Catch up on stories from the past week (and beyond) at the Slashdot story archive

 



Forgot your password?
typodupeerror
×
United States Businesses

Nvidia CEO Says US Will Take Years To Achieve Chip Independence (bloomberg.com) 121

Nvidia Chief Executive Officer Jensen Huang, who runs the semiconductor industry's most valuable company, said the US is as much as 20 years away from breaking its dependence on overseas chipmaking. From a report: Huang, speaking at the New York Times's DealBook conference in New York, explained how his company's products rely on myriad components that come from different parts of the world -- not just Taiwan, where the most important elements are manufactured. "We are somewhere between a decade and two decades away from supply chain independence," he said. "It's not a really practical thing for a decade or two."

The outlook suggests there's a long road ahead for a key Biden administration objective -- bringing more of the chipmaking industry to US shores. The president has championed bipartisan legislation to support the building of manufacturing facilities here. And many of the biggest companies are planning to expand their US operations. That includes Taiwan Semiconductor Manufacturing Co., Nvidia's top manufacturing partner, as well as Samsung and Intel.

This discussion has been archived. No new comments can be posted.

Nvidia CEO Says US Will Take Years To Achieve Chip Independence

Comments Filter:
  • FABs are the problem (Score:5, Informative)

    by saloomy ( 2817221 ) on Wednesday November 29, 2023 @01:12PM (#64041521)
    They are more complex than refineries and are built to a specific size, like 5nm or 8 nm. You only build the latest one and make the newest chips there. Then, when a newer one is made, the previous ones are shifted to make less-complex chips like modems and controllers. There is a long pipeline of older and older chip FABs that would all have to be replaced with domestic FABs before the US can say it is independent. I remember there was an old one from a decade or two ago in Poughkipsee, NY called IBM Fishkill that was 300nm IIRC. So, there is some domestic FABs that exist, but due to water consumption, and other regulatory hurdles, most are overseas.
    • by jacks smirking reven ( 909048 ) on Wednesday November 29, 2023 @01:28PM (#64041593)

      I think there is some conflation with "independence" and "domestic". It's 2023, globalism is the ruling order of the economy and the USA is the hegemon of a large group of allies. Independence can and probably should mean a supply chain that is not dependent on global rivals like China and Russia or vulnerable geopolitical situations like Taiwan. It's all about risk management and right now the risk and fear is about China controlling supply and also the fact it could greatly disrupt the supply from Taiwan.

      I don't think the US has to have every fab for every node in it's borders but increasing supply in places like Canada, the EU, Japan, Mexico, Korea at this point and time are just as good as "domestic" although we need a strong base of chip manufacturing on our shores but we don't need all of it.

      • South Korea has alot of FABs, Japan has some FABs. Europe is part of the tech supply chain and Canada can certainly provide most of the mineral exports.
        • by dgatwood ( 11270 )

          South Korea has alot of FABs, Japan has some FABs. Europe is part of the tech supply chain and Canada can certainly provide most of the mineral exports.

          One caveat, of course, is that Japan and South Korea aren't much farther from China than Taiwan, and if China takes Taiwan by force, there's no guarantee that they'll stop there. Europe seems like a much safer fallback.

          • by G00F ( 241765 )

            yea, if china took Taiwan that entire area would be held or hotly contested where any supply chain depending on Japan/South Korea would be dead.

            I would make sure all military chips can be supplied from nato nations, and a certain amount domestically.

            I would think most of it would operation, function, and perform fine in 32nm. (IIRC most motherboard chips are 55nm or bigger)

      • The US should just one-up China by offering Taiwan statehood like we did Hawaii in 1959. Or, we could make it a US territory like Puerto Rico. They would love the tax advantages. It would be a political coup for the US administration. Then, we could put our military there. 'nuff said.

        The Chinese wouldn't have the balls to go to the carpets over it. They'd lose, and they know it.
        • While i like this this idea and it does give me a bit of a patriotic chubby I doubt the Taiwanese people want this nor would the US citizenry accept it.

          • The US would accept it, for sure if it ended strife with China. As for the Taiwanese: https://www.taipeitimes.com/im... [taipeitimes.com]
            • How would it end strife with China? It puts them in an even more uncomfortable position.

