The Battle to Control the World’s Most Powerful Technology

[0] From The New York Times, I'm Katrin Benhold, and this is The Daily.

[1] The American company, NVIDIA, has created one of the most sought -after inventions in the world today, a computer chip that powers artificial intelligence.

[2] But the company has been in an ongoing battle with the U .S. government over its chips being sold in China.

[3] Today, my colleague Anna Swanson, on her investigation into America's escalating war to control this crucial technology, and how it may be losing that war.

[4] It's Wednesday, September 4th.

[5] So, Anna, it's really hard to get into China these days as a journalist, but you were there recently on a mission to find something.

[6] Tell us about that.

[7] Yeah, so I traveled to China actually with Secretary of State, Tony Blinken, when he went to China in April.

[8] I'm a trade reporter, not a State Department reporter, but the State Department reporters were on spring break with their families, so I stepped in and gladly went on that trip.

[9] There's been a lot of tension between China and the United States, and the Secretary of State was there on one of many recent missions to try to stabilize and manage this relationship.

[10] But I had been working on this other story about technology that the United States has banned from going to China, and I was really curious if it is available in China.

[11] So I hopped on a train and I went to southern China to a city called Shenzhen.

[12] This is Shenzhen, it's just right across from Hong Kong on the mainland.

[13] And it is a capital of the tech industry, a really entrepreneurial city.

[14] It's Sunday afternoon.

[15] There are lots of shoppers out, buying every kind of electronics imaginable.

[16] And I went to this amazing electronics market.

[17] It stretches for about half a mile.

[18] But this market is like a maze.

[19] It's just blocks of these confusing buildings.

[20] So we're doing a lot of running around today.

[21] It's this really bustling place, and it's a place that's notorious for selling basically any kind of electronic that you might want.

[22] Cables and electronic devices, fake Dyson hair dryers and Apple Watches.

[23] And I'm wandering around, talking to different vendors, trying to see if anyone sells this technology.

[24] And actually very quickly, I did find some who were willing to readily sell it.

[25] So what exactly is this piece of technology that you found?

[26] So these are the most advanced AI chips made by the U .S. company, NVIDIA.

[27] They're essential now for.

[28] for creating artificial intelligence.

[29] And they've really landed themselves kind of at the center of this technology war between the United States and China.

[30] So how does it happen that this banned product ends up being for sale at this Chinese market?

[31] So the answer to that question is kind of complicated.

[32] But it's the culmination of this multi -year cat and mouse game between the U .S. government and invidia.

[33] And I call it a game, but the stakes are really high.

[34] It's happening in the context of this bigger military and geopolitical conflict between the United States and China.

[35] Okay, so there's a lot to digest here.

[36] Let's maybe break this down.

[37] When I hear cat a mouse, I'm kind of thinking Tom and Jerry, right?

[38] A big player going after a smaller player, but perhaps a very nimble one, who always somehow manages to escape.

[39] So I guess my question is, in your scenario, who's the mouse?

[40] So, NVIDIA in this case is the mouse.

[41] This company was founded in 1993 by a group of Silicon Valley engineers, and they've always had this knack for making business decisions that people didn't really understand at the time, but turned out to be very prescient and really paid off.

[42] Like what?

[43] Invidia is coming out with a 3D game software.

[44] So the first was early in their history.

[45] they were making chips that powered graphics for video games.

[46] And the company at one point almost went bankrupt, but it ended up inventing a kind of novel graphics card that was programmable, and it ultimately dubbed this the graphics processing unit or the GPU.

[47] NVIDIA's graphics processing units, or GPUs, aimed to make video games ever more realistic.

[48] And I don't think the rest of the world, you know, realized what the implications of the GPU were for computing until much later.

[49] A single chip able to process a myriad of calculations all at once, not sequentially like more standard chips.

[50] But actually, Envidio was thinking about this pretty early.

[51] By 2006, they had started working on this software platform that allowed you to use these chips that were formerly for gaming for a lot more.

[52] tasks like science and research.

[53] And then things really kick off with the sort of birth of AI, which comes in 2012.

[54] At that time, researchers started using GPUs to create an image recognition model.

[55] So they're creating a model that can recognize pictures of things like cats or strawberries.

[56] And using these GPUs, they create this model that basically blows the performance of all the previous models out of the water.

[57] Over the next decade, researchers are making progress in AI using Nvidia GPUs.

[58] It's their underlying tech stuff that powers everything that we see in this space.

[59] They're so far ahead of everybody.

[60] And then in 2022, there's this huge breakthrough when ChatGPT, which is powered by Invidia chips, is released.

[61] And that's when everybody really woke up to the power of this technology.

