Intel outlines a plan to get back in the game: Pause fab projects in Europe, make the foundry unit an independent subsidiary, and streamline the x86 portfolio

Intel
(Image credit: Intel)

Intel chief executive Pat Gelsinger revealed the company’s next steps this week in the form of three main goals that might turn around the struggling technology company. But tech enthusiasts worry: Will it be enough?

In a recent meeting, Intel leadership discussed how to pursue their long-term strategy. The Board, comprised of independent directors, emphasized the importance of Intel’s efficiency, profitability, and market competitiveness. In the end, they agreed upon three core objectives: enhancing Intel’s foundry operations, optimizing costs to hit the $10 billion savings target, and refocusing on its core x86 technology while driving forward its AI initiatives. 

Notable announcements include a multi-billion-dollar partnership with Amazon Web Services (AWS) and the decision to put fab and packaging facility projects in Germany and Poland on hold for the next two years while continuing to build capacity in the U.S.

Foundry business remains key to Intel's future success

Intel's internal manufacturing remains crucial to the company's long-term success. However, the company plans to make Intel Foundry an independent subsidiary, completing the separation of profit and financial reporting from Intel Products started earlier this year.

This new structure will offer more apparent distinctions for customers and suppliers while allowing Intel to explore external funding and optimize capital strategy for the Foundry and Products units. The leadership team remains unchanged, reporting to the CEO, with an independent board providing oversight to ensure transparency and accountability.

A more focused Intel Foundry will enhance collaboration with Intel Products, while Intel's design and manufacturing capabilities will remain a competitive advantage.

Intel says that its Foundry unit has seen increased customer interest, with a significant uptick in demand for its advanced packaging technologies. Intel has tripled its deal pipeline since the start of the year, and the collaboration with AWS (more on this in a while) is a testament to this growth.

Germany and Poland facilities are on hold; U.S. projects to continue

Intel is also enhancing its capital efficiency within the foundry business. After a period of significant investment in manufacturing, the company is shifting toward a more balanced development pace.

Intel remains committed to its manufacturing projects in the U.S., so the fabs and packaging facilities in Arizona, Oregon, New Mexico, and Ohio will be built according to Intel's plans. It may not be particularly surprising that Intel received billions of dollars from the U.S. government to expand its production capacity in America.

Given the current demand outlook, Intel will pause its fab and packaging facility projects in Germany and Poland for two years as part of the overall business optimization. Intel recently expanded the capacity of its fab near Leixlip in Ireland and expected this facility to remain its lead European hub. The company did not reveal whether demand expectations concern foundry or internal products, but it is obvious that Intel limits the growth of its production capacity in general.

Also, Intel will build the shell for its advanced packaging facility in Malaysia but will not equip it before demand picks up.

Amazon Web Services to use custom Intel Xeon processors

One significant development is Intel's expanded partnership with Amazon Web Services (AWS). Under this multi-year agreement, Intel will produce custom AI fabric chips for AWS using its Intel 18A technology and a custom version of the Xeon 6 CPU based on Intel 3, as planned. In the future, Intel will produce AWS designs using Intel 18A, Intel 18AP, and Intel 14A process technologies. AWS has become Intel Foundry's long-term client, which is a big deal.

The manufacturing and packaging contribute to Intel's ongoing collaboration with AWS, which already uses Intel Xeon Scalable processors (many of which are the so-called off-roadmap semi-custom processors). Intel calls this collaboration a 'better together' strategy, integrating foundry services, infrastructure, and x86 products.

Streamline product portfolio

Intel’s x86 instruction set architecture (ISA) and expertise in its enhancement have been its core strengths for decades, so Intel is certainly not giving up on x86.

Instead, Intel is streamlining its portfolio, including developing competitive x86 offerings across its client, edge, and data center lineups. Intel’s collaboration with AWS exemplifies its commitment to providing customized x86 solutions for customers to better compete against Arm and RISC-V.

The company also sees opportunities to simplify its portfolio to increase efficiency and innovation. To that end, it integrates its Edge and Automotive businesses into its Client Computing Group (CCG), allowing it to expand its leadership in AI PCs and vertical edge solutions. Networking and telco will remain the focus of Intel’s Network and Edge (NEX) group, while its Integrated Photonics Solutions division will merge into the Data Center and AI (DCAI) group.

Building a leaner Intel

These changes are vital to creating a leaner, more streamlined, and efficient Intel by making internal changes and reducing the company's workforce by 15,000.