              • Because that would be the end of the conflict. The tension would be ended. At the end of the match, you look in the ring, and the guy dancing was the winner. Now, it's like we are about start another round. China would not be able to counter a US takeover. Not for a long, long time. It would be the equivalent of trying to conquer Hawaii.
              • It doesn't end strife with China.
                Despite saloomy getting all hot and sweaty at the thought of sending other people off to fight in another pointless war, it's never going to happen.
                China doesn't have the capability to invade Taiwan, and even if they did, it would tank their economy. They're not idiots, but they do depend on trade. There is no easier way to screw up your trade than declaring war on your biggest trade partners.
                • I don't know, Russia kind of proved that smart and savvy autocrats... aren't. They surround themselves with yes men and end up believing whatever they want is the correct thing. Then they can't back down after taking action whatever the cost.

            • by znrt ( 2424692 )

              taiwanese overwhelmingly know that the current status quo is the best possible scenario for them for the time being. they might not like china, but they don't trust the americans either although they might welcome their influence as part of the international support precisely because it somehow contains china and makes this balance possible. push for independence is minoritary for the same reason. they know full well they're in a catch-22 situation and the best possible strategy is to stay low for the time

              • by AmiMoJo ( 196126 )

                I don't think China cares as much about Taiwan as some people think. There is a lot of posturing and rhetoric, but their main concern is that the US doesn't get any ideas about bases in Taiwan and the like. It's the same idea as their support of North Korea - keep a buffer between them and the US.

      • South Korea is even more vulnerable to attack / invasion by China than Taiwan. Unlike Taiwan, it's a simple land invasion they're virtually guaranteed to win.

        That's not to say Taiwan isn't vulnerable. If it wanted to deny its fabs to the western world, China could just lob artillery over the strait at key locations until they are forced to negotiate or surrender.

        • While that is true from a geographic situation what makes Taiwan vulnerable is the political environment in which it exists and it's history with China and the fact that China holds a claim over the island.

          With Korea the situation is far more clear cut. SK is a sovereign nation, carved out in a conflict that China was in involved with and China has it's ally in NK. A Chinese attack on SK is not a grey area whatsoever, it would be viewed as a full act of aggression with little provocation and thus the imm

          • >China holds a claim over the island
            You have that exactly backwards.

            • by AmiMoJo ( 196126 )

              The Taiwanese government's official position is that they are the legitimate government of all of China, including Taiwan and the mainland.

              The CCP's official position is that they are the legitimate government of all of China, including Taiwan and the mainland.

              • That sounds downright delusional coming from Taiwan. Not saying you are wrong, just that one tiny little island that's essentially been allowed to conduct themselves autonomously is hardly the government of mainland China. Sounds like trying to say USA is really the property of the native tribes.

                • That sounds downright delusional coming from Taiwan. Not saying you are wrong, just that one tiny little island that's essentially been allowed to conduct themselves autonomously is hardly the government of mainland China. Sounds like trying to say USA is really the property of the native tribes.

                  Actually it sounds like trying to say exactly what the "Land doesn't vote, people vote" folks have been saying about the USA for the past 8 years -- that 10 metropolitan areas should govern all 50 states.

                  Most capitals in human history have been tiny compared to the lands they control. Beijing is less than half as large as Taiwan. It's no more absurd to think that "one tiny island" could govern a whole nation than to think that "one [even smaller] city" can. Because by that standard, China as a unified count

                  • All governments are delusions. But they are shared delusions - either voluntarily through some form of ratified charter, or induced at the point of a rifle.

                    Might makes right. If you have the support of the majority or the power of enough rifles, you get to call all the shots.

                    Regarding the 10 metros controlling USA, those 10 metros have the majority of people living in them. We aren't a direct democracy though, so popular vote means diddly squat. Not to mention, if we did have a direct democracy, some people would indeed vote differently.