[62] NVIDIA is posting record revenue thanks to the AI boom.

[63] They're talking about $24 .5 billion in revenue.

[64] Basically over the last year and a half, NVIDIA's revenue and its stock have gone bananas.

[65] That's the technical term for it.

[66] Well, you know, Nvidia's up 1 ,000 % since ChatGPT was introduced to the world.

[67] Its value becomes bigger than Amazon, Google, or Facebook.

[68] InVidia, how much further will it go?

[69] Wow.

[70] So this is basically the biggest company many people have never heard of.

[71] Definitely.

[72] And the secret to their success is they are selling the vast majority of hardware and software now that underpins the AI boom.

[73] So it's a little bit cheesy, but people like to say if the AI boom is a gold rush, Nvidia is the one selling the picks and shovels to everyone.

[74] So what you're saying is that Nvidia is really synonymous with AI, this incredibly powerful technology that everyone in the world is now trying to get a hold of.

[75] So that's the mouse.

[76] Who's the cat in this cat and mouse game?

[77] So the cat in this scenario is the US government.

[78] The US government starts to become very interested in this technology.

[79] It's watching Nvidia chips power science and AI developments all over the world, including in China.

[80] And in China, it's making big strides in terms of commercial industry, but it's also helping to develop the military.

[81] So basically, Nvidia is a concern for the U .S. government because it's selling to China, which is the U .S .'s main military and economic rival, and it's making a key component that can be used in advanced weapons technology, which is worrying.

[82] Right.

[83] So I think the vast majority of the chips that were sold to China by NVIDIA weren't used for military purposes.

[84] They're used for commercial technologies like powering social media feeds and keeping high speed trains from derailing.

[85] But they do go to designing new weapon systems.

[86] And there are also concerns about them being used for surveillance, for cybertax, and for things like disinformation.

[87] And that's why the U .S. government is trying to figure out a way to stop the flow of this technology to China.

[88] Okay.

[89] So how does the U .S. try to do that?

[90] So the U .S. has a few different restrictions when it comes to technology.

[91] One of the most basic and widely used ones is something called the entity list.

[92] And what exactly is the entity list?

[93] So the entity list, it's like a kind of It's a list that has foreign companies, universities, and organizations that the U .S. considers a threat to national security and wants to stop the flow of technology too.

[94] The entity list was started in the Clinton administration, but it was used more sparingly at first, and then the government starts to use it much more aggressively under the Trump and Biden administrations, and particularly against China.

[95] And one of the companies that gets added to this list is an NVIDIA customer.

[96] And what makes this particular customer of NVIDIA so dangerous that the U .S. government decides to put it on this entity list?

[97] So this is a company called Sucon.

[98] It's a technology company that does advanced computing, and it's buying American technology from NVIDIA, Intel, and Microsoft.

[99] And in 2019, the U .S. government adds them to the entity list for helping to advance the Chinese military.

[100] And Sugan has also built this computing center in Xinjiang that the Chinese government is using to surveil Uyghurs, the minority population there.

[101] Okay, so Nvidia has been selling chips to a company that effectively has been partnering with the Chinese government.

[102] But now it can no longer sell its chips to Sugan.

[103] Yes, but the thing about the entity list is it's pretty easy to find workgrounds.

[104] How so?

[105] Because the entity list is based on names and addresses of particular companies.

[106] companies.

[107] And companies can find ways to work around it by changing those.

[108] Okay.

[109] That seems like a big problem.

[110] Yeah.

[111] So in this example, a bunch of former Sugan executives have left the company.

[112] It's been entity listed.

[113] And six months later, they create a new company, which is called Netrix.

[114] They have a lot of ties.

[115] Netrix executives are all former Sugan executives.

[116] They use Sugan's technology.

[117] They buy parts from Sucon, and they serve some of the same customers.

[118] And we also found in reporting that Nettrix's owners shared a location with some of Sucon's executives as well.

[119] But the companies say that they're separate companies.

[120] And NVIDIA and Intel and Microsoft all quickly form ties with this new company, Netrix, and continue to sell technology to it.

[121] And that's technically legal?

[122] So this is kind of a gray area.

[123] The companies need to do due diligence to ensure that Netrix is not a threat company for Sukon.

[124] And the companies say that they've followed all the rules.

[125] But, you know, even if it's legal, you have to ask, is this really doing much to accomplish the government's goals of stopping the flow of advanced technology to China?

[126] Okay.

[127] So basically what you're saying is that Nvidia is still finding ways to sell its technology to people, the government, strictly speaking, doesn't want it to sell to.

[128] Or to go back to our metaphor, the mouse seems to have escaped the cat.

