Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/us-takes-security-first-focus-in-doling-out-39-billion-chip-aid.17486/
        )

    [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] => 2021370
            [XFI] => 1050270
        )

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

US Takes Security-First Focus in Doling Out $39 Billion Chip Aid

I think one of the "struggling" chipmakers she was talking about is Intel.
I could also see Micron and maybe GF (pending how sticky their nodes are now that shortages are easing) being in this boat too. Either way this is further evidence that CHIPs not meant to prop up losers. Micron and Intel (as well as TI, Samsung, TSMC, Global wafers, ect) will get subsidies for investing not for existing. Although I wouldn't be surprised if firms with R&D/pay taxes in the US will get slight preferential treatment for reasons that should be painfully obvious.
 
I could also see Micron and maybe GF (pending how sticky their nodes are now that shortages are easing) being in this boat too. Either way this is further evidence that CHIPs not meant to prop up losers. Micron and Intel (as well as TI, Samsung, TSMC, Global wafers, ect) will get subsidies for investing not for existing. Although I wouldn't be surprised if firms with R&D/pay taxes in the US will get slight preferential treatment for reasons that should be painfully obvious.

The ability to provide chips for DoD, DoE, CIA and any important national security related applications are probably the number one priority to US government, according to Secretary Raimondo's talking. Essentially she is saying those 3-letter agencies need real products to bring fighter jets, missiles, nuclear bombs, and satellites to the real world. They do care but are not interested too much about chip makers' roadmaps, 5-year turnaround plans, business visions, or theories.
 
The ability to provide chips for DoD, DoE, CIA and any important national security related applications are probably the number one priority to US government, according to Secretary Raimondo's talking. Essentially she is saying those 3-letter agencies need real products to bring fighter jets, missiles, nuclear bombs, and satellites to the real world. They do care but are not interested too much about chip makers' roadmaps, 5-year turnaround plans, business visions, or theories.
If that was the case TSMC and Samsung would be totally unsuitable for this purpose. Yeah the manufacturing would be onshored, but the IP would be in the hands of foreign nationals. If say the ROC is conquered by the PRC, then that IP would be in PRC hands (a far worse scenario than just the supplier being taken out of commission). At that point why bother onshoring at all? If the DOD was only reason for CHIPs, then only US firms would get the money and no matter how bleak things start looking for intel and micron the DOD would prop them up.

If instead the purpose is to build up a stronger ecosystem, more resilient supply chains, and create a new growth engine for the USA; then giving funding to all and ensuring a wide variety of companies across the semiconductor manufacturing chain makes sense.
 
If that was the case TSMC and Samsung would be totally unsuitable for this purpose. Yeah the manufacturing would be onshored, but the IP would be in the hands of foreign nationals. If say the ROC is conquered by the PRC, then that IP would be in PRC hands (a far worse scenario than just the supplier being taken out of commission). At that point why bother onshoring at all? If the DOD was only reason for CHIPs, then only US firms would get the money and no matter how bleak things start looking for intel and micron the DOD would prop them up.

If instead the purpose is to build up a stronger ecosystem, more resilient supply chains, and create a new growth engine for the USA; then giving funding to all and ensuring a wide variety of companies across the semiconductor manufacturing chain makes sense.

TSMC has been making several important chips for US military (such as F35 stealth fighter jets) or space related application in Taiwan for many years already. To bring TSMC leading edge nodes manufacturing capability to US is a logical next step.
 
TSMC has been making several important chips for US military (such as F35 stealth fighter jets) or space related application in Taiwan for many years already. To bring TSMC leading edge nodes manufacturing capability to US is a logical next step.
My question is how does this make US technology more secure? I don't buy the argument that now the DOD doesn't need to worry about the collapse of the ROC. TSMC AZ will in the best case only be able to maintain the nodes they currently produce at that site without the rest of TSMC. As I mentioned there is also the risk of DOD IP/designs falling into the hand of the PLA. That happening is FAR worse than getting their future capacity cut off. In short in my opinion if all that mattered was securing DOD technology/supply you would go to intel or GF and even if the chip needs to be 2x more expensive and some huge die monstrosity you would do that. This solution is also FAR FAR FAR cheaper than the $52B for CHIPs since the DOD could just leverage existing capacity. Those two firms are the only two that could provide guaranteed capacity, a guaranteed roadmap, and guaranteed security for their IP.
 
