Former Intel directors believe Intel must split in two to survive
This is the only way, they say.

Four former members of Intel's board of directors believe that the company should spin off its manufacturing operations and assert that this is the only way for Intel to survive. They wrote in a column for Fortune that the U.S. government should use its almost $20 billion in grants and loans pledged federal funding as leverage to force Intel to split into two entities.
With losses over the past 2.5 years amounting to $5.87 billion and tens of billions of lost market capitalization, it is clear that Intel is struggling with its IDM 2.0 strategy to make it a successful chip designer and a thriving chipmaker. Whether the company will make money after cutting costs by laying off some 15,000 employees remains to be seen. Still, David Yoffie, Reed Hundt, Charlene Barshefsky, and Jim Plummer believe that the only way for Intel to survive is to split into two independent entities.
Why split?
They argue that Intel's fabs, engineers, and intellectual property are vital for U.S. national security and technological advancement. Intel is the only American company that manufactures advanced logic on a large scale. Without it, the U.S. risks becoming overly reliant on Taiwan's TSMC (concentrated mainly in Taiwan), which presents major geopolitical and supply chain risks. The former directors stress that the U.S. needs its advanced semiconductor capabilities to remain competitive in areas like AI and defense technologies.
But Intel has lagged in the race with TSMC, losing its lead in leading-edge process technologies in recent years. Furthermore, while companies like Broadcom, Nvidia, and Qualcomm need an alternative to TSMC for manufacturing, they are reluctant to choose Intel Foundry. At the same time, it remains controlled by Intel, their direct competitor. Samsung Foundry faces the same issue as prominent designers like Apple and Nvidia avoid its foundry due to competition concerns.
While Intel's products division remains competitive, Intel Foundry has been bleeding money. Furthermore, Intel's current management still has to prove that they can run contract chipmaking operations, as the company has consistently fallen short in execution, missing deadlines and targets, the ex-directors say.
The former Intel directors believe that if Intel continues to operate this way, it may go bankrupt, which would be a disaster for the national economy and security.
How to split?
The former directors urged Intel to separate its foundry business from its design business, similar to what AMD did in 2009 when it created GlobalFoundries. Intel's product design unit must form a long-term supply contract with the new independent foundry. The former BOD members assert that this arrangement would help stabilize the foundry's operations initially before the manufacturing unit lands orders from third parties.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
The U.S. government has a critical role to play. Through the CHIPS Act, it has earmarked $39 billion to boost semiconductor production in the U.S., including up to $8.5 billion in grants and $11.5 billion in loans for Intel. However, the former BOD members are concerned that Intel's mismanagement could turn it into another high-profile failure like Solyndra. This solar company went bankrupt after receiving significant government funding.
The former directors believe that to avoid such a scenario, the government needs to push Intel to fully separate its manufacturing and design divisions into independent entities. This would not only help Intel's foundry compete but also offer American, Korean, Japanese, and European companies access to a crucial second source for advanced chips, improving global supply chain security.
The ex-directors believe time is of the essence. It took Intel just a few years to fall behind in the semiconductor race, and it could take several years to catch up. If the spinoff is delayed, the U.S. risks falling further behind in the semiconductor industry while TSMC continues to accelerate its lead. Quick, decisive steps from both Intel and the government are needed to secure the future of American semiconductor manufacturing, the former members of Intel's board say.
A major catch
The former directors have a clear scenario for Intel's manufacturing spinoff: the same was true for AMD's split of its foundry unit in 2008 – 2009. However, there is a catch to this example.
GlobalFoundries, which exclusively produced chips for AMD and had a plethora of other customers (to some degree as a result of Chartered Semiconductor acquisition in 2009 – 2010) like Broadcom, NXP, and Qualcomm, was rarely profitable before its IPO (which coincided with the semiconductor demand boom in 2020 – 2021) and had to cease developing leading-edge nodes in 2018 to cut its losses.
It did not help much, as GlobalFoundries had a low utilization rate. Consequently, Mubadala, which still controls most of GlobalFoundries, lost over $22.4 billion on GF from 2009 to 2021. With such an example in mind, it is hard to imagine many investors lined up to take over Intel's manufacturing operations. Also, Intel has more fabs and more projects than AMD or GloFo ever had, so its foundry division will probably make astonishing losses if it becomes genuinely independent, particularly if Intel products company decides to produce more chips at TSMC.
Of course, it may not be allowed to do so by an exclusive wafer supply agreement, but in this case, its products may end up being non-competitive with products from AMD, Nvidia, and Qualcomm made by TSMC. In this case, even Intel's products company will become unprofitable. Also, since Intel will likely remain the majority owner of the new Foundry company, those losses will be attributable to Intel.
Anton Shilov is a contributing writer at Tom’s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.
-
waltc3 FABs are hugely expensive endeavors. nVidia and Apple, not to mention AMD, are fabless. All are booming. Intel doesn't FAB for them, AFAIK, for obvious reasons. The biggest upside to US TSMC is that even if China decides to try and nationalize Taiwan, TSMC could survive and prosper in the US. Bringing manufacturing industries back to the US was always a no-brainer, and props to the former president's administration for making it happen and initiating the process.Reply -
bit_user
TSMC has fabs around the world, including China, Japan, and South Korea. Their first US fab was in Washington State, dating back to 1996. They're also building more in Japan and Europe. Samsung also had a fab in the US, dating all the way back to 1996.waltc3 said:Bringing manufacturing industries back to the US was always a no-brainer, and props to the former president's administration for making it happen and initiating the process.
Looking at US-based manufacturers: Intel, Micron, and Global Foundries all have fabs around the world. I'd say there's always been a lot of geographical diversity in the industry, but I think the CHIPS legislation was a good move to focus new investments in the US. -
Notton IMO, Intel and Samsung Foundries don't sell well, not because of competition issues, but because they sell a lame product.Reply
For Intel they can't seem to do a power efficient design. You don't want your ASIC, IoT, ARM chip running hot and power hungry, that's a tough sell.
For Samsung, it's yield rate, clock speed, and power efficiency. It's so bad only a handful of ASICs for mining even bother with ordering. -
thestryker
Intel's only DUV node that utilizes industry standard tools is Intel 16. This means relying on Intel to do the process design work for the rest which is a non-starter for pretty much anyone doing volume fabrication.Notton said:For Intel they can't seem to do a power efficient design. You don't want your ASIC, IoT, ARM chip running hot and power hungry, that's a tough sell.
Intel 4/3 and 20A (canceled)/18A also use industry standard tools, but Intel 4 and 20A aren't full nodes so they'd be of limited use in the first place. 18A isn't going to be available for usage until some time in 2026 and is the first node Intel has been pushing for external customers. Intel 3 should also be available, but I'm not sure what their volume availability looks like here though it should improve over time.