                    Given how most of the states' voting rules are setup, most of our votes don't even count. If you aren't in a battleground sta

          • by gtall ( 79522 )

            China holds a claim over whatever the hell the runt running the joint wants. Tibet should be given back to the Tibetans. Xi Jinping is annexing the S. China Sea because no country in the area can stop him. Taiwan is only meaningful to Xi Jinping because he's loser due of how he runs China and needs a distraction; he thinks if he takes Taiwan that somehow he isn't just another male with an inferiority complex.

          • by iAmWaySmarterThanYou ( 10095012 ) on Wednesday November 29, 2023 @03:59PM (#64042055)

            The current government of mainland China has absolutely no historic or legal claim to Taiwan. None. Zero. Zip. Nada.
            The former government fled there after losing the main land to Mao's uprising. Mao's boys never set a single toe on Taiwan nor ever tried.

            Taiwan is the last bastion of the government that had control of the mainland. The war never ended as both governments still exist today.

            If by some miracle Taiwan was able to launch an invasion of the mainland and recapture it they have an infinitely better claim on the mainland than the communists have on Taiwan.

            • Even Taiwan wouldn't say that, though. With both sides claiming that mainland and Taiwan are part of "One China" they are implicitly stating that "China" has claim to Taiwan.

            • by AmiMoJo ( 196126 )

              It's a little more complicated, because at the time of the split the government of Taiwan was fascist. I know you probably think of the communists as just as bad, but back then there was a legitimate reason to want to liberate Taiwan from fascist rule.

              Of course, Taiwan became a democracy on its own, so today we would obviously prefer that the democratic government become the one for the PRC as well, but it's hard to see how that could happen.

              • It isn't complicated at all. I made no distinction between their government forms as far as historic or legal claims are concerned.

                Being a western style democracy does not increase legal claims to land nor does being a fascist state change historical fact regarding who controlled what and when.

                • by AmiMoJo ( 196126 )

                  Perhaps in theory, but in practice being a democracy means they will have a lot more support from powerful and influential nations, and being fascist similarly de-legitimises any claims in those same eyes.

                  • In practice the mainland communists have no historic or legal claim to Taiwan.

                    In practice the Taiwanese government in exile has a legal and historic claim to the mainland which it has zero possibility of ever reclaiming or enforcing.

                    The government forms have nothing at all to do with legal or historical claims. Nothing.

                    Your argument is based on political support from other powerful nations. Aka force. If we're going to switch from historical and legal claims to might makes right then that's an entirely d

                    • by AmiMoJo ( 196126 )

                      I hope force will not be used. There are other consequences though, as Russia has discovered.

                    • Might makes right. That's how humans operate. Everything else is really just screaming in the wind.

                    • Yes, me too. But I don't think it will happen. The odds are not well stacked in China's favor on this one either militarily, politically, or economically but others have certainly fooled themselves into war in the past they were better off not starting.

                      Fingers crossed that they're more practical than earlier aggressors in similar situations.

                    • It's more complicated than that. Power isn't just military force.

                      If you're a country like China with a decent military you can't go off Willy nilly invading your neighbors because you may find the rest of the world refuses to trade with you, sell you critically needed energy supplies and so on.

                      Soft power is as real as guns and missiles. In a highly connected world playing nice is generally a better plan. Or you could say fuck you all and be North Korea which is a total basket case as a result of ignoring

          • Yeah, as if nations like China or Russia need a real reason to attack places that have something they want. China has been building a fury domsstically over all things Taiwan for a decade. They don't give a rat's ass about how it looks internationally; that's for the consumption of the suckers who actually think any of that stuff matters.

          • the official US position is still the "One China" policy.

            Depending on context, the US may say that ROC is the One China. So far, though, position is always given in vague terms and is intentionally ambiguous. It could be considered an invasion of the "One China" - either way, if there's an invasion the US will be forced to declare by action or inaction which One China is the one they meant.

    • by edwdig ( 47888 )

      I don't think regulatory hurdles are a big factor. It costs billions to build a new fab. Whenever someone like Intel builds a new one, they shop around for the best tax deal they can get. They also put a high value on spreading out the fabs. They don't want a natural disaster to wipe out too much production at once. They also want to spread things out for political reasons. Basically the same factors that make us want local production make them want to spread their risk around. No one wants a sudden politic

    • The ecological disaster fabs are is the problem. Semiconductor fabrication is an environmental nightmare. One big enough that even crooked governors couldn't turn a blind eye to. You're basically facing the choice of either creating a toxic hellhole around the fab plant or up the price to fabricate those things by a magnitude.

