and change wasn't coming fast enough. So they gave him a board ultimatum: a choice the board sets, with no third option. Retire, or be removed. Reuters and Bloomberg each reported it independently. He chose to retire. The chair, Frank Yeary, orchestrated the exit and pushed for spinning off the foundry. Less than a week before he left, Washington had finalized Intel's CHIPS award.

Mara: So the patience he'd asked for in 2021 had run out on the board. By then Intel was trading below book value, for the first time since 1981. Book value is roughly what the company's assets are worth on paper, minus what it owes. The market was valuing Intel at less than that.

Eli: Publicly, it read differently. On December 2nd, Intel announced Gelsinger had retired, effective the day before. One Intel distribution executive, who asked not to be named, said retirements there usually come with a long transition, 'not, ‘I’m retiring, and today’s my last day.’'

Mara: The official reading is Intel's own and Yeary's, Eli: a retirement. And the separation agreement is built like one. Eighteen months of base salary. A multiple of his target bonus on top. Most of his 2024 bonus as well. What it didn't settle was who would run Intel the next morning, or what became of 18A.

The Return of the Skeptic

Eli: It came before dawn on Monday, December 2nd: an email inviting Intel staff to an all-hands that morning. Taking their questions were two interim co-CEOs, two senior executives splitting the top job until a permanent one was found. One was the finance chief. The other, Michelle Johnston Holthaus, was also named CEO of Intel Products. Yeary became interim executive chair, and set the tone: 'As a board, we know first and foremost that we must put our product group at the center of all we do.'

Mara: Products at the center, but the foundry plan and 18A stayed in place, with new stress on profitability and returns. Then in March 2025 the board named a permanent CEO: Lip-Bu Tan. The director who'd walked out the summer before. Yeary went back to independent chair.

Eli: Gelsinger, on his last earnings call, had said: 'We know what we need to do – now it is all about execution.' In 2025 he said he'd wanted to finish what he started at Intel, and wasn't given the opportunity.

Mara: Then the full year landed. Analysts had projected Intel's first annual net loss since 1986. They got it: $18.8 billion, about five times what they'd expected. And the heaviest weight sat in the factory business. Intel Foundry lost about $13.4 billion over the year. As an external foundry, building chips that other companies design, it depends on those companies showing up. Intel's own filing said it had 'few external customers to date.'

Eli: Lip-Bu Tan, as CEO, has said 'the company invested too much, too fast, without sufficient demand.' Others read it as time cut short. Randy Copeland called the exit unwise, given how much Intel had transformed in three and a half years. In that reading, Wall Street pressure and board impatience cut a long manufacturing rebuild short.

Mara: So, to the listener's question. Pat Gelsinger stepped down because Intel's board lost confidence in his costly plan: billions lost in the foundry, AI sales that didn't arrive, investors who kept selling. And the directors gave him a choice: retire, or be removed.

Eli: In February 2021 he walked outside the Noyce Building asking for five years. He got a little under four. When he left, the last of his nodes, 18A, had a promising defect number and its volume production still ahead.

Mara: What's still unknown starts in that boardroom. Watch for the minutes of the late-November meeting, or how individual directors voted. And watch whether 18A's yields in late 2024 actually met Intel's own bar for high-volume production, beyond the defect number it made public.

Eli: And watch the foundry itself: whether Intel keeps majority control of the subsidiary, in line with its CHIPS Act terms, or spins it off as its own public company.

Mara: Look up Five nodes in four years.

Eli: And stay with us on business.

Mara: That's our story for today. I'm Mara.

Eli: And I'm Eli. Thanks for listening to Angle, and until next time.