With efforts underway to cut around 15,000 jobs by year-end, the company has already reached more than half of this target through voluntary retirements and separation packages. In addition, Intel plans to reduce or exit two-thirds of its global real estate by the end of the year as part of its cost-saving initiatives. To bolster its financial position, Intel is also preparing to sell part of its stake in Altera, generating funds ahead of Altera's anticipated IPO.

Anton Shilov
Contributing Writer

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.

  • Marlin1975
    Seems this is more about moving the chairs on the deck as the ship sinks.

    May fool some investor's but does not change anything of substance.
    Reply
  • JamesJones44
    Marlin1975 said:
    Seems this is more about moving the chairs on the deck as the ship sinks.

    May fool some investor's but does not change anything of substance.
    I'm not sure I would call splitting into two companies just moving chairs. Intel sheds the responsibility of producing chips and just becomes a design house just like AMD. Intel Foundry or whatever it ends up being called will not slave to whatever Intel wants.

    How that works out is anyones guess, but it's more than just shifting assets around and saying we're new.
    Reply
  • Jimbojan
    Intel's revenue is better than AMD's, it makes more money than most the companies. The only problem Intel has now is to invest heavily, because the foundry, but the investment is not lost, these investment values will be able to recover when Intel uses its foundry to make the chips for itself and for customers.
    Reply
  • thestryker
    I don't think any of this is meant to be revolutionary, but rather an explanation of what is going on and signs for the future. Intel needed to officially announce an external design win for 18A pretty badly so that's a big deal for the foundry side. While there aren't any huge guarantees with the AWS deal it gives them a good opportunity to show publicly what they can deliver.

    I fully expect "investors" to really push short term hard which tends to undermine long term success so delaying the European fabs as opposed to canceling was a welcome surprise. Building a foundry business requires fab space which Intel doesn't really have enough of, but it also requires customers to fill them. Hopefully the delay doesn't put them too far behind the curve though with ASML's rate of production on High-NA machines it also might not be an issue if they are targeting 14A.

    Moving edge and auto over to client seemed weird, but I think it's probably mostly a sign of the marketing layoffs. It's likely this was solely done for cost cutting measures.

    Hopefully the real estate changes aren't of the "we'll sell the property and lease" variety that idiot investors love pushing for to get those short term gains, but I won't be holding my breath on that one.
    Reply
  • why_wolf
    Basically they are doing the prep work to sell the foundry business. You always spin a division off as fully separate subsidiary first to get the books clean. Then you sell it.

    Short term this might make sense to Wall Street investors. Long term this is Intel giving up on the ability to ever have a technologically superior product to anyone else as they will all be using TSMC which puts the same ceiling on anything they design.

    From a government and industry perspective it is vastly more important that Intel Foundry survives, not Intel Design.
    Reply
  • thestryker
    why_wolf said:
    Basically they are doing the prep work to sell the foundry business. You always spin a division off as fully separate subsidiary first to get the books clean. Then you sell it.
    This is absolutely the way things usually go, but there is something to be said for making the lines very clear from a public perception stance. I hope this is the anomaly where the reason they're doing it is for the perception shift as I don't think it can stand as a leading edge foundry on its own any time soon.

    The Tower deal falling through was a huge blow and the new "12nm" node with UMC won't be online until 2027. We won't know about the full cost/benefit of High-NA machines until probably 2027 at the earliest. There are simply too many variables until probably the next decade.
    Reply
  • bit_user
    I'm really worried what this could mean for their dGPUs.

    Also, it seems like a big lift to keep their datacenter GPUs (e.g. Falcon Shores) alive, given their datacenter AI strategy seems centered around Habana/Gaudi. That leaves Falcon Shores as something of a HPC-specific product, which doesn't seem like a market that's been doing very well for Intel, lately.
    Reply
  • Root Canal
    Intel should first use its own fabs for all its major client and serve tiles. Credibility, customers and revenue will follow.
    Reply
  • joeer77
    Intel took too long to adopt EUV lithography. They pushed FinFETs too long. There is where they went wrong.

    TSMC and Samsung are making gate all around logic chips for some time. Get with the program Intel or get crushed. Start taking risks again. Accept some lower wafer yields. Accept lower margins.
    Reply
  • thestryker
    joeer77 said:
    Intel took too long to adopt EUV lithography.
    This is correct.
    joeer77 said:
    They pushed FinFETs too long...

    TSMC and Samsung are making gate all around logic chips for some time.
    This is not correct as there are zero volume production nodes today using GAAFET.
    Reply