      Neither is something you want, so China it is. Or India. Or some African hellhole. Somewhere where your constituents who want cheap electronics crap don't complain about their water ta

    • I remember there was an old one from a decade or two ago in Poughkipsee, NY called IBM Fishkill that was 300nm IIRC. So, there is some domestic FABs that exist, but due to water consumption, and other regulatory hurdles, most are overseas.

      Actually the USA has close to 100 operating fabs. Out of those who disclose their nodes the lowest in the list is 14nm with a quite a few operating in the but due to water consumption, and other regulatory hurdles, most are overseas

      No, due to the fact that there are 195 countries on the planet most are overseas. The USA is currently the third biggest semiconductor manufacturer in the world. The regulatory hurdles are not a consideration.

      • I don't know how I deleted half my post while hitting submit... Here's the full one:

        Actually the USA has close to 100 operating fabs. Out of those who disclose their nodes the lowest in the list is 14nm with a quite a few operating in the sub 200nm regime. These are not old, or outdate, or even repurposed fabs. In fact some of them are brand new, e.g. Texas Instruments only opened a new fab with a 200nm node size last year.

        But this is not what anyone is talking about. When we are discussing independence we are almost exclusively focusing on high-tech. 12nm and lower, not the ability to produce all semiconductors across the entire range.

        but due to water consumption, and other regulatory hurdles, most are overseas

        No, due to the fact that there are 195 countries on the planet most are overseas. The USA is currently the third biggest semiconductor manufacturer in the world. The regulatory hurdles are not a consideration.

        • When we are discussing independence we are almost exclusively focusing on high-tech.

          Independence is being able to build your own chip fab, from components you can make yourself. Not just importing the components to build a chip fab. But the ability to build all those components yourself. From things you can make in you own country. Anything less is relying on those other countries.

      • There are more advanced chips being made in the US. Intel operates fabs that make 10nm and 7nm chips, and is working on making more advanced processes such as Intel 4 and 20A. The GlobalFoundaries fab in Malta NY gets down to 12nm (the most advanced process that company makes). The upcoming TSMC plant will make 3nm chips if it ever gets finished.
  • Balderdash (Score:5, Insightful)

    by ranton ( 36917 ) on Wednesday November 29, 2023 @01:32PM (#64041611)

    It all depends on how much money we spend on the transition. If we treated it as a true security threat, hundreds of billions could be spent annually to address this much quicker. 20 years ago top of the line fabrication was at 90nm. We are at 3nm today. The industry has completely retooled its fabrication plants a half dozen times or more in the past 20 years. I have no doubt we could catch up in 5 years if we really cared.

    I do think at our current pace it could take 20 years though, but that is our decision to allow it to take that long.

    • Those 90nm fabs are still making things like USB controllers, display controllers, hardened processors for use in space, toy electronics, etc... Thats why 20 years. To get EVERYTHING to be domestic, you have to replace all the fabs of the last few decades that still contribute to the US consumption. It wouldn't be worth it to stand up a new 90nm fab, since it still costs billions but cant produce state of the art stuff, unless you willingly decided to make a cost-losing operation. But, why would you? You ca
      • by ranton ( 36917 )

        unless you willingly decided to make a cost-losing operation. But, why would you?

        Because the costs are being subsidized by the American taxpayer. Either this is a security risk for the US or it isn't. If it is, that is what the government is there for. An aircraft carrier has zero ROI, except to manage security risk.

        • An aircraft carrier does have zero ROI, but how many zero ROI projects can or should we sustain? We do not need (yet?) to make our own chips since we can again... buy them for less. If we can not buy them at all, then it would make sense to secure our supply by either producing them domestically or by conquering the land that has them with a few of those said aircraft carriers which would now have an ROI.
          • If we are planning seriously for another world war then we have to assume that there will be supply chain interruptions as our suppliers get conquered or just bombed, and the safest place for us to have the fabs is here at home.

