Pat Gelsinger's Fall from Grace

choi

TSMC rivals found out that second place is worthless in chip foundry: United Microelectronics Corporation (UMC 聯華電子) was world's second largest foundry until 2012. See Dylan McGrath, Globalfoundries Projected to Surpass UMC in Foundry Rankings. DesignLines, Aug 21, 2012
https://www.eetimes.com/globalfo ... n-foundry-rankings/
("Globalfoundries (Milpitas, Calif.) is expected to achieve foundry sales of $4.29 billion in 2012, up 23 percent from 2011, according to a mid-year update of IC Insights' McClean Report [for paid clients only]. UMC, long the No. 2 ranked foundry by sales, is expected to report sales for the year of about $3.78 billion, roughly flat with 2011, according to the report")

Now Intel has learned the same bitter lesson.

(1) 실리콘밸리=오로라 특파원, AI 파도서 좌초… 인텔의 '구원투수'도 짐 쌌다; 인텔 CEO 겔싱어 3년 만에 사임. Seoul: Chosun Ilbo 朝鮮日報, Dec 4, 2024
https://www.chosun.com/economy/t ... W5FOTGN22JWPK4JHIY/
("올해 2월 인텔 연례 행사에서는 2030년까지 파운드리 업계 2위가 되겠다고 공언했다. 파운드리 1위인 대만 TSMC와 2위 삼성전자가 구체적인 계획조차 내놓지 않은 1나노 공정의 반도체 생산도 내년 시작할 수 있다고 장담했다. 하지만 그의 계획에는 '실현 가능성' 에 대한 의문이 꼬리표처럼 따라다녔다. 반도체 업계에선 '너무 꿈같은 얘기를 해서 신뢰성이 떨어진다' 는 말이 나왔다. * * * 업계에선 인텔이 내년 양산을 계획하던 18옹스트롬(1.8나노에 해당) 공정의 수율은 10%가 채 안 된다는 분석이 나왔다. 이에 고객사 브로드컴은 인텔 반도체 주문을 취소했다")

(a) The newspaper has English website, which does not have translated version of this report.
(b) Google Translate:
The reporter, title (followed by subtitle after the semicolon) and quotation, respectively, are

[reporter:] 실리콘밸리=오로라 특파원 Silicon Valley = Aurora Correspondent,
[title:] AI 파도서 좌초… 인텔의 '구원투수'도 짐 쌌다 Stranded in AI Waves; Intel's 'Relief Pitcher' Packs Up Too;
[subtitle:] 인텔 CEO 겔싱어 3년 만에 사임  Intel CEO Gelsinger resigns after three years.

[quotation 1:] 올해 2월 인텔 연례 행사에서는 2030년까지 파운드리 업계 2위가 되겠다고 공언했다. 파운드리 1위인 대만 TSMC와 2위 삼성전자가 구체적인 계획조차 내놓지 않은 1나노 공정의 반도체 생산도 내년 시작할 수 있다고 장담했다. 하지만 그의 계획에는 '실현 가능성' 에 대한 의문이 꼬리표처럼 따라다녔다. 반도체 업계에선 '너무 꿈같은 얘기를 해서 신뢰성이 떨어진다' 는 말이 나왔다.
At Intel's annual event in February of this year, he [Pat Gelsinger] declared that he would become the second largest foundry company by 2030. He also boasted that production of semiconductors using the 1nm process, which the number one foundry company, Taiwan’s TSMC, and number two Samsung Electronics have not even released specific plans for, could begin next year. However, questions about the 'feasibility' of his plan followed him around like a tail. People in the semiconductor industry said, 'It's too dreamy, so it's not credible.'

[quotation 2:] 업계에선 인텔이 내년 양산을 계획하던 18옹스트롬(1.8나노에 해당) 공정의 수율은 10%가 채 안 된다는 분석이 나왔다. 이에 고객사 브로드컴은 인텔 반도체 주문을 취소했다
Industry analysts say that the yield of the 18 angstrom (equivalent to 1.8 nanometers) process that Intel was planning to mass produce next year is less than 10%. As a result, its client Broadcom canceled its order for Intel semiconductors.

(c) This Chosun report said something that nobody knew or reported before: "the yield of the 18 angstrom (equivalent to 1.8 nanometers) process that Intel was planning to mass produce next year is less than 10%.


(2) Above, quotation 2 stated that its client Broadcom canceled its order for Intel semiconductors. For this, see
Max A Cherney, , Exclusive: Intel Manufacturing Business Suffers Setback as Broadcom Tests Disappoint. Reuters, Sept 4, 2024
https://www.reuters.com/technolo ... sources-2024-09-04/

the first two paragraphs:

" Intel's contract manufacturing business has suffered a setback after tests with chipmaker Broadcom failed, three sources familiar with the matter told Reuters, dealing a blow to the company's turnaround efforts.

"The tests conducted by Broadcom involved sending silicon wafers - the foot-wide discs on which chips are printed - through Intel's most advanced manufacturing process known as 18A, the sources said. Broadcom received the wafers back from Intel last month. After its engineers and executives studied the results, the company concluded the manufacturing process is not yet viable to move to high-volume production.

