Intel (INTC +1.14%) recently launched Intel Foundry Services, a new division of its internal foundry that will manufacture third-party chips. It kicked off that expansion with a $20 billion investment in two new foundries in Arizona, and it plans to build more plants across the U.S. and Europe.
Intel has engaged with more than 50 "potential" customers for the Foundry Services business so far, and it believes its plants can help alleviate the global chip shortage. The Biden administration's plan to spend $50 billion on expanding America's domestic chipmaking sector could also support Intel's ambitions.
Intel considers this expansion to be a crucial part of the IDM (Integrated Device Manufacturer) 2.0 strategy it unveiled in March. With that plan, Intel's new CEO Pat Gelsinger shut down the idea that the company needed to transition to a "fabless" model like Advanced Micro Devices (AMD 2.33%) and outsource the production of its chips to a third-party foundry.
That's a bold move, but it seems to ignore many of Intel's past and current problems. Let's see why Intel's foundry plans still don't make much sense.
Image source: Intel.
Intel has already lost the process race
In the past, many chipmakers manufactured their own chips. But as chips became smaller and more difficult to manufacture, many chipmakers shut down their capital-intensive foundries and outsourced their chips to major contract chipmakers like Taiwan Semiconductor Manufacturing (TSM +0.42%) and Samsung.
Meanwhile, Intel's own foundry fell behind TSMC and Samsung in the "process race" to create smaller and more powerful chips. Intel fumbled its transition from 14nm to 10nm chips in 2018, then delayed its transition from 10nm to 7nm chips last year. Those missteps resulted in CPU shortages, which enabled AMD, which uses TSMC's foundries, to gobble up Intel's market share in PCs.
TSMC started mass producing 7nm chips and 5nm chips in 2018 and 2020, respectively. Samsung also started to mass-produce 5nm chips last year. Intel doesn't expect to launch its first 7nm chips until 2023.
Intel wants to spin a lot of plates at once
Intel's manufacturing issues were so severe that Gelsinger agreed to outsource the production of some of its CPUs to TSMC earlier this year. Gelsinger also said Intel would engage with other smaller third-party foundries like UMC (UMC 0.91%) and GlobalFoundries.
Image source: Getty Images.
Those moves suggested Intel could divest its foundries, as AMD did with its stake in GlobalFoundries nearly a decade ago, and focus on designing better chips and letting more advanced foundries do the heavy lifting.
However, Intel's IDM 2.0 plan suggests it will only temporarily rely on those foundries until it can solve its own R&D and manufacturing issues. Gelsinger believes doubling down on manufacturing could help Intel eventually restore its lead in the process race.
Yet catching up to TSMC would be extremely difficult, since the Taiwanese foundry already plans to spend more than $100 billion over the next three years on R&D and capacity expansion to maintain its lead.
Intel Foundry Services plans to manufacture a wide range of x86, Arm, and RISC-V chips for third-party chipmakers. This business could generate some extra revenue for Intel, but it's still baffling, for three reasons. First, it's more logical for Intel to dedicate its new foundries to solving its own manufacturing issues instead of rushing to serve other chipmakers. Second, the Foundry Services segment will likely generate lower-margin revenue than its core business. The segment's higher costs will also likely impact its ability to repurchase shares or raise its dividend.
Lastly, Intel seems confident about starting a new foundry business because TSMC and Samsung's plants are fully booked right now. However, TSMC and Samsung are also expanding their capacity to meet the market's demand -- which could convince Intel's "potential" foundry customers to stick with the market leaders instead.
The bottom line
Last year, I believed Intel had two ways to resolve its R&D and manufacturing issues: It could either go all-in on first-party manufacturing or become a fabless chipmaker like AMD. I didn't expect Intel to simultaneously expand its own capacity while providing third-party foundry services to other chipmakers. It's a bold move, but I think it's far too late for Intel to catch up to TSMC and Samsung in the process race -- even if the U.S. government supports its effort with generous government subsidies.
