US Govt's sluggish Chips Act payouts slam the brakes on Samsung's fab — company delays mass production at Texas fab to await further CHIPS funding

Samsung's Semiconductor Plant in Pyeongtaek, South Korea
Samsung Electronics Pyeongtaek Line 1, a South Korean fabrication plant focused on the production of fourth-generation V-NAND (64 layers) chips, seen during its opening in 2017. (Image credit: Samsung)

Samsung is delaying the start of mass production at its Texas foundry to 2025, according to a report by Business Korea. The new fabbing plant, which was originally set to begin mass production in the second half of 2024, is now expected to have only limited production by that time. Samsung is apparently scaling back its Texas operation due to uncertain financial factors, including CHIPS Act subsidies and the global economy.

The Korean conglomerate will invest $200 billion in Texas alone, with 11 total foundries to produce 4nm chips, the first one being built at Taylor, Texas. The Taylor fab has been delayed, much like TSMC's Fab 21 in Arizona. But the delay at Samsung's first fab is apparently intentional, with the corporation deciding to postpone mass production in favor of a smaller output levels until 2025. In contrast, TSMC's foundry hit conflicts with local workers and unions.

For comparison, Samsung's brand-new 4nm foundry at Pyeongtaek in South Korea is capable of producing 28,000 wafers a month. Output from the Taylor, Texas, fab is said to be 5,000 wafers per month-- about a sixth of what the Pyeongtaek facility is capable of.

Business Korea claims that finances are a key concern of Samsung, especially when considering CHIPS Act subsidies and the state of the global economy. The CHIPS and Science Act is supposed to grant subsidies to semiconductor companies like Samsung to encourage the construction of foundries in the U.S. However, these subsidies are still largely in the pipeline, with just $35 million of the total $52 billion granted so far.

Samsung is also concerned that even when subsidies are finally released, Intel could receive the lion's share, with a report from last month alleging Intel was set to receive up to $4 billion early. The U.S. arm of Samsung is apparently lobbying politicians to distribute the funds more equally and not favor Intel, arguing that Samsung has only invested so much in Texas because it trusted that the CHIPS Act would go into effect.

The health of the global economy is also on Samsung's mind. Although the U.S. has seemingly achieved a so-called soft landing and has avoided a true recession, other parts of the world aren't so lucky. Samsung is still reeling from low SSD and RAM prices, even though revenue is finally on the rise. The PC market is forecasted to finally recover to 2020 levels in 2025, but the high peaks of 2021 are apparently well in the past. All of this just makes Samsung tepid about its $200 billion investment in the U.S., which could spell success or disaster for the company.  

Matthew Connatser

Matthew Connatser is a freelancing writer for Tom's Hardware US. He writes articles about CPUs, GPUs, SSDs, and computers in general.

  • SemiChemE
    Doesn't Samsung already operate a fab in Austin, Texas? Sure, it's older technology (65nm down to 14nm), but it should still count as a fab. So, unlike TSMC, Samsung already knows how to build and operate a fab in the U.S. This seems to be more of a response to recent market forces (Semiconductor slowdown since 2H22), and possibly some lobbying to get a larger share of Chips Act money.
    Reply
  • JamesJones44
    I'm not so sure the US has made a soft landing yet. There is a lot of soft areas in the economy and we are in the time of year where if you have budget you are going to spend it or lose it. The real test will be Q1 2024 and we won't know until Q2 2024 if there was a slow down after the start of the new year. If we get past those gates without a recession then I would consider the US to have engineered a "soft landing".

    Samsung is right to be cautious on the economic/government outlook until we can see clearly through the clouds of gov stimulus and monitory tightening.
    Reply
  • phead128
    SemiChemE said:
    Doesn't Samsung already operate a fab in Austin, Texas? Sure, it's older technology (65nm down to 14nm), but it should still count as a fab. So, unlike TSMC, Samsung already knows how to build and operate a fab in the U.S. This seems to be more of a response to recent market forces (Semiconductor slowdown since 2H22), and possibly some lobbying to get a larger share of Chips Act money.
    TSMC has a fab in Washingston state (TSMC Fab 11) since 1996, the first pure-play foundry fab in the US.

