When we sat down to work this morning, starting out the final month of the 2024 year, we did not think that we would be writing about Pat Gelsinger "retiring" from the company that he loves best and most.

Intel's prodigal and now former chief executive officer was one of the few executives in the semiconductor industry who was trained by the company's co-founders Gordon Moore, Robert Noyce, and Andy Grove. And no matter how you cut it, the fact that Gelsinger is leaving Intel does not bode well for the company that he came back to lead during the most serious crisis it has faced since leaving the DRAM memory business in the late 1980s.

This crisis is arguably more severe, which is why Gelsinger left VMware in January 2021 to take the helm of Intel. At the time, in What Gelsinger Can Do To Unscrew Intel, we outlined the parallels from 1987 when Grove ascended to the CEO job and when it exited the DRAM business and when Gelsinger was taking the top job. We are not going to reprise all of those thoughts here.

But what we are going to do is draw parallels to IBM's self-described "near death experience" in the late 1980s and early 1990s, when the mainframe market imploded at the same time that RISC/Unix systems ascended in the datacenter and client/server computing shifted a lot of application work from datacenter systems out to the desktop. This was when John Akers, the consummate mainframer and Big Blue insider, was at the helm and did not read the market forces even close to correctly.

It is not a pretty comparison between IBM and Intel or Akers and Gelsinger.

Back in the late 1980s, when Big Blue was at its absolute peak, the company had over 400,000 employees, with half of them being in the United States. This is when we entered the datacenter market ourselves, and it was an exciting time with maybe 25 different chip architectures in systems and maybe 40 operating systems that had some market penetration. But the inevitable consolidation was coming, and IBM was the biggest and richest target for the competition.

In 1991, IBM laid off around 30,000 employees and booked a $3.4 billion charge, reporting a $2.8 billion loss for the year. In 1992, IBM's revenues shrank a bit and it lost nearly $5 billion and the employee count dropped by close to another 43,000. This left IBM with just above 300,000 people worldwide and an employee base that had shrank by nearly a quarter in two years. In 1993, revenues went down a few more percent and IBM cut another 45,000 employees and posted a whopping $8.9 billion in restructuring charges and an $8.1 billion loss against $62.7 billion in sales. That was it for Akers, who "retired" in April that year and American Express executive Louis Gerstner was brought in by the board of directors to replace him. In his first year (1994), IBM boosted revenue by two points to $64 billion and actually posted a $3 billion profit, but Gerstner continued the layoffs; the employee count shrank by 36,000 to 220,000 worldwide.

All during this time, there was talk of breaking IBM into "Baby Blues," separating its PC, mainframe, X86 server, other proprietary servers (AS/400), Unix servers, middleware and database software, printers, and services businesses into separate companies to give them freedom to maneuver in the market. Gerstner decided - and we think correctly - that keeping IBM together and pivoting strongly to services and getting its software running on other systems was smarter than breaking the company up. IBM bought Lotus for its middleware, sold off low-end printers, and righted its financial ship through huge stock buybacks that rewarded IBMers and Wall Street for growth in earnings per share. (We call this burning the furniture in the house to stay warm.)

We could argue as to whether or not this strategy has worked. But a lot of the IBM that was then is still around now, despite it subsequently selling off DRAM memory, disk drives, PCs, high-end printers, X86 servers, and various services businesses. The mainframe and Power server businesses are relatively healthy and profitable, and IBM owns Red Hat and soon HashiCorp. It is a one possible equilibrium state Big Blue could have engineered.

The question we have - and that the Intel board of directors has been facing - is will we be able to say the same about Intel ten or twenty years from now?

Which brings us all the way back to Gelsinger, who is absolutely as Intel Inside as Akers was True Blue.

If Gelsinger had a health issue that means he cannot fulfill his fiduciary duties as chief executive officer, then Intel would have to say that in its announcement of Gelsinger's "retirement" because it is such a material fact about the state of the company and the strategy that was put into place by Gelsinger several years ago. It doesn't look to us like Gelsinger is tired, much less retired, and that means all we can surmise is that Gelsinger is being shown the door as a repudiation of the IDM 2.0 strategy where Intel revives its floundering foundry business and revitalizes its dependent chip design business at the same time.

Maybe Gelsinger could not make the very deep job cuts and strategy changes that were necessary to get Intel financially healthier sooner - just like Akers could not. Both are too close to the companies they run, and they would have to bear that shame personally. Running companies like IBM and Intel is extremely difficult, and economies and markets change fast. Look at how quickly Nvidia went from interesting HPC accelerator to the largest and most profitable systems business we have seen since IBM's System/360 mainframe in the 1960s and 1970s.