            • In theory we could stockpile and should anyway but for a lengthy engagement we'd be screwed right now.

              Not every fab needs to be cutting edge though. I doubt the controller chips in a cruise missile need to be 3nm, for example. Any old thing should do. It's got a short one way existence anyway.

              • You cant stockpile chips that have yet to be developed. If, for example, Carrier wanted to make a new thermostat, They do not go drop the latest Xeon and run a full server. They would need some custom controls (maybe not but bear with me here). So, they design an SOC or use an old snapdragon. Then they make some custom chip that has the I/O they need to communicate with their system, and have a company like Global Foundry make those chips to drop into their boards. Those chips do not need to be 5nm, which
                • We're talking war not thermostats.

                  The next real war between super power near peers is unlikely to last long enough for either side to develop, test, build, and deploy anything new in sufficient quantity before it no longer matters.

                  For example, US v China over Taiwan will last as long as it takes for a few carrier groups to show up plus a few weeks. At most. By then we've either achieved total air and sea superiority or we lost a few carriers and we're fucked. It won't take years to find out.

      • by dgatwood ( 11270 )

        Those 90nm fabs are still making things like USB controllers, display controllers, hardened processors for use in space, toy electronics, etc... Thats why 20 years. To get EVERYTHING to be domestic, you have to replace all the fabs of the last few decades that still contribute to the US consumption. It wouldn't be worth it to stand up a new 90nm fab, since it still costs billions but cant produce state of the art stuff, unless you willingly decided to make a cost-losing operation.

        In principle, AFAIK, there's no reason that a 3nm fab can't fabricate parts designed for a larger process at their original size. After all, the fact that a fab *can* create features that are tiny doesn't mean that it can't still create giant features. You pretty much have to do that all the time on high-density parts for things like power handling. We don't typically do that, of course, because there aren't nearly enough small-process fabs to meet demand for high-density parts as it is, but we could.

        So

  • by MacMann ( 7518492 ) on Wednesday November 29, 2023 @02:17PM (#64041749)

    Bloomberg wants people to register to read their articles, something I won't do. Not when I can just do a search on the headline to find someone willing to provide the same news without nagging me to pay up.

    How about this instead: https://www.msn.com/en-us/mone... [msn.com]

    Mining for raw materials will be a problem, and just opening up new mines will take 20 or 30 years. I suspect rare earth elements are a concern, not because they are rare since they are not, but because the USA closed most domestic production. Some rather silly rules on mining in the USA meant it was just cheaper to buy from China than produce domestically. There's plenty of rare earth elements in the USA, and perhaps we could see the USA ramp up production relatively quickly because many of those mines have been merely left idle than reburied. It would still take years to get all the people and equipment in place to restart mining, and a few more years to ramp this up to where the USA reached independence, but perhaps not 20 years.

    I'll see people talk about how some element is so incredibly abundant that it should be easy to mine more. With silicon there's plenty of it around but the silicon we need for integrated circuits needs to be in a purity that is rare to find in nature. We could use a lower grade ore but that means putting more work into the refining. Take your pick on the element that we need, perhaps lithium is another example. Lithium exists in large quantities as salts dissolved in seawater but getting that out is very energy intensive, and could cause considerable environmental damage if not done with care.

    • If there was a war or China cut us off I think you'll see ground breaking at rare earth mines in the US pretty damned fast.

      Our restrictions are regulatory not physical or financial.

  • Nvidia is not a US company at heart, and has no plans to be so. The problem that I see is that AMD is in the same boat.

    Looks like Intel has my support by default. At least they are trying to maintain some semblance of US manufacturing.
  • "Years" - really? That seems optimistic. And yep - turns out it's a couple of decades. That sounds more reasonable. Unless you mean "years" in the sense that it's any time between next November and Ragnarok... in which case, sure... years.

  • If the outlook really is 10-20 years, then the best time to start shifting things was 10-20 years ago.

