Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/how-america-can-avert-the-next-chip-shortage.15477/
        )

    [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
)

How America can avert the next chip shortage

Daniel Nenni

Admin
Staff member
Thoughts on this would be greatly appreciated:

American semiconductor production has arrived at a watershed moment.


Congress is on the verge of taking an historic step to fuel U.S. leadership in the research, design, and manufacturing of microchips, a technology at the heart of America’s economy, national security, and critical infrastructure. The U.S. Innovation and Competition Act (USICA)—which the Senate has passed—would provide $52 billion in chip research investments and incentives for companies to establish manufacturing in the United States. New House legislation called the America COMPETES Act, expected to be considered this week, would do the same.

In essence, the legislation would level the global playing field and ensure more of the semiconductors our country needs are made on U.S. shores, while also turbocharging domestic chip innovation through critical research investments.  

The trendlines of the past few decades show why Congress and the White House must act swiftly. Just 12 percent of all modern chip manufacturing capacity in the world is in the United States, down from 37 percent in 1990. During the same period, chip manufacturing capacity in China surged from 1 percent of the global footprint to 15 percent. China’s share is projected to reach 24 percent by 2030, while the U.S. share will continue to erode without swift government action. Meanwhile, federal investments in chip research have held flat as a share of GDP, while other countries have significantly ramped up research investments.

U.S. policymakers can help reverse these trends and set in motion a resurgence of chip production and innovation in the U.S.

The current chip shortage—which makes it difficult to purchase everything from automobiles to medical devices—has highlighted the crucial role of semiconductors in driving our economy and future innovation. And it has thrown into bold relief the sobering reality that we need to manufacture more of this critical technology here in America.

The subsidies other countries offer chip companies to establish their manufacturing facilities, known as fabs, are far higher than what the U.S. provides. The cost to operate a U.S.-based fab is 25 percent to 30 percent higher than the cost to run an equivalent fab in Taiwan or Singapore, and 50 percent more expensive than in China. Government incentives offered by other countries are the biggest contributor to the cost advantage other countries have over the U.S.

Federal investments to narrow the cost gap, combined with America’s built-in advantages—homegrown American talent, strong protection for intellectual property, and reliable infrastructure, among others—are precisely what’s needed to make this country the preferred choice when it comes to building new fabs. 

One study concluded a $50 billion federal investment program would create an average of 185,000 temporary American jobs annually and add $24.6 billion annually to the U.S. economy as new fabs are constructed from 2021-2026. The study also finds such federal investments would add 280,000 permanent jobs to the U.S. economy beyond 2026, including 42,000 direct semiconductor industry jobs.

Beyond creating jobs and boosting the economy, shoring up American semiconductor manufacturing will also serve a longer-term good by creating a more resilient supply chain. There are more than 50 points in the chip value chain where one region holds more than 65 percent of the global market share, leaving production vulnerable if a natural disaster, infrastructure crisis, or conflict occurs. Diversifying manufacturing locations will lead to more stable supply chains and help mitigate future chip shortages.

The legislation will also bolster national security. Last year, a bipartisan group of national security leaders, including former directors of the CIA and NSA, wrote a letter to President Biden requesting federal funding for semiconductor research and manufacturing incentives. In it, they cited the integral nature of this technology for global leadership: “Who leads in future technology developments, standards, and the means of secure production and supply will be the true decider of the type of world we build with our allies and partners. The strategic funding plan put forward today will create a great economic return and improved resilience for generations.”
Semiconductor companies, recognizing the importance of building more resilient supply chains, have already begun making critical investments in U.S. facilities. In only the last 12 months, more than a half-dozen fab construction projects—totaling tens of billions of dollars in private investments—have been announced. These announcements show chip companies believe the U.S. government is serious about enacting legislation to make the U.S. a more competitive location for chip manufacturing. Enactment of federal manufacturing incentives would spur hundreds of billions of dollars in additional private investments.

Company investments, combined with robust federal investments, will dovetail to revitalize the American semiconductor industry and ensure our country’s technological leadership for a very long time to come. It’s time to enact federal semiconductor investments to keep America on top in this foundational technology.

John Neuffer is president and CEO of the Semiconductor Industry Association.
 
SIA has commission this report and they have been lobbying for it for 2 or 3 years. Dozens of influencers are involved so we can talk about a campaign at this point. National security is a fig leaf IMO since DOD is not a big consumer of chips and they have no practical limit on capex. They could build a fab it they wanted to. That's effectively what they do with planes so the same could apply to chips. The fact they they do not says a lot about their real priorities. (Yes, I know that they run Lincoln Labs and Los Alamo and a few others but these sites are no tsmc). At any rate:

First point is that the US is doing very well in the overall ecosystem. We may not have the newest fabs but we have critical OEMs, leading edge R&D, many fabless makers etc. Moreover, when one thinks about the usage of chips (software), we are strong indeed. With unemployment under 4%, we have a vigorous job market and I can't imagine that more bunny suits will make a difference. We need well paying jobs for under-educated folks and semi is NOT that.

