Intel to Build Two U.S.-Based Fabs

Posted on March 24, 2021 by Paul Thurrott in Hardware, Mobile with 27 Comments

Amidst calls for it to use more efficient third-party silicon fabs, Intel this week announced that it would instead build more of its own. And it’s going to do so in the United States, one of the most expensive places to manufacture electronics in the world. In the meantime, it will also expand its use of third-party fabs.

“We are setting a course for a new era of innovation and product leadership at Intel,” CEO Pat Gelsinger said. “Intel is the only company with the depth and breadth of software, silicon and platforms, packaging, and process with at-scale manufacturing customers can depend on for their next-generation innovations. IDM (integrated device manufacturing) 2.0 is an elegant strategy that only Intel can deliver – and it’s a winning formula. We will use it to design the best products and manufacture them in the best way possible for every category we compete in.”

The announcement came during an online event dubbed Intel Unleashed: Engineering the Future. At the event, Mr. Gelsinger said that his team was “fired up” and that Intel’s best days are ahead. But what onlookers were perhaps most interested in is how Intel intends to compete with fabs that are now delivering 5 nm and 7 nm designs for its competitors when Intel’s own facilities are still stuck on less efficient 10 nm and 14 nm manufacturing processes.

“Our 10 nm delay pushed out 7 nm and ultimately put us on the wrong side of the EUV (extreme ultraviolet lithography) maturity curve,” Gelsinger said, referring to an advanced production process that Intel will use for its next-generation chipsets. “Today, I’m pleased to share that we have now fully embraced EUV, and we’ve rearchitected and simplified our 7 nm process flow, increasing our use of EUV by 100 percent.”

Intel will expand its own manufacturing capabilities in the U.S. and Europe, and it plans x86, ARM, and RISC-based designs. But it will also expand its use of third-party foundries to include “products at the core of Intel’s computing offerings for both client and data center segments” starting in 2023. “This will provide the increased flexibility and scale needed to optimize Intel’s roadmaps for cost, performance, schedule, and supply, giving the company a unique competitive advantage,” the firm says.

On that note, Gelsinger referred specifically to Intel’s “very strong relationship” with ASML, the worldwide leader in EUV manufacturing, and he touted a 2023-era Meteor Lake chipset that will be initially “taped-in” in the second quarter of this year. This chipset is x86-based but it has a new modular design that represents the future of today’s system on a chip (SoC) designs that Intel calls system on a package (so, SoP, I guess).

Intel’s manufacturing expansion represents a $20 billion investment in the future, the company says, and it will create over 3,000 permanent high-tech, high-wage jobs plus more than 3,000 construction jobs and approximately 15,000 local long-term jobs.

Tagged with

Join the discussion!


Don't have a login but want to join the conversation? Become a Thurrott Premium or Basic User to participate

Comments (31)

31 responses to “Intel to Build Two U.S.-Based Fabs”

  1. red.radar

    Wow, High tech manufacturing in the US and Europe. I am rooting for them.

  2. brettscoast

    You have to commend Intel for putting up some serious money investing in future chip manufacturing creating lost of jobs in the process. They have fallen behind AMD and other chip makers so they are playing catch up but the massive investment should be welcomed for what it will bring in the future.

  3. shameer_mulji

    In reply to blue77star:

    "Alder Lake is going to be awesome, in fact Intel super fin 10nm is better than TSMC 7nm and Intel fixed their 7nm and two new fabs are all about 7nm and compared to TSMC it is going to be much better. "

    I'll believe it when I see it.

  4. MoopMeep

    Its interesting that they shut down the local Fab here in Massachusetts a few years ago.

    They could have just upgraded that Fab and those people would of had jobs still.

    • Elwood P Suggins

      In reply to MoopMeep:

      Not all fabs are created equal.

      That fab in Massachusetts was originally built by DEC in the early 1990s - ancient by today's standards. Intel already spent $$$$$ upgrading it once. It's unlikely it would have even been possible to upgrade that facility again to meet today's nm standards - and even if it had, it would probably still have been cheaper to knock it down and start from scratch.

      I suspect it was far cheaper for Intel to just continue building their fabs in Arizona and Oregon where they own plenty of cheap land, have favorable tax treatment, a trained workforce, and all the satellite business infrastructure in place to support their already existing state-of-the-art fabs.

  5. jbinaz

    How will having the fabs here in the U.S. affect the cost of the chips?

