The Surprising People Getting Very Rich Off AI

Bloomberg Audio Studios, podcasts, radio, news. This summer, Bloomberg reporter Cindy Wang traveled to Kaohsiung. It's an industrial port city in southern Taiwan. She was there to visit a company most people have never heard of, Kingslide.

It's just very unassuming, factory buildings. Nothing like fancy, nothing like what you can see in Silicon Valley or something like that. You wouldn't imagine that that's sitting at the center of the AI boom.

But it is right at the center of the AI gold rush. In August, Kingslide's 89-year-old founder, Ling Songji, briefly became Taiwan's richest person. Ling's now one of the world's wealthiest people, according to the Bloomberg Billionaires Index. But he didn't get rich making AI chips or software. Kingslide makes metal rails.

That may sound like a very ordinary piece of hardware. But as AI servers have become much bigger and heavier and more expensive and more complicated, these rails have become increasingly important.

These rails work exactly like the ones in your kitchen drawer. But Kingslide's industrial rails can hold AI servers weighing hundreds of pounds so that they can be pulled easily from a rack for maintenance. And American tech giants like NVIDIA are spending billions of dollars to build data centers that house these servers.

The AI story has really gone from the software story to one of the biggest physical infrastructure build-outs really in the world.

That's Bloomberg Opinion columnist Catherine Thorbeck.

I think when we're talking about especially chip suppliers and sort of all the companies that go into building data centers, you know, we think of NVIDIA GPUs and we think of you know, TSMC chips, but there's all these tiny little parts that without access to them, you can't build your data centers.

And the providers of these tiny little parts are creating a class of new billionaires all over Asia.

Whenever people talk about AI billionaires, they usually think about the obvious names, Jensen Huang of NVIDIA or OpenAI like Sam Altman and all these big and well-known names. But Our reporting showed that there is another group emerging. These are the entrepreneurs who sell the picks and shovels of the AI boom.

This is The Big Take Asia from Bloomberg News. I'm Wan Ha. Every week, we take you inside some of the world's biggest and most powerful economies and the markets, tycoons, and businesses that drive this ever-shifting region. Today on the show, the little-known manufacturers behind the AI boom. how an unglamorous corner of the supply chain is minting a new batch of billionaires, and the risks of having critical AI infrastructure concentrated in just one part of the world. Since ChatGPT burst onto the scene in 2022, the AI race has been framed largely as a software battle. Who can build the most powerful models? But the smartest model is useless without the machines that power it, the servers sitting in

massive data centers. And that's turned the AI race into an infrastructure one. Here's Bloomberg Opinion's Catherine Thorbeck.

So the numbers are really quite staggering. The biggest tech companies are expected to spend, I believe it was roughly $ 700 billion this year. And Goldman Sachs actually projected that spending on AI infrastructure could reach between $ 920 billion and $ 1. 4 trillion in 2027. So the really enormous checks are going into physical infrastructure. So GPUs, servers, data centers, cooling, electricity, and sort of the grid. You know, R & D and model development does still matter enormously. But I think the biggest constraint isn't necessarily now whether you can you know, do the software development to build a better model. It's whether you can get enough chips, enough power, and enough data center capacity to really run it.

And the companies providing the essential nuts and bolts of this infrastructure are part of a supply chain deeply entrenched in Asia. Take Kingslide. The Taiwanese firm started out in the 1980s, making furniture handles and drawer slides for American and European brands. But as Chinese manufacturing took off in the 2000s, buyers could get the same parts for less.

forcing the company to pivot.

At that time, a big client of theirs, it's an American client, told them that amid the rising of the Chinese manufacturing, this client demanded Kingslight to cut their prices by 30%. Otherwise, the client would like to choose a Chinese supplier instead. So Kingslight decided to walk away from the customer. And eventually they decided that they have to pivot to serve the server clients. And just in a very short period of time, they was able to secure orders from companies like IBM or Compaq, these server makers.

So they started to supply the server rail kits since 2000. And that's how They could spend decades honing their craft and build up this capability. And when the AI boom finally arrived a couple of years back, they were able to keep up with this increasingly and very strong and very fast demand from very important clients now.

In less than four years, the company's market cap grew more than 3,000% to $ 43 billion. But Kingslide isn't alone. Cindy identified at least seven companies, from Taiwan to China to Switzerland, that now dominate quiet corners of the AI supply chain. Together, they've created over $ 62 billion of additional wealth for their major shareholders.

They all supply a glamorous, indispensable part of the AI supply chain, like the Guangdong DTEK technology. They've made the specialty drill bits for the printed circuit boards.

What many of these new cash flush companies have in common is that they were experienced manufacturers with innovation and networks developed over decades.

For example, like Toto, which is famous for the toilet. But now it also makes this electrostatic chucks that's used in the memory chip manufacturing. Who would have thought that a toilet maker would be connected with this AI boom? Another Japanese food giant, it's called Ajinomoto. It's better known for its umami seasonings and soup bases.

But now it.

Also makes insulating materials that's critical to advanced semiconductor packaging.