My question is how does this make US technology more secure? I don't buy the argument that now the DOD doesn't need to worry about the collapse of the ROC. TSMC AZ will in the best case only be able to maintain the nodes they currently produce at that site without the rest of TSMC. As I mentioned there is also the risk of DOD IP/designs falling into the hand of the PLA. That happening is FAR worse than getting their future capacity cut off. In short in my opinion if all that mattered was securing DOD technology/supply you would go to intel or GF and even if the chip needs to be 2x more expensive and some huge die monstrosity you would do that. This solution is also FAR FAR FAR cheaper than the $52B for CHIPs since the DOD could just leverage existing capacity. Those two firms are the only two that could provide guaranteed capacity, a guaranteed roadmap, and guaranteed security for their IP.

1. DoD and DoE did ask those US suppliers (who use TSMC's fabs in Taiwan) to find US fabs to manufacture their chips. But for various reasons, they found it won't be practical or feasible.

2. DoD and DoE are very experienced in bringing foreign suppliers' products and manufacturing capability to the US soil. M777 howitzer (from UK), Naval Strike Missile (from Norway), Constellation-class frigate (from Italy and France) are just some examples. It's a very practical and cost efficient way to bring needed TSMC capability to US to secure the supply. DoD and DoE have little interest to build a new supply chain with also-run players to replace an existing, efficient, and capable supply chain.

3. Once the DoD/DoE chips are made in (and only in) TSMC Arizona, all the security and secrecy procedures can be implemented and enforced. It is not much difference than bringing suppliers from UK, Norway, France, or Italy to produce products on US soil.

4. DoD and DoE have many expensive yet critical projects that need semiconductors. They can't jeopardize those project schedules by using questionable vendors. For example the US Nuclear Weapons Modernization is a 30 years $1.5 trillion project that is crucial to US national security and safety. In this project the computer simulation done by the supercomputers is the must because US had stopped any nuclear explosion test long time ago. There are several DoE supercomputer construction projects, among them:

a. Aurora Supercomputer at Argonne National Lab. HP Enterprise, based on Intel CPU and GPU.
  • Announced in 2015, was planned to be operational in 2018 but delayed multiple times, go live postponed to 2023 (hopefully?)
b. Frontier Supercomputer at Oak Ridge National Lab, HP Enterprise, based on AMD CPU and GPU.

Announced in 2019, go live in 2022, ranked number one in World Top500 supercomputer 2022 list. The world's first exascale supercomputer.

I believe a project manager at National Nuclear Security Administration (NNSA) can tell which vendor he or she can trust more.
 
Last edited:
b. Frontier Supercomputer at Oak Ridge National Lab, HP Enterprise, based on AMD CPU and GPU.

Announced in 2019, go live in 2022, ranked number one in World Top500 supercomputer 2022 list. The world's first exascale supercomputer.
Almost certainly not the world's first exascale supercomputer. China is rumored to have at least two operational exascale systems.


The TOP500 list only includes publicly announced systems. The US government is rumored to have multiple "dark" exascale or near-exascale systems operational, which are probably top secret. The real issue with any of these scale-out / accelerator-based systems is software, so I'm skeptical that LINPACK results are all that telling anyway.
 
Almost certainly not the world's first exascale supercomputer. China is rumored to have at least two operational exascale systems.


The TOP500 list only includes publicly announced systems. The US government is rumored to have multiple "dark" exascale or near-exascale systems operational, which are probably top secret. The real issue with any of these scale-out / accelerator-based systems is software, so I'm skeptical that LINPACK results are all that telling anyway.

Yes, that's possible but unverifiable. CCP has a solid track record to fake the achievement at any cost regardless it's money or human life.
 