    Don't get me wrong, I'm all for anything we can do to reduce dependency on China and other nations that view the U.S. and other western nations as competitors. But can they make them inexpensive enough? Labor costs have to come into play somewhere, right?

    • ecumenical

      In reply to jbinaz:

      Just to be clear, there are already lots of fabs in the US. GlobalFoundries (former AMD and IBM fab facilities) in New York and Vermont, Intel of course in Oregon and Arizona, TSMC in Washington state, and Samsung/TSMC are both looking to build in Arizona/Texas.

      Not to mention that other main locations for fabs are Taiwan, Singapore, and Germany, which aren't necessarily cheaper or underdeveloped compared to the US, to put it mildly.

    • mattbg

      In reply to jbinaz:

      I've always assumed it was primarily labor costs and environmental regulations. I imagine supply chains and where your materials are located and manufactured and how much those materials are subsidized by the country they're manufactured in also factors in (Tesla is using a different battery composition in China in some cars in part because China is subsidizing the production of that particular type of battery).

      Every facility needs employees, but if anything was near-fully automated you'd think it would be semiconductor manufacturing.

      Something that is also creeping up on us on the climate change front is the "supply chain footprint" and how businesses may want to only deal with companies that are seen as clean in their "clean supply chains" to have certain environmental credibility. I believe that's one reason Microsoft and others are starting to talk about zero emissions, etc. If you need your cloud services or software to come from a clean source in order to claim a green supply chain then it makes a difference. I wonder if that has anything to do with it - no idea.

  6. MikeCerm

    Intel's processors are not really competitive with AMD at the high-end or ARM at the low-end of the market, but at least you can get 'em! I know that sounds like a joke, but it isn't. The fact that you can buy Intel CPUs at retail and you literally can't buy AMD's stuff is a real problem. If Intel can finally get their act together and make competitive products with adequate supply while AMD is still fighting for limited fab space at TSMC, Intel could see a big resurgence. I'm actually genuinely excited for for Intel's new 10-nm Jasper Lake "little core" CPUs. Early benchmarks show them performing 1.5x better than Gemini Lake (well into "good enough for most people territory), and if they bring even modest improvements to battery life then it basically boxes ARM out of the laptop market entirely.

  7. Cdorf

    I'm excited for them, I loved the WW episode you had with Intel last week, and they alluded to "Pat will be there so yeah its going to be important." What if the M1 is exactly what the PC industry needed?

    • Paul Thurrott

      Competition is always good, both for the players involve and for the people who buy their products. I present Internet Explorer 6 as the obvious example of what happens when there's no competition.
  8. maddycom

    Wisconsin has this unused FoxConn plant with water from the Great Lakes looking for something to do with it.

  9. avidfan451

    And what of the comments that Intel will actively court Apple with these new fab facilities, a week after their extremely negative ad campaign directed at the M1?

      • avidfan451

        In reply to paul-thurrott:

        Just noticed you didn’t address them in your article, after your in-depth analysis of their recent ad campaign. Seemed like a subject ripe for your thoughts, especially since Gelsinger mentioned Apple again in his Yahoo Finance interview after the Intel Unleashed presser. Not a big deal, just curious what you thought. ??‍♂️

    • avidfan451

      In reply to avidfan451:

      I get both these points; however, this is the CEO announcing the new fabs and stating they’ll court Apple. You can’t tell me the CEO is unaware of the ad campaigns currently run by his company. I do wonder if the ads are long for this world now…

    • samp

      In reply to avidfan451:

      Big tech companies now are so big one part could be cat and dog whilst the other Romeo and Juliet.

      See Microsoft & Google, Samsung & Apple (as Saarek said), Samsung & Sony, Ford & Toyota (on hybrid systems, I suppose that counts as tech), Amazon & Apple, Google & Mozilla, Amazon & Am Ex, etc.

    • Saarek

      In reply to avidfan451:

      No different to when Samsung was stealing Apple iPhone design left right and centre whilst making the chips for their iPhones. Different arms of the same company that essentially bear no relation to each other.

  10. bluvg

    Kudos to Intel for ignoring the outsourcing calls from investors and pundits who never seem to look beyond the next few months. TSMC's capacity appears just about pegged anyway, and Taiwan is potentially a geopolitical liability. The world's current demand requires more capacity, granted new fabs take a lot of time.

    I'm guessing these calls were likely made pre-Gelsinger, but I think he's proving he's the right person at the right time.