(a) However, this Reuters report did not say WHY Broadcom was disappointed. Specifically, Reuters did not say the yield was 10%.
(b) On the other hand, common sense dictates that Intel must fail in pursuing 19A, as it has not installed ASML's EUV (extreme ultraviolet) equipment (Intel did order this year, but the equipment has not arrived; it takes time to manufacture the equipment).


(3) Dan Gallagher (columnist in WSJ section "Heard on the Street"), Are Intel's Problems Too Big to Fix? Wall Street Journal, Dec 4, 2024, at page B13
https://www.wsj.com/tech/are-int ... big-to-fix-442a7dd7
("And while the company's stock jumped after its third-quarter report showed data-center revenue beating Wall Street's expectations after four straight quarters of misses, the unit is still generating half the annual revenue it made in 2020, just before Gelsinger took over")

"data-center revenue * * * the unit [data-center chips] is still generating half the annual revenue it made in 2020" does NOT mean that data-center revenue makes up half of Intel's 2024 revenue (the reality is that the unit [data-center chips] is still generating half the annual revenue it made in 2020 is almost the total of Intel revenue, not selling AI chip and failing foundry business). Rather, the sentence means Intel's data-center revenue is just half of its 2020 revenue in the same section (data-center).
----------WSJ
Pat Gelsinger’s quest to save Intel has come to an abrupt end. Whoever runs the storied chip maker next is still going to have to pull off a heroic feat.

Intel announced Monday that Gelsinger is retiring as chief executive and stepping down from the board of directors. The move had the tone of a peaceful transition, but the reality was still clear, with Gelsinger calling the day “bittersweet” and board chair Frank Yeary noting the goal of “restoring investor confidence” in the press release. Intel’s stock lost 61% of its value between Gelsinger’s first day on the job in early 2021 and Friday’s close, making it the worst performer on the PHLX Semiconductor Index over that time, according to data from S&P Global Market Intelligence. The S&P 500 has gained 53% over the same period.

A CEO exit following such an uninspiring run naturally sparks some hope; Intel’s shares jumped more than 5% Monday morning before ending the day with a slight loss. But the move raises even more questions about the company’s path, both in the short and the long run. Current financial chief David Zinsner and Michelle Johnston Holthaus, who has been running Intel’s PC chip business, will act as co-CEOs while the board seeks a permanent successor. That keeps the top job in limbo as Intel supposedly is nearing the end of its ambitious race to catch up its manufacturing processes with chip-making giant Taiwan Semiconductor Manufacturing.

The culmination of that race is a production process called Intel 18A. Intel said on its last earnings call on Oct. 31 that the first chips manufactured on that process are expected to start shipping in the middle of next year. Much rides on their success as 18A is the final phase of Gelsinger’s plan to have Intel race though five so-called nodes in four years (Intel used to spend at least two years on a single node).

A CEO switch so late in that cycle, therefore, naturally raises some eyebrows. “As the standard-bearer for the company’s ‘five nodes in four years’ guiding mantra, Mr. Gelsinger’s sudden departure leaves us unsure of the strategic path ahead for Intel,” wrote Joshua Buchalter of TD Cowen in a note to clients on Monday. In his own report, Stacy Rasgon of Bernstein said, “We might have expected Pat to at least make it until 18A is out the door (at which point we would see how it stacks up), and as he hasn’t, one has to wonder whether his departure foreshadows any negative implications for the health of the process roadmap.”

Intel’s other major problems involve both selling the chips that it makes and finding companies willing to use its factories to make their own. Neither effort is going great at the moment. Intel’s foundry business, which handles manufacturing for external clients, has lost more than $11 billion in the first nine months of 2024—nearly double what it lost in the same period last year. And while the company’s stock jumped after its third-quarter report showed data-center revenue beating Wall Street’s expectations after four straight quarters of misses, the unit [data-center chips] is still generating half the annual revenue it made in 2020, just before Gelsinger took over.

That is due to Intel losing share to Advanced Micro Devices AMD -2.00%decrease; red down pointing triangle in server CPU chips and booming demand for Nvidia’s NVDA 0.16%increase; green up pointing triangle GPU chips used in artificial intelligence computing. Intel’s own attempt at a data center GPU has been a bust; the company admitted on its last call that its recently launched Gaudi GPU chip will fall short of its target of $500 million in revenue this year. Nvidia’s current data center GPU family, known as Hopper, is projected to hit nearly $83 billion in revenue for the fiscal year ending in January, according to consensus estimates from Visible Alpha.

Some think Gelsinger’s exit raises the possibility of some sort of deal—perhaps one that separates Intel’s product and chip design business from its money-losing foundry arm. But that would be extremely difficult since the $7.86 billion Intel is receiving from the U.S. government through the Chips Act requires the company to retain at least 50% ownership of its fabs. And any foreign buyer is unlikely to pass muster with regulators given Intel’s position as the largest U.S.-based chip manufacturer.

Intel, in short, has no easy options, and even few very difficult ones. Chip manufacturing is a complicated process that requires years of research and development on each process and product. Much of the company’s current predicament stems from strategic missteps made well before Gelsinger returned to the company. “We think the potential for a new strategy raises some optimism—but Intel is in a difficult position and the path forward will be difficult no matter the leadership,” wrote Chris Caso of Wolfe Research on Monday in a note to clients.

Even with nearly two-thirds of its market value gone under the last CEO, Intel’s new boss will have a surprisingly tough act to follow.