Second point: who are these companies that we are supposed to help? Intel: returned billions to shareholders. GF: gave up on R&D. Micron: new fabs are all overseas, ON, Microchip, HP, Skywater, Skyworks, etc.? TI has done well without fed money and they did not wait for the CHIPS act to buy the Micron facility and build 2 new factories in Sherman. Samsung and TSMC will keep the crown jewels in Korea and Taiwan, unless WW3 breaks out at which point it won't matter.

Third point: at the end of the day, semi manufacturing is capital intensive and risky as it relies on full factory utilization and ongoing R&D spending. If semi was this attractive, Amazon, Google, Apple would have bought fabs - they bought design houses instead... The CHIPS act will not bring new entrants. It certainly won't create new 200 mm factories, that ship sailed.

We could definitely strengthen R&D and fundamental research. Subsidizing fabs at the tune of $2-3b ea. will make a marginal difference at best.

The other angle is the global nature of the industry which has kept costs on a downward curve. Ever bought a Soviet phone? I did not think so. Closed loops are inefficient and tsmc said as much. I mean there is wafer making, tantalum smelting, carrier manufacturing, packaging in general etc. You can be independent from other countries or you can have an iphone 13. You can't do both.

Finally, all of this is premised on the idea that China stole from us (they did but we were willing partners) and that we can stop/slow the progress of a 1.5b nation. I do not like contemporary China but to think that we can bottle them up is a fool's errand. Some realpolitik is needed.
 
No one can avoid the next shortage and the next over supply.
It's not just semiconductor. Almost everything.
 
No one can avoid the next shortage and the next over supply. It's not just semiconductor. Almost everything.
True but semiconductors even more so since it is a worldwide ecosystem, so many cooks are required to be in the kitchen. Even if we do build semiconductors in the US they will lag what is done in Taiwan and Korea. Kind of like what China has done, they have spent HUNDREDS of BILLIONS and are still several generations behind. But hey, FIFTY BILLION is better than nothing but let's set the proper expectations so we don't get schooled by semiconductor shortages again, absolutely.
 
SIA has commission this report and they have been lobbying for it for 2 or 3 years. Dozens of influencers are involved so we can talk about a campaign at this point. National security is a fig leaf IMO since DOD is not a big consumer of chips and they have no practical limit on capex. They could build a fab it they wanted to. That's effectively what they do with planes so the same could apply to chips. The fact they they do not says a lot about their real priorities. (Yes, I know that they run Lincoln Labs and Los Alamo and a few others but these sites are no tsmc). At any rate:
DoD is not a large volume consumer compared to other industries but all of the capabilities intrinsic to defense modernization are rooted in hardware. DoD cannot actually build their own fab and even if they could what would it focus on? 5nm silicon? Legacy? FPGAs? RF? There is not one fab to meet all of DoD needs nor does DoD as you pointed out have the volumes to sustain a fab economically or to produce at high yield. That was the already done in the past and failed.

What is the national strategy is regarding semiconductors? What problem are we trying to solve? Does it not make sense to have enough capability on-shore so that if there is a supply chain disruption the country is able to scale up to larger volumes. Having 92% of the world's advanced semiconductors made in Taiwan is just silly. Monopolies are bad. You also can't separate national and economic security.

This is a hard problem and I see folks just throwing up their hands and saying $50B won't matter but what is important is collectively between government and industry how do we get the smartest folks together to make it matter - not hordes of lobbyists or the usual non sense. All we will end up with is $50B peanut butter spread to all.
 
Nice article Dan!!! What do you think about a possible China invasion of Taiwan and what effect that would have on the situation?
 
Nice article Dan!!! What do you think about a possible China invasion of Taiwan and what effect that would have on the situation?

Personally I do not think it will happen, especially with the semiconductor shortage narrative that has been trending for the past year. People now realize how important semiconductors are to modern life and no matter how you slice it Taiwan is a critical part of the semiconductor ecosystem. China, the US, or EU can strive for semiconductor independence but it will never happen. In fact, if anything, the semiconductor industry could bring the world together if we play our cards right. The Taiwan silicon shield could cover the word, absolutely. That would be my "I have a dream".
 
I agree But .... isn't that why China might want to take over Taiwan?
70 years ago, 60 years ago, 50 years ago, all the way to 2022, CCP always has all kinds of excuses for invading Taiwan. We don't need to be too serious about an aggressor's reasons why they want to do bad things.

I can assure you CCP will find a reason even if there isn't a logical one.
 
That bill contains spending on lots of things, not just the chip problem. It would help the ecosystem, and everyone who wants to spend money wisely, to partition the "one huge bill" into several smaller ones.