    The second best time to start making the shift is today. And the US is making moves in that direction.
  • There is a lot of drama in the world currently, but interest groups like Nvidia have their agendas. Paint a dramatic picture and money will pour in. As many people equate efficiency/shorter waiting times with quantity of money, it is what will be pushed for. And probably achieved.
    Chip independence may be critical or not, it may take 20 years or 2, but that is not an issue. How much money will flow because of all of it is the issue here.
    I know it sounds cynical but it is exactly what at least most of it is a

    • A lot of people made a lot of money shifting the US economy towards globalization. Unwinding that is going to cost someone money, and likely it will be the American taxpayer.

      • Globalization was never the problem. Interdependence and redundancy are good globally. Concentrating our dependence on specific nations is the problem. It's also cheaper and more resilient of a supply chain if we just find a way to get more redundancy whether it's domestic or just more countries.

        Things like the UN / NATO are not going to help - they are on the war/peace and humanitarian side of things. But it's the right kind of idea.

        • It's always been a race to the bottom. You can't expect there to be lots of redundancy when there is a strong preference to buy from the lowest cost supplier. The supplier with the least regulations, lowest labor cost, most government financial backing, and most manipulated currency. We can't compete globally with a centrally planned economy. Once we run out of ideas and patents, we're going to have a very difficult time supplying anyone but ourselves with US built chips. And even then, at a greater cost th

  • With high confidence I predict that the only country with full semiconductor independence in 20 years from now will be China. Because of US tech blockade.

  • by TomGreenhaw ( 929233 ) on Wednesday November 29, 2023 @05:47PM (#64042341)
    Nvidia designs but does not fabricate its devices. It would be in his company's best interest if the US government heavily subsidized domestic semiconductor fabrication as it would lower his cost as well as reduce supply chain risk.

    All of this is due to concerns that China will invade Taiwan and that the US would be at war with China. This is nonsense. Contrary to propaganda and media hype intended to whip up nationalism and boost government corporate handouts, the reality is that neither China nor the US can afford a war with each other. In that event, China's economy would take a catastrophic economic hit and the US would see a dramatic inflationary spike with both countries experiencing wrenching political turmoil. Both leaders met recently and confirmed that neither could "turn their back on each other". For a variety of reasons on both sides, the confrontational approach is being tried and will ultimately be found wanting. A far more logical, desirable and likely course of action long term would be warming relations between China and the US, with a US diplomatic effort to conclude China's forever civil war and a close partnership between Mainland China and Taiwan.

    Moreover, there is a misconception that innovation is linear and that it advances steadily over time. History shows that there are huge leaps punctuated by periods of slower improvement. There is no reason to believe that computing couldn't be revolutionized by paradigm shifts in technology. Examples could include genetically engineered biological neural networks that drive AI, space based server farms using super cooled digital systems based upon Josephson junctions with CPUs running at THz clock rates, spintronics, or even advances in chip substrates that allow dramatically higher clock rates and lower power consumption. 3 nanometer design rules are likely close to a physical limit anyway and it is a matter of time that intel catches up to TSMC with superior FET designs on top of that.

    I have always been a huge fan and loyal Nvidia customer. That said, assumptions that Nvidia somehow is the only game in town and that their CEO's opinion is infallible are false. While I agree that it would take years to replicate TSMC's capability on US soil, it would not take decades. As long as it's cheaper elsewhere, that is where US companies like Nvidia, Apple, Intel, AMD, and Qualcomm will fabricate their US designs. In the event of an emergency, imagine what a partnership with the US government, western academia and these US companies could accomplish. Does China really want to force that cooperative effort? Of course not and their leadership is smart enough to understand.

    Finally, the US does not need chip independence although it could easily achieve it. These arguments are just more indirect US bashing propaganda by implying the US is somehow behind. It's relationships with Europe, Japan and Korea are unshakeable as the world's democracies tend to be reliable trading partners.

    It is likely that the US will never have total chip independence because it's unnecessary and simply doesn't make sense.
  • But at least we started.
  • And that is a big problem.

  • This is probably the first time a CEO has given a most likely correct estimate of how long something will really take. You can't build an industry overnight, but you can destroy one much faster.

"I got everybody to pay up front...then I blew up their planet." "Now why didn't I think of that?" -- Post Bros. Comics

Working...