    https://www.wafertech.com/en/about/index.htmlhttps://www.oregonlive.com/silicon-forest/2023/12/tsmc-finally-puts-its-name-on-its-aging-northwest-chip-factory.html
    So TSMC has been building and operating fabs in US for decades before Samsung even arrived. TSMC had admitted it knows full well it's US fabs will never be as productive as Taiwanese fabs, due to work culture difference. To act like TSMC is some newbie is disingenuous.
    Reply
  • kjfatl
    It is likely that Samsung is delaying the fab as a result of equipment availability delays. Intel, in particular has filled the pipeline of fab equipment suppliers. By delaying a few months, or even over a year, Samsung may end up with a better, more profitable fab.
    Reply
  • Joseph_138
    SemiChemE said:
    Doesn't Samsung already operate a fab in Austin, Texas? Sure, it's older technology (65nm down to 14nm), but it should still count as a fab. So, unlike TSMC, Samsung already knows how to build and operate a fab in the U.S. This seems to be more of a response to recent market forces (Semiconductor slowdown since 2H22), and possibly some lobbying to get a larger share of Chips Act money.
    14nm isn't good enough. That fab may be able to produce chips for low and medium tech devices, but for the high tech devices, that we depend on, we need the newest fab processes, like 7 and 8nm, and we don't have those.
    Reply
  • djcalligraphy
    "...slam the BREAKS on Samsung's fab"Wouldn't that be "brakes?"
    Reply
  • thisisaname
    Governments may be quick to promise cash but they are not quick to hand it over.
    Reply
  • SemiChemE
    phead128 said:
    TSMC has a fab in Washingston state (TSMC Fab 11) since 1996, the first pure-play foundry fab in the US.

    https://www.wafertech.com/en/about/index.htmlhttps://www.oregonlive.com/silicon-forest/2023/12/tsmc-finally-puts-its-name-on-its-aging-northwest-chip-factory.html
    So TSMC has been building and operating fabs in US for decades before Samsung even arrived. TSMC had admitted it knows full well it's US fabs will never be as productive as Taiwanese fabs, due to work culture difference. To act like TSMC is some newbie is disingenuous.
    True, but the TSMC fab is a very old 200mm fab that hasn't been significantly upgraded in over a decade. The Samsung fab is a state-of-the-art 300mm facility and one of the largest fabs in the U.S.
    Reply
  • SemiChemE
    Joseph_138 said:
    14nm isn't good enough. That fab may be able to produce chips for low and medium tech devices, but for the high tech devices, that we depend on, we need the newest fab processes, like 7 and 8nm, and we don't have those.
    I never said Samsung shouldn't be building a new fab, just that they know how to operate a high-end state-of-the-art fab in the U.S. Note, that 14nm was state of the art until TSMC rolled out 7nm in 2018 (just over 5 years ago). Further, 28nm is still even today, the highest volume node in the industry due to the significantly lower costs compared to finfet. The high-end 7nm and beyond nodes are really only needed for very high-end, products and the retail volumes needed to justify such a high-end fab, aren't there. That is why the Texas fab is so small (60K wafers per year is tiny, tiny). That's why I'm arguing that the Samsung slow down is due to market forces, rather than slow Chips Act payouts as the article implies.

    Most of those outside the industry, don't seem to understand how the skyrocketing design costs of Finfet really caused most of the industry to stall at the 28nm node. The lower geometry nodes can only be justified either for very high-end niche products (eg. server chips & GPUs), where customers are willling to pay a large premium, or for extremely high-volume products (like cellphone chips from Apple and Qualcomm), where steep design costs can be spread across billions of units. Unfortunately for Samsung, TSMC has locked down most of the high-volume business, which leaves Samsung pursuing the niche applications in a small fab.
    Reply
  • phead128
    Joseph_138 said:
    14nm isn't good enough. That fab may be able to produce chips for low and medium tech devices, but for the high tech devices, that we depend on, we need the newest fab processes, like 7 and 8nm, and we don't have those.
    90% of the semiconductor market is <14nm and below. Most of your everyday items like automobiles, jet fighter planes, etc...use legacy nodes. Only iPhones or mobile phones things that need exceptional battery life need 7-8nm.
    Reply