Another oddity, is that the article claims that the cost of a US chip fab is 25% to 30% higher -- without mentioning that we just doubled the money supply creating massive inflation and increased the minimum wage which drives up costs. There are two big reasons why it costs more to run a chip fab here.
Don't go replying that the fabrication people are higher paid so the minimum wage argument doesn't apply. It does because the higher-ups base their idea of what they should be paid on those below them. It's all relative. Of course, there are real expenses that need paid back like student debt, but again, those in the school system get their idea of what they should be paid from those below them. It's a pyramid.
 
DoD is not a large volume consumer compared to other industries but all of the capabilities intrinsic to defense modernization are rooted in hardware. DoD cannot actually build their own fab and even if they could what would it focus on? 5nm silicon? Legacy? FPGAs? RF? There is not one fab to meet all of DoD needs nor does DoD as you pointed out have the volumes to sustain a fab economically or to produce at high yield. That was the already done in the past and failed.

What is the national strategy is regarding semiconductors? What problem are we trying to solve? Does it not make sense to have enough capability on-shore so that if there is a supply chain disruption the country is able to scale up to larger volumes. Having 92% of the world's advanced semiconductors made in Taiwan is just silly. Monopolies are bad. You also can't separate national and economic security.

This is a hard problem and I see folks just throwing up their hands and saying $50B won't matter but what is important is collectively between government and industry how do we get the smartest folks together to make it matter - not hordes of lobbyists or the usual non sense. All we will end up with is $50B peanut butter spread to all.
Before or around 2030, there are several critical national security related programs required significant and advanced semiconductor supplies.

DoD, DOE, NSC, and those other three-letter federal agencies need to deliver programs such as: 6th generation fighter jets, B21 Stealth Bombers, F22 upgrade program, Hypersonic missiles, DDGx (Next-Generation Guided-Missile Destroyer program), Constellation class class multi-mission guided-missile frigates program (FFGX), new enhanced missile defense system, Exascale supercomputer for defense and nuclear weapon program, US Nuclear Modernization program, new joint cyberwar capacity, and new encryption and code breaking capabilities to name a few.

My feeling is DoD, national security related agencies, and their industry partners/suppliers must already have settled on certain semiconductor platform and technology requirements. Otherwise there's no way they can deliver those complicated weapon systems successfully on time before or around 2030. They need real semiconductor products to design, test, and perform final assembly. Most importantly they can't afford to rely on any vendor's roadmaps for program success.

IMO, TSMC Arizona N5 Fab is a critical milestone DoD planners already had in their project plans. To many DoD's fabless suppliers, it offers a great opportunity to move their chip production to a domestic manufacturing site without changing their design and supply chain arrangements.
 
Last edited:
70 years ago, 60 years ago, 50 years ago, all the way to 2022, CCP always has all kinds of excuses for invading Taiwan. We don't need to be too serious about an aggressor's reasons why they want to do bad things.

I can assure you CCP will find a reason even if there isn't a logical one.
That's what worries me.
 
I may be saying crazy things, but I can see that we spend billions of dollars worldwide to make all aspects of computing systems more cost-efficient. Chip manufacturing stands away from there. The costs are growing exponentially year by year - even the cost per transistor. That means that for a component so crucial to the well-being of the society, we leave a critical resource to hands of the (very) few - so few that maybe in the future we wont have to use plural but rather just say ... TSMC. I am wondering if time has come to spend those billions on ways to make chip manufacturing cost-efficient at the smaller scale level, so that the barrier to enter the race can be lower and the risks more moderate. Historically, this has been done for other industries, so this is something to consider. I am not convinced that giving billions of dollars, public land, tax incentives and the rest to those very few to build a fab in a specific territory is the answer, as this leads to making those few even more powerful and even more controlling of chip manufacturing globally, making the barrier for any new competitor even higher (as if it was not high enough the way it is right now).
 
<snip> I am wondering if time has come to spend those billions on ways to make chip manufacturing cost-efficient at the smaller scale level, so that the barrier to enter the race can be lower and the risks more moderate. <snip>
It's already happening, but not in billions but in much smaller donations and with developers who have minimal training in the field and are donating their time. You can subscribe to their mailing list here:
https://list.libresilicon.com/
 
Don't confuse chip shortage with chip technology. TSMC and Samsung have a big market share in high-end devices. However, many fabs (including many newish ones) do not make 5N devices. We are out of chips because we are out of fab capacity and especially lower-tech factories. There are many reasons for that (crypto - widespread over-ordering - WFH. etc.). There's been an underinvestment in lower-tech factories going back to the transition from 200 to 300 mm.

If we really wanted to source domestically, we would have spent spending billions year after year. Nothing like that has happened and nothing like that will happen.

See AMAT annual 2021 report. Sales to China in 2019, 29%, in 2020, 32%, in 2021 33%. Equivalent numbers in USA 13, 10 and 9 % (yes, it's been dropping).

So AMAT is shipping 3X the number of machines to China year after year. If you add Taiwan and Korea, you are 75%. 3 out of every 4 tools is going to Asia. Once shipped and ramped, such hardware will never move.
 
Back
Top