Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/analyst-insight-why-intel-foundry-matters-more-than-you-think.23937/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2030770
            [XFI] => 1060170
        )

    [wordpress] => /var/www/html
)

ANALYST INSIGHT: Why Intel Foundry Matters More Than You Think

Daniel Nenni

Admin
Staff member

An Intel Xeon 6+ server chip, code-named Clearwater Forest (Credit: Anshel Sag)

There’s been no shortage of talk about Intel’s prospects as a company, with even more talk about the future of Intel Foundry. This is especially true after the official announcement of the Intel 18A process node ramping production, which is the culmination of Intel’s five nodes in four years (5N4Y) strategy to return the company to process leadership and reintroduce its foundry business. Intel 18A is finally coming to fruition in both client and edge products (Intel Core Ultra Series 3 — code-named Panther Lake) and datacenter processors (Clearwater Forest).

As I explained in my analysis of the Panther Lake processors built on the 18A node, Intel Product isn’t just building the CPU tiles at Intel, but also the GPU tiles for the majority of the SKUs. This is a drastic departure from only a year ago, when the Lunar Lake family of chips was manufactured predominantly at TSMC and packaged by Intel. Being able to shift most production for both the compute and GPU tiles away from TSMC not only gives Intel the scale it needs for its foundry side, but also eventually creates considerable cost savings, which in turn fosters better margins to help fuel profitability — all while delivering a competitive product.

Packaging is becoming another interesting area within the semiconductor industry as chiplets grow in popularity because they enable better yields and larger-scale chip designs with more memory and compute per chip. Intel has a long track record in advanced packaging semiconductor products going as far back as 2019. Back then technologies like Foveros enabled low-power chiplet designs that Intel had previously been unable to deliver. Foveros is still a keystone technology that enables Intel’s Panther Lake to be flexible and change out different tile configurations as necessary based on the application. Intel’s Clearwater Forest also implements the new Foveros Direct, which uses copper hybrid bonding over microbumps to allow for 3-D stacking of chips. Intel also has EMIB technology for mostly datacenter chips; it uses this in addition to Foveros Direct in the new Clearwater Forest chip to interconnect the multiple 18A dies together. I’ve had the opportunity to see Intel’s packaging prowess on display at its facility in Malaysia, which it has been expanding to satisfy increased demand.

What About the Rest of the Ecosystem?​

Intel 18A and products such as Panther Lake that use it are exciting for the rest of the ecosystem as well. When Intel offers a more competitive product that supports better battery life and performance, it does so at a lower cost to the OEM. Obviously, Intel Product will capture a good portion of the margin on the chips when selling them to OEMs, but the relationship between Intel and its OEM partners is symbiotic; when Intel delivers competitive products in volume, that ultimately benefits the entire ecosystem — and consumers, too. In the bigger picture, a competitive and healthy silicon ecosystem for PCs delivers better products at more competitive prices.

Going beyond PCs, there are also benefits for the smartphone industry when Intel Foundry is successful and competitive. While there aren’t any Intel Foundry mobile customers to announce yet, there is still a considerable opportunity for the mobile industry to leverage Intel Foundry to lower costs and diversify the supply chain. Currently, the vast majority of the smartphone industry manufactures its chips using TSMC. Apple, for example, manufactures 100% of its chips with TSMC, and if you add Qualcomm and MediaTek into the equation, you are probably looking at something in the ballpark of 90% of global smartphone SoCs produced by TSMC. Samsung is the next closest competitor to TSMC, although most of its semiconductor business focuses on solid-state memory and storage and has little to do with smartphones. That said, Samsung has recently transitioned to using more of its own SoCs in its smartphones, including the entire Galaxy Z Flip7 line, of which it expects to ship millions during the next year.

Offering a Competitive Alternative to TSMC​

Intel Foundry entering this market with a competitive process node — like its hotly anticipated Intel 14A process, next up on the roadmap after 18A — could also prevent TSMC from continuing to raise its prices by double-digit percentages, which we have seen it do almost every year for the last five years. Many of these price increases aren’t public, so it’s hard to nail down exact figures, especially as they vary from one customer and process node to another. But as each node becomes more expensive to produce and the size and cost of these foundries continues to go up, so do the prices. Unless TSMC — or a competitor — offers some relief, that cost eventually gets passed along to the chip vendor, the OEM, and ultimately the consumer. That means more expensive phones, PCs, TVs, and even cars.

What’s Next for Intel Foundry?​

While Intel Foundry has made its current customer list for 18A clear, there is still a chance that other chip companies could come along and take advantage of Intel’s advanced foundry capabilities, whether that means manufacturing dies or packaging chips using dies from multiple foundries. Intel has already proven that it can mix and match dies from TSMC and Intel on its own chips within its packaging facilities.

So, although Intel Foundry is still working to attract new customers, if Panther Lake and Clearwater Forest prove to be the successful and competitive products that I think they will, there are sure to be companies that show more interest in Intel Foundry to take advantage of a winning formula. Today, that’s still somewhere down the road. But I believe that a combination of need for cheaper and lower-power chips and a desire for a geopolitically diverse supply chain for chips will help drive more business towards Intel Foundry and its advanced process nodes, whether it’s for 18A, the forthcoming 18A-P, or 14A a few years down the road.

 
Back
Top