When demand for AI infrastructure exploded, these companies weren't starting from scratch. But they moved fast to capitalize on the AI boom.

For example, Kingslight has realized that the biggest moat for their business would be the patents. So they have grown their engineering team from just seven people in the early days to now it's almost over 500 engineering teams. And they have successfully cultivated over 3,800 patents to help protect their technology and technology. to secure their leadership role in this industry. And they realized that only by relying on this patents and this technology they have built over the years that they can secure their place in this industry supply chain.

Now, Cindy, the majority of the companies you profiled, five out of seven, are based within the greater China region. Is there a reason for that?

I think one reason could be that Greater China has an unusually deep manufacturing ecosystem. Like in Taiwan and China, it all sits within a broader network of suppliers, engineers, factories, electronic manufacturers that have developed over several decades. And Taiwan is in particular deeply embedded in the global tech supply chain.

from semiconductors to electronics manufacturing and specialized components. Like TSMC right now is the world's largest chip maker. It has supplied over 90% of the world's most advanced chips. So that means when a demand for a new technology suddenly explodes, there is already an ecosystem of companies that can respond.

And that ecosystem is It takes decades to build and it's hard to be replaced anytime soon.

This raises the question, what happens when so much of the infrastructure powering AI is concentrated in one part of the world? And what's being done about it? That's after the break. Thank you. The U.S. may be leading the race to build the most sophisticated AI models, but the hardware that powers it, from servers to rails and chips, is largely made in Asia. Here's Bloomberg Opinion's Catherine Thorbeck.

When it comes to chipmaking, I mean, Asia has a huge leg up, specifically Taiwan, also South Korea. You know, there are just these sort of entrenched chipmaking ecosystems that have been around for a while that are sort of in a flywheel of supporting each other and suppliers are very close by and you can sort of access them.

But the same concentration that makes these supply chains so efficient can also make them vulnerable. And that's where the conversation shifts from economics to geopolitics.

A Taiwanese company like Kingslide isn't something that Beijing can really switch off. But I think there's a lot of concern around geographical concentration. When it comes to sort of Chinese companies, Beijing definitely has a lot of leverage and it's shown that it's not afraid to use it.

The U.S. has tightened restrictions on advanced chip technology, including equipment and software. China has responded with export controls on critical materials like rare earths.

I think the lesson from the U.S.-China tech fight is really that once supply chains can become geopolitical weapons, the collateral damage can really spread very.

Very quickly.

Every supplier in Taiwan that I have talked to, they have very strong awareness about this technology. broader picture of this U.S.-China conflict or tension that everybody is facing now. And every supplier is thinking about how they can diversify their production base.

All of this is pushing the U.S. government and companies to look for ways to reduce dependence on Asian manufacturing. But that's easier said than done.

I think it's very, very difficult. It's not just throwing money at one factory or throwing money at one fab. It's really building up the skilled workers, the specialist suppliers, the engineering know-how, the tacit knowledge and the logistics, and also, you know, having customers sort of in the same cluster so they can all kind of work together and speed it along. You can subsidize a factory, but you can't really subsidize the 30 years of accumulated manufacturing knowledge and tacit knowledge and skills that it really takes to create these ecosystems.

And even if manufacturing can be diversified, there's another question hanging over the industry. How sustainable is the AI infrastructure boom itself?

We're talking about a trillion dollars in AI infrastructure, which is an enormous amount of money. And at the same time, like none of the model makers have really found, you know, a sustainable path to profits. They're bringing in a lot of revenue, but they have to spend enormous amounts of money to get there. You know, the most sort of exposed specialized suppliers and infrastructure players that expanded the fastest to sort of meet this boom, if something does happen, then they're, you know, obviously very, very exposed. And I think that, you know, a correction could hit very hard.

Which is why, despite their pole position now, there's still uncertainty for companies like Kingslide, the companies expanding in Taiwan and starting a trial production facility in Texas. Like many companies riding the AI boom, it's investing on the assumption that demand will keep growing. Jane Ling, Kingslide's president, told Cindy that the pressure and stakes now feel much higher.

Every six months, there is a new generation of of these AI servers, there are some new demand coming from the clients and that may come all at once. And so she has a very vivid description about her mindset nowadays. It's like every day she is both happy and she's worried. She is happy whenever they are able to meet the client's demand, but she's also worried at the same time because Now that they have accomplished this test, then what about the next one? That was this kind of sense of urgency and anxiety that's driving everybody in this AI boom moving forward.

This is The Big Take Asia from Bloomberg News. I'm Muan Ha. To get more from The Big Take and unlimited access to all of Bloomberg.com, subscribe today at Bloomberg.com slash podcast offer. If you liked the episode, make sure to subscribe and review The Big Tech Asia wherever you listen to podcasts. It really helps people find the show.

Thanks for listening. See you next time.

The Surprising People Getting Very Rich Off AI

Kapitelklicken springt direkt hin

Shownotes

Transkriptvom Podcast bereitgestellt · Zeitmarken springen hin