Doctor Lyle Lovett on Intel and the CHIPS Act

For critiques of industrial policy, subsidies, tariffs, the CHIPS Act and the like, many eminent economists are available.

Our favorite is Dr. Lyle Pierce Lovett, Texas A&M ’78, for his sage advice to CEOs everywhere. Just one mistake, one wrong move, advises Dr. Lovett, can turn it all upside down, no matter how good you’ve been so far. (We apologize for the awkward paraphrase but copyright laws on song lyrics are a bear).

Just one mistake. That’s the real argument against government subsidies of any sort, the real argument against giving them and the real argument against taking them. It is not possible for a company to accept significant subsidies and not make decisions around them. And one bad decision can kill a company.

Exhibit A: Intel

Pat Gelsinger returned to Intel (NASDAQ: INTC) as CEO in February 2021, after the company had suffered a series of reversals, including losing the technology leadership of the CPU market to long-time also-ran Advanced Micro Devices, Inc. (NASDAQ: AMD).

The root problem was a breakdown in the company’s manufacturing division. Starting in the mid-2010s, a pattern developed. The company continued to design breakthrough processors but was unable to manufacture them on schedule. It’s move to the “10nm” generation was delayed more than two years, coming into volume production toward the end of 2018. Taiwan Semiconductor (NYSE: TSM) launched its “10nm” generation in 2016. INTC continued to lag, getting to the next “7nm” generation late again with some critics saying it was a minimal advance over the “10nm.”

That February of 2021, Gelsinger faced a choice. Fix manufacturing or get out of the factory business and outsource Intel designs to “foundries” such TSM, the world leader in manufacturing chips designed by other companies. Intel rival AMD had been outsourcing for almost a decade since spinning out Global Foundries in 2009.

Except for Intel (and Samsung (KSC)) every important chip maker in the world—Nvidia (NVDA), Apple (AAPL), Qualcomm (QCOM) Huawei, MediaTek (24540, Broadcom (AVGO) Sony (SONY), Marvell (MRVL), Microsoft (MSFT) and more—outsources production. The big newcomers to chipmaking—Microsoft (NASDAQ: MSFT), Google (NASDAQ: GOOG), Amazon (NASDAQ: AMZN), and Meta (NASDAQ: META)—who now roll (some) of their own to support their data-center/AI business—all outsource.

INTC, the big cheese, stood alone.

In March, a month after becoming CEO, Gelsinger gave his answer. INTC would not abandon manufacturing; it would double down. Gelsinger announced the launch of Intel Foundry Services which would offer its manufacturing prowess to all comers.

The CHIPS Act, offering U.S. semiconductor manufacturers at least $52 billion in subsidies, would not pass until 17 months later in August 2022, though it had been introduced to Congress in 2020. We don’t know whether the prospect of subsidies influenced Gelsinger’s decision in March 2021, but they have since become a bedrock, nay an obsession of Intel’s business planning.

Gelsinger and other INTC execs talk about the CHIPS subsidies all the time. Gelsinger has successfully lobbied politicians to force defense contracts his way rather than to rivals who manufacture overseas. If the company were to abandon manufacturing now, it would leave $19.5 billion in grants and favorable government loans on the table.

Yet the company almost certainly will get out of manufacturing, just not voluntarily. Such a failure has been the manufacturing reboot that INTC likely will be forced out.

To this day, INTC remains two generations behind TSM in manufacturing processes. Just a year after launching its 3nm process, TSM is already getting more than 20% of its revenue from that technology. INTC, which has not reached commercial sales volume with its 3nm product, hopes to do so next year. It has abandoned its “2 nm” plans to go directly, it says, to 18 angstroms. For now, that remains vaporware.

True, INTC’s most exciting chip, its “Gaudi 3” AI accelerator, is manufactured at “5 nm.” A brilliant design, Dr. Robert Castellano of the Information Network, reports that on price/performance, the Gaudi 3 outpaces the Nvidia H100 by almost 3x.

Gaudi 3 is manufactured by… TSMC.

Intel Foundry Services is on track to lose $10 billion this year, having lost some $12 billion across 2022-23. External customers are few, 2023 external revenues were under $1 billion, leaving the company well out of the world’s top ten foundries.

Those cash losses, however, understate the damage.  The relevant number is the $17 billion a year in revenue Intel is not getting compared to revenues of $72 billion for 2019 (before the Covid boom). Those revenues were lost, in part, because customers don’t believe INTC can deliver on the manufacturing side.  That’s a backward move of almost 25% (Using INTC’s peak Covid revenue of $77 billion in 2022, the spread would be $24 billion or almost 30%).

Over the same 2019 to 2024 period, rival and outsourcer AMD’s revenues more than tripled.

Since 2021, when Gelsinger announced INTC would double down on manufacturing:

• Net debt has ballooned from $8.8 billion to $23.7 billion
• Operating margin has collapsed from 25% to negative territory
• Return on Assets has dropped from 12.1% to less than 1%
• Return on Equity was 22.5% and now stands at 0.9%
• Gross Margin has dropped from 55% to 35%; even more devastatingly gross profit has dropped from $44 billion to a run rate of $18 billion.

Noting that potential IFS customers seem reluctant to pump money into a rival, in September INTC announced it would make IFS an independent subsidiary with its own directors. That might boost sales, but we doubt it was the main motive. More likely, INTC is getting ready for a restructuring.

Intel is cutting costs and delaying plans with all the fervor of a firm staring into the abyss. It likely will be split into at least two companies, with IFS spun out entirely, leaving Intel to choose outside foundries on their merits.

The break-up would be costly: not many buyers will lay out cash for a deal that might lower their market capitalization almost as much as getting rid of IFS would raise Intel’s. We expect any deal would leave INTC with at least some of the debt built up by constructing IFS facilities. If it could be confirmed that the $19 billion in government subsidies would go to the independent IFS, that would sweeten the deal.

Any yet… we are still long Intel

Despite, and in a way because of all this foolishness, at the Gilder Technology Report, we still list INTC in our Paradigm Portfolio of recommended stocks. Intel remains a brilliant semiconductor design company. Its recent CPU designs are again challenging AMD for the price/performance lead and sometimes winning. The PC resurgence via distributed AI is in its favor.

Most of all, however, we believe recent failures will force the company to do what it should have done 10 years ago: focus on its strengths. Peter Drucker sagely advised: “don’t solve problems, pursue opportunities.” Kicking and screaming Intel is about to be dragged onto the right side of that line. The sooner that happens, the more likely the stock is to soar.

P.S. COSM Technology Summit, Oct. 31-Nov. 1, Bellevue Hyatt, Seattle, Washington: This conference is sponsored by the Discovery Institute and me. I’ll be one of the speakers and will address the latest technology trends, including the rapid rollout of artificial intelligence and the huge opportunities that will be provided by graphene. To register, go to https://cosm.tech/. Use Promotion Code EAGLE-GG to save $200.