    • Paul Thurrott

      Well, they're outsourcing AND building capacity. I assume those 2023-era chips he discussed will be built in some other fab, not in Intel's.
    • solomonrex

      In reply to bluvg:

      What do you mean? They're already outsourcing, it was announced last year. That's what makes this so odd. Who is going to use these fabs if Intel won't? I assume it's a long term play either to undermine TSMC's profits or to build up that side of the business a little for the spinoff.

      • Truffles

        In reply to solomonrex:

        I was listening to a podcast last year where the guest was a chip manufacturing guy who wanted to remain anonymous. He said that TSMC pretty much became a monopolist at the high end of the market after the US imposed a global ban on the supply of chip manufacturing machines to China and banned existing support contracts. This particularly hurt the Europeans where the leading edge companies are currently located.

        The way he described it was that when fabs buy these machines they're pretty much taking a multi-billion dollar punt on a science experiment (several years ago Intel made a wrong bet, while TSMC and others made better bets). Apparently the support contracts are essential because leading edge chip making machines have uptimes measured in hours and are impacted by obscure things such as local variations in gravity. As such, it can take several years for the production lines to be 'tuned' to reliably churn out chips.

        Anyway, the US sanctions meant that Chinese fabs started to lose capacity to supply the low end of the value chain and couldn't creep up the chain by investing in newer machines. 18-months later the lag has caught up and there's a global shortage of chips while Chinese fabs run at very low capacity.

        He said that US companies wouldn't be willing to invest the billions needed to take up the lag because it'll take several years to build and tune the lines during which time the sanctions policy might change overnight, leaving the world with a glut of chip making machines and rendering them unprofitable. Interestingly, he speculated that the only way to change the fundamental business calculation would be if the US government guaranteed to somehow underwrite the risk so that companies like Intel are guaranteed a healthy ROI regardless of sanctions. I'm guessing that that's what has happened.

        Obviously the Chinese government noticed that the US policy was to keep Chinese companies at the low end of the value chain, and China's response has been to directly invest in building a tech stack from the ground up that is entirely local. He was saying that China might reach parity with the future state-of-the-art in around 2030, but it would take only about 5yrs before it can become a high volume and reliable producer of the current generation 5nm chips using a China-only tech stack.

        A lot of the podcast was technical/jargon heavy and got way out of my depth, but above are the key points that I took away from listening.

        • bluvg

          In reply to Truffles:

          "China's response has been to directly invest in building a tech stack from the ground up that is entirely local."

          This has long seemed to me like an obvious unintended consequence of short-sighted sanctions. It may have been inevitable anyway, but it seems likely to be regrettable long-term. The US has no long game, and that problem is baked in structurally.

          • Truffles

            In reply to bluvg:

            I agree, the US has no long game.

            The guy said that China's government didn't have any policy with respect to chip fabs, and were content to let them thrive or die based on market forces just like any other manufacturer had to. That's why China's fab industry had been only slowly growing because they needed to accumulate sufficient turnover to get the commercial loans needed to buy increasingly advanced (and expensive) chip making machines. One company in particular had reached that break-out point and bore the brunt of the sanctions (I didn't recognise the name, but everyone in the podcast seemed to know it).

            Only when the sanctions started did the Chinese government take notice of the industry, and once the full stack decision was made, the engineers and scientists who were underutilised due to the sanctions were moved from production roles to R&D roles and charged with replacing all imported tech from the ground up. At the same time dozens of technical institutes were created to provide the large number of specialists the industry will ultimately need.

      • bluvg

        In reply to solomonrex:

        True, they're already outsourcing some, but I don't think that's their desire. There have been Intel chip shortages, too, and their recent Qtr report showed they're still selling better than most think. This statement seems to be making clear their business strategy. I doubt the primary aim is about blunting TSMC, other than by regaining lost tech leadership.

    • digiguy

      In reply to bluvg:

      If he is the right person, the pandemic might have given them the additional leeway to stir the company and make it competitive again over the next coupld of years before it's too late.

  11. will

    The best thing to kick Intel into gear recently has been Apple. A decade ago it was AMD with X64, now Apple is doing the same with ARM and efficiency.

  12. darkgrayknight

    This is a great start and we need more fabs.

  13. wshwe

    The one edge Intel has is their fab capacity. Last year I bought a gaming laptop with Intel inside. I have no affiliation with Intel.

Leave a Reply