Is There A Black Swan Ahead For Intel?

Intel Seems To Be In Deep Trouble

Note: Beyond The Hype Is publishing a speculative thesis only because it has the potential to be a black swan event for Intel. With the risks involved, Intel investors need to tread carefully and consider appropriate risk strategies.

The last few months have been interesting and full of negative developments for Intel (INTC) – more so than what has been usual lately. The Company is going through several things that can be seen as a “first” for Intel – and not in a good way. Diving right in, here are the three major developments:

Intel is getting destroyed in servers and the client side is not looking very good

Advanced Micro Devices (AMD) has been steadily gaining share against Intel in the server space, but few would have forecast the ignominy that was the Xeon 6 launch. Intel’s Sierra Forest and Granite Rapids server processors were launched to almost no fanfare. For a couple of years now, Intel’s Pat Gelsinger has been pitching these products as the turning point and foreboding the reversal of market share losses to AMD. But the reality looks quite different. Not a single major hyperscaler was present at the launch of these next generation server chips. At Computex, servers with Xeon 6 chips were largely invisible except to the extent that they got paired with Nvidia’s (NVDA) Hopper and Blackwell systems. Not one server maker at Computex the Beyond The Hype visited had a public display of Blackwell systems with an AMD CPU. As discussed previously, Nvidia is using Intel CPUs instead of AMD CPUs in Hopper and Blackwell reference systems for strategic reasons. Nvidia sees AMD as a GPU competitor and is far less concerned about Intel’s GPU efforts. While Nvidia’s strategic move helps Intel server CPU demand, the bottom line is that Intel’s Xeon 6 launch is foreboding even more massive market share losses to AMD in the server domain. Intel data center prospects look quite poor for H2 and beyond.

The client side, where Intel has effectively defended market share for the last few years, is not looking that good either. Intel’s AI PC offering is running months behind AMD’s Qualcomm’s (QCOM) comparable efforts. Given the timing of the holiday season, even an “on-time” late-Q3 arrival of Lunar Lake followed by Q4 arrival of Arrow Lake makes it risky for OEMs to be adopting these chips for the 2024 holiday season. More so considering the problems Intel has been having with Raptor Lake lately. Would OEMs want to risk a booming holiday season with new Intel chips when Intel has been unable to fix a year plus old Raptor Lake bug? (more on this later).

The combination of Raptor Lake problems and late arrival of Lunar Lake and Arrow Lake makes it likely that Intel will lose big chunks of client market share in H2 and 2025.  

All-TSMC Lunar Lake and mostly-TSMC Arrow Lake Hit Margins

It is now well known that Intel under its previous CEO Bob Swan negotiated with TSMC (TSM) to build leading edge Intel x86 CPUs using the N3 process. Based on public information, it appears that all the tiles on Intel’s upcoming Lunar Lake CPU will be made at TSMC. Arrow Lake, which is expected to follow Lunar Lake, will have most of the tiles made at TSMC but Intel may manufacture tiles for a few of the lower end SKUs using Intel 20A or 18A process. If the information about Intel using its own processes for the lower-end products true, this should be troubling to Intel investors since the implication is that Intel sees TSMC N3 as superior to 20A or 18A.  

Setting aside questions about process superiority, we can be assured that TSMC is not giving the best pricing to Intel given that Pat Gelsinger has constantly presented TSMC as competition and as a transient supplier. In the near term, moving an estimated $12B worth of chip production to TSMC means about $12B of business that Intel fabs do not get. The resulting underutilization at Intel will not be pretty and is likely to keep margins depressed for much of 2024 and 2025.

When one combines the TSMC cost structure with expensive packaging, it is all but certain that Intel would have a high-cost CPU SKU with Arrow Lake (especially on the desktop front). AMD, with a superior cost structure, will be able to take its Zen 5 to price points Intel will not be able to reach except at atrocious margins. This dynamic makes Intel desktop client an especially vulnerable segment for Intel even in 2025.