[129] And it's kind of incredible how easy it seems to be to skirt these very serious efforts by, you know, the most powerful government in the world to stop the sale of these chips.

[130] U .S. officials are aware that the entity list has some serious drawbacks.

[131] And very early in the Biden administration, they start thinking about more effective ways.

[132] to restrict the flow of technology to China.

[133] And then in the middle of this cat and mouse game, something happens.

[134] They get a wake -up call, which really gets everyone's attention.

[135] There's new concern tonight about China's military capabilities emit a report the country recently tested a nuclear -capable hypersonic missile.

[136] So in the summer of 2021, China tests this new weapon, a hypersonic missile.

[137] The report says the missile circled the earth before speeding toward its target, demonstrating an advanced capability in space that, quote, caught U .S. intelligence by surprise.

[138] And it's important because it could theoretically dodge around U .S. missile defense systems to deliver a nuclear warhead to the United States or take out U .S. aircraft carriers, which is really important in the event of a potential war over Taiwan.

[139] Wow.

[140] All this essentially means China is close to being able to launch a nuclear war.

[141] against any other nation without any warning and there'd be no defense against it.

[142] And U .S. officials are really worried about it because this technology surpasses what the United States has developed.

[143] And much is still unclear about this technology, but the U .S. government suspects that NVIDIA chips have helped to make this type of weapon possible.

[144] So in a way, this missile confirmed the U .S. were fierce.

[145] Basically, American technology helping China to build a weapon that was actually more advanced than the U .S. equivalent.

[146] Yes, this really scared people across the government that the Chinese military might be able to leap ahead in certain areas.

[147] So the United States decides to double down on its efforts to restrict Nvidia chips.

[148] We'll be right back.

[149] So, Anna, after this Chinese missile that was powered in part by American technology, totally freaked out the U .S. government.

[150] What happens next in this cat and mouse game with NVIDIA?

[151] So in 2022, the U .S. tells NVIDIA to stop selling its most advanced chips at the time to China.

[152] Overnight chipmaker and NVIDIA shares fell 7 .7 % after chipmaker said the U .S. government is restricting some sales into China.

[153] So at this point, the Biden administration is going not just after a specific Chinese company, but it halts the same.

[154] sales of these most advanced chips to the entire country of China.

[155] It's a move that could hamstring Chinese firm's ability to carry out advanced work like image recognition and cost Nvidia hundreds of millions of dollars in business.

[156] So I guess in theory, this should be a total shutdown, right?

[157] How does Nvidia react to this?

[158] So Nvidia quickly adapts.

[159] NVIDIA soon found a workaround, developing a slightly less sophisticated chip which got around the U .S. controls.

[160] They take their most advanced chip, and they develop a slightly downgraded version that's within the limits set by the U .S. government, and they quickly start selling it to China, and Chinese companies start snapping up this product.

[161] And U .S. officials are definitely surprised and quite annoyed, and the refrain within the government is that NVIDIA is following the letter.

[162] but not the spirit of the rule.

[163] And to NVIDIA and its defenders, this kind of charge is ridiculous.

[164] They say, if the speed limit is 65 miles per hour and I'm going 63, am I breaking the law?

[165] Of course not.

[166] I mean, does NVIDIA have a point?

[167] In the end, it's up to the government, right, to design its policy in a way that it achieves its policy goal.

[168] And this one does not seem to be working all that well.

[169] So, yeah, the U .S. realizes that these degraded NVIDIA chips will basically allow China to do about the same thing as before in terms of AI.

[170] So the U .S. government decides to lower the speed limit and also ban these newer chips that NVIDIA has made.

[171] But I think this all raises an really interesting question about these rules.

[172] How far does the U .S. government need to go?

[173] And how much does the U .S. want to erode the business of this American company in China so that we don't risk U .S. national security.

[174] The government has to walk this very fine line between business interests and national security, and they're still trying to figure out where this line is.

[175] Okay, but while the government is trying to figure out where this line is, we do know that these banned invidia chips are still available in China, as you yourself saw at this market in Shenzhen.

[176] Yeah, I found people there who are readily willing to sell them.

[177] There's no evidence that NVIDIA sold these chips directly to the market or broke any loss here.

[178] But there are certainly avenues for the chips to end up there.

[179] So for one thing, these chips are all manufactured in Taiwan.

[180] And then they're sent to companies around Asia that put them into computers or sell them on.

[181] And it seems like there are plenty of middlemen and smugglers.

[182] that are then forwarding these chips on to China and making a good deal in the process.

[183] So based on everything you've told us, am I correct in thinking that Nvidia has won this cat and mouse game?

[184] Well, they don't think they're winning.

[185] Why is that?