Intel is still an innovator in chip packaging and transistor design, despite its lapses in the past decade, and given a decent manufacturing process, we still think that Intel can create a compelling X86 processor for PCs and servers.

It is important to remember that Intel still has two-thirds of the share of the X86 server market, and the X86 server recession that has plagued Intel and bothered AMD - and that we think was caused in part by a shift in focus on spending for AI systems - is largely over.

But the problem is that the X86 server matters less and less in the datacenter except as a legacy platform for Windows Server applications, which is itself the last proprietary operating system on Earth with volume economics driving it. And it has been shrinking ever so slowly over the past decade as Linux is on the rise and the variety and volume of server CPU platforms based on the Arm architecture that can run it are also increasing. The hyperscalers and cloud builders as well as a number of HPC centers in the world have their own Arm processors. Nvidia has one, and we think at some point AMD might need to blow the dust off its own Arm server chip efforts. Combined, these organizations represent a growing base of server CPU acquisitions, and as we have said before, we can envision a day when AMD, Intel, and the Arm collective all have large and relatively equal shares of the server CPU pie.

This is not a world where Intel can extract huge profits from X86 CPUs, as it was able to do in the 2010s when it essentially had a monopoly on datacenter compute. (Much as Nvidia does on datacenter GPU accelerator compute today, and is benefitting mightily from.)

As we said four years ago, Intel has to decide if it wants to be a foundry or a chip designer. It may not be able to do both and still be Intel. And maybe instead of keeping these two companies together, it does indeed need to separate them so they can go their own ways and either live or die separately.

We think the Intel Products group will have an easier time switching to Taiwan Semiconductor Manufacturing Co as a foundry than the Intel Foundry will have making a dent in TSMC as a secondary source for chips. The forecasts for Intel and TSMC show this - and we did the math back in April showing that TSMC will have an AI chip manufacturing business that is larger than all of Intel Foundry before too long.

Maybe the answer is that TSMC buys the Intel foundry business to get a footprint - a real one - in the United States and a place to carry on once China invades Taiwan. (This gets more and more likely as time passes, and with a Trump administration starting in a few weeks and an expected ramping of the trade war with China, it may happen sooner rather than later.) TSMC would, in this doomsday scenario, keep current Intel processes alive, give Intel employees jobs, and give Intel Products - what we will call Intel if this comes to pass - a foundry for the future as TSMC builds out capacity in the United States.

This is what IBM tried to do in 2014 when it paid GlobalFoundries to take its IBM Microelectronics division that, among other things, etched its Power and z mainframe chips. (This didn't work out so well, with IBM suing GlobalFoundries three years ago and switching to Samsung as its foundry partner, but don't let that scare you.)

The client PC business at Intel is relatively healthy at around $30 billion per year with on the order of $10 billion in operating profits, which is one reason why Michelle Johnston Holthaus, a sales executive from the PC business who was elevated to general manager of its marketing and communications operations in 2017 and who was made general manager of its Client Computing Group in January 2022 by Gelsinger himself, is now CEO Intel Products and co-CEO of all of Intel Corp alongside of chief financial officer Dave Zinsner. Frank Yeary, who was the independent chair of Intel's board, is now the executive chair, and Naga Chandrasekaran, who was named general manager of Intel Foundry in July this year, remains in that role.

The statement put out by Intel does not indicate a change in Intel's strategy, but its leadership, but it also says that Gelsinger retired when we all believe he was asked to leave because the turnaround was not happening fast enough.

"While we have made significant progress in regaining manufacturing competitiveness and building the capabilities to be a world-class foundry, we know that we have much more work to do at the company and are committed to restoring investor confidence," Yeary said in that statement. "As a board, we know first and foremost that we must put our product group at the center of all we do. Our customers demand this from us, and we will deliver for them. With MJ's permanent elevation to CEO of Intel Products along with her interim co-CEO role of Intel, we are ensuring the product group will have the resources needed to deliver for our customers. Ultimately, returning to process leadership is central to product leadership, and we will remain focused on that mission while driving greater efficiency and improved profitability."

Yeary went on to say that Intel would simplify its product portfolio and advance its foundry capabilities. There is no hint of selling off the foundry business. And for all we know, Gelsinger wanted to break Intel up and the current board does not - just like Akers thought about it and Gerstner put the kibosh on the idea.

We don't know.

What we also don't know is how a Trump administration might use executive orders and national security to compel US-based tech giants to use an indigenous US foundry - whether it is owned by Intel, TSMC, any cluster of private equity companies, or maybe even Uncle Sam which could buy Intel Foundry for its own purposes without denting the Federal balance sheet. Without question, the semiconductor business and national security are in jeopardy because Intel dropped the ball on advanced manufacturing processes a decade ago.

It will be very interesting to see who Intel tries to bring in as CEO, and when, and what happens then.