Without seeing the application itself and being a lawyer qualified to dissect the thing, it looks like things are being well thought out. Foxcon Wi or government welfare these do not seem to be.
 
Without seeing the application itself and being a lawyer qualified to dissect the thing, it looks like things are being well thought out. Foxcon Wi or government welfare these do not seem to be.
As you could tell from my previous posts, I just don't like this Act. It does nothing to address a bunch of fundamental differences in construction costs and inefficiency between the US and Taiwan. It does nothing to address permitting inefficiencies. It includes a boat-load of expensive and (IMO) useless bureaucracy. It does nothing to level the playing field with China and Taiwan WRT taxation, especially the write-off of capital expenditures. And it includes some very tangled rules on stock buybacks, and I'm guessing dividends as a percent of net income will be a political factor. And DoD chip volumes are so low compared modern efficient fab capacities that calling this ACT "fundamentally a national security initiative" is silly.

I think the US does have to subsidize semiconductor manufacturing if we want it here, but I think this Act is going about it the wrong way.
 
As you could tell from my previous posts, I just don't like this Act. It does nothing to address a bunch of fundamental differences in construction costs and inefficiency between the US and Taiwan. It does nothing to address permitting inefficiencies. It includes a boat-load of expensive and (IMO) useless bureaucracy. It does nothing to level the playing field with China and Taiwan WRT taxation, especially the write-off of capital expenditures. And it includes some very tangled rules on stock buybacks, and I'm guessing dividends as a percent of net income will be a political factor. And DoD chip volumes are so low compared modern efficient fab capacities that calling this ACT "fundamentally a national security initiative" is silly.
Valid concerns. I was speaking from the perspective of if this is what CHIPs, it seems like the US GOV is putting guardrails in place so that this doesn't turn into something that squanders the people's resources.

I think the US does have to subsidize semiconductor manufacturing if we want it here, but I think this Act is going about it the wrong way.
Besides the permitting thing what would you do instead? The ROC/ROK offered and still offer generous tax credits on equipment and cheap guaranteed financing, wouldn't the US need to offer even better incentives to attract folks like TSMC, Micron, SK, and Samsung to build new greenfield fabs in a region which they have no (TSMC/SK) to little (Micron/Samsung) presence?
 
Besides the permitting thing what would you do instead? The ROC/ROK offered and still offer generous tax credits on equipment and cheap guaranteed financing, wouldn't the US need to offer even better incentives to attract folks like TSMC, Micron, SK, and Samsung to build new greenfield fabs in a region which they have no (TSMC/SK) to little (Micron/Samsung) presence?
Personally, I don't think TSMC needs any incentives at all. I think they're incentivized by customer demand to reduce geopolitical risk. IMO, I don't see that changing for the foreseeable future. And I still hold the position that just building fabs in the US doesn't really reduce TSMC's geopolitical risk. I don't think the AZ fabs can be just in a way nationalized (put into a US-based company) and expected to continue producing and evolving without the ROC infrastructure.

Along the same lines, I'm not sure what to think about Samsung. In theory their Korean fabs are even more vulnerable than TSMC, and threatened by a more bellicose regime than Taiwan is. I am curious about what Samsung HQ thinks about its Texas operations. Not one peep out of them about the cost of building a fab in Taylor, TX versus one in South Korea.

If we really want fabs in the US, my opinion is:

- No-cost land leases for semi manufacturing.
- Equalizing total US tax rates (federal, state, and local) with Korea and Taiwan, even if it means national emergency legislation.
- Enact equal to the global most accelerated depreciation on semi equipment, which might be one of the most important factors for fabs.
- Expedited federal permitting that overrides state and local laws.
- Incentivize US universities to create degree programs most in need by the semi manufacturing industry.
- Create a national scholarship program for qualified students to enter the those degree programs. Require rigorous testing to get a scholarship.
- Expand the NSF grant level for semiconductor research in US universities.
- And one I go back and forth about, create a tax incentive for companies to use US-manufactured chips.

My view - if we're serious, let's get serious.
 
Back
Top