[186] So they've been shut off from a massive market in China.

[187] So China sales have gone from about 20 % of the company's revenue to more like 10 now.

[188] And it's creating a massive open market there for Chinese companies to potentially develop rival products to what NVIDIA is offering.

[189] So you're saying it's almost like these U .S. restrictions have created a vacuum in which China is developing its own chip making capacity.

[190] Is it possible that these restrictions hurt rather than help U .S. policy goals?

[191] Well, I think the restrictions are still definitely slowing down China's industry and China's chip development efforts because the country can't get the most advanced technology anymore to make its own chips.

[192] But it is creating this kind of huge, protected market for Chinese companies to develop.

[193] And it's certainly strengthening China's resolve the amount of money that it's pumping into technology in order to develop, alternatives to the United States.

[194] This is a trend that China was pursuing before these U .S. restrictions, but now it has no other alternative.

[195] And what do we actually know about China's competition to NVIDIA?

[196] Do we have evidence that they are catching up?

[197] So Huawei, the Chinese technology company, has been rolling out some AI chips, and they still seem to be lagging behind NVIDIA.

[198] But Huawei is reportedly rolling out another AI chip this year that will be as good as NVIDIA's most advanced products.

[199] Meanwhile, NVIDIA is also racing ahead.

[200] It's planning to introduce a new, more advanced chip later this year that will be a lot better.

[201] But it's certainly a race.

[202] And what does the U .S. government make of the fact that China is now investing a ton in this AI space and, in fact, that its leading company in the space, Huawei, is starting to release chips that could rival NVIDIA's chips.

[203] So they know that they're not stopping China entirely.

[204] But in the meantime, the United States is investing in its own chips industry.

[205] Nvidia and other companies are continuing to innovate, and the U .S. is trying to race ahead.

[206] And the U .S. government says that that lead matters.

[207] So, Anna, we spent most of this episode talking about, a cat and mouse game between the U .S. government and NVIDIA.

[208] But as you said yourself earlier, behind all of this is the race between the U .S. and China for technological and military supremacy.

[209] And so I guess the key question here is, as we're ending, what does this story about chips tell us about the state of that race?

[210] Yeah, so there is a lot of concern over the economic implications of this and the kind of collateral damage for U .S. companies that we've been talking about.

[211] And I think, you know, the government doesn't want to cause that, doesn't want to harm U .S. companies, but it thinks that the stakes are a lot higher here.

[212] That artificial intelligence and technology developments can change the course of how the U .S. and China develop and their economies and their national security.

[213] And essentially, America's technological supremacy and military supremacy has now been unquestioned for decades.

[214] And China's technological development is starting to put us into uncharted territory.

[215] And there are concerns that as China develops, that could change the balance of power in ways that in a worst -case scenario could potentially put peace at risk or lead to a war, between the countries.

[216] So there are big questions here.

[217] Can the U .S. maintain its technological supremacy for how long?

[218] And what happens when China catches up?

[219] And I think we're just at the beginning of that story.

[220] Anna, thank you very much.

[221] Thanks for having me. We'll be right back.

[222] Here's what else you need to know today.

[223] On Tuesday, a federal judge in Manhattan denied an effort by former President Trump to claim immunity in his New York state hush money case, in which he was already convicted.

[224] Earlier this year, a jury had found Trump guilty of covering up payments he made while in office to a porn star who had threatened to go public about an affair.

[225] Trump invoked a recent Supreme Court ruling granting president's broad immunity for official acts, but the judge said that ruling was irrelevant in this case.

[226] Hush money payments, he said, were private, unofficial acts outside the bounds of his case.

[227] executive authority.

[228] The sentencing in this case is scheduled for mid -September, less than seven weeks before Election Day.

[229] And a Russian missile strike killed more than 50 people and injured many more, according to Ukrainian officials.

[230] It was one of the deadliest attacks of the two -and -a -half -year war.

[231] Two ballistic missiles hit a military academy and a neighboring hospital in the eastern Ukrainian city of Poltava.

[232] The attacks come at a particularly active moment in the war.

[233] Ukraine is pressing an offensive into Russian territory, while Russian forces are advancing in Ukraine's east.

[234] Today's episode was produced by Ricky Nevetsky, Muj Zady and Eric Kruppki, with help from Luke Wunderplug.

[235] It was edited by Lisa Chow, contains original music by Dan Powell, Marin Lozano, Pat McCasker and Diane Wong, and was engineered by Chris Wood.

[236] Our theme music is by Jim Brunberg and Ben Landsberg of Wanderly.

[237] That's it for the daily.

[238] I'm Catcham Benholt.

[239] See you tomorrow.