Presentation

Eric Hanselman

I'm Eric Hanselman, Chief Analyst for Technology, Media and Telecom at S&P Global Market Intelligence. Today, we're doing something a little different. We're actually doing a twin podcast version of both for Next in Tech, and joining me is Kristen Hallam, the host of Economics and Country Risk. We're going to be doing joint podcast together. Kristen, welcome to Next in Tech. I guess I'm welcoming myself into Economics and Country Risk.

Kristen Hallam

Yes. We give you a warm welcome to Economics and Country Risk and thank you for having us on Next in Tech, Eric.

Eric Hanselman

We both are thrilled to have guests on today who are deep into the semiconductor supply chain aspect and the various pieces opposed of it. Joining us today are Chris Rogers and John Abbott. Welcome to you both.

Chris Rogers

Thanks, Eric.

John Abbott

Thanks very much for having me here.

Eric Hanselman

It's great to have you both on. Chris and I have both been talking about clearly, a lot of the upheaval that's going on, we got well-acquainted with supply chain broadly. But of course, now we've got a whole set of additional factors that are in play. From the Next in Tech side, there's a lot of tech that's happening. From the ECR side. There's an awful lot that's taking place and impacting those environments as well. Yes, Kristen?

Kristen Hallam

That's correct, Eric. We're here today to talk about the collaboration that John and Chris have done to produce some really rich insights around these questions.

Eric Hanselman

Well, so let's dive in. I guess I'd like to start off with maybe just doing some definition of that. Really, what's the semiconductor supply chain composed of? Really, what are the mechanics of how it's changing. It's a complex environment with a lot of different players on both the production and consumption sides of these capabilities and things that are playing out in some interesting ways.

John Abbott

Yes. It's John here. I could dive in by saying if you look at it from the bottom to the top, you can divide the making of silicon chips into obtaining the raw materials, refining them, making wafers that have most floors, then etching the circuit patterns on to them, doping them with the impurities to get the right sort of mix of semiconductor properties, slicing them into dies, packaging up into the finished chips.

After that, there's usually further work to be done assembling the chips into that system-level modules. For all that stuff, you need very expensive and specialist machines to do the tooling, adding incredibly small scales, which increasingly get smaller and smaller, and you need the electronic design software to produce the circuit patents and control those machines so that you can massively etch those things on from the actual designs. Complex.

Eric Hanselman

John, the raw materials are just sand, right?

John Abbott

Yes. That's starting at the very bottom. You need first, to look at the raw materials that produce the silicon wafers and that's basically silica sand, and that sand that contains at least 95% silicon dark side in the form of course. Obviously, sand is pretty plantable for certain uses, at least, it's a nonrenewable resource.

A silica sand has to be very pure. It has to be mined, milled, purified using water and chemicals, such as boron and phosphorus. Before it melted down at very high temperatures and spun into a form of a single crystal ingot and the skills and equipment required for those tasks are highly restricted and the mining operations can have a major environmental impact.

That sometimes faces a position where people who don't want it in their backyards. There are also rare elements such as tantalum and neodymium I always think about that Tom Lehrer’s song on The Elements when I start talking about that. It's not really the scarcity of those materials that's the main problem, but it's more the complexities and sometimes the environmental impact of mining and refining them that really is the problem.

Eric Hanselman

Yes. I mean we're not talking about regular beach sand here. This is something that actually is mined, a relatively limited number of mines. But I think the bigger part of this is that the point that you made, which is that this is a matrix of suppliers across the supply chain.

This is a whole set of folks that are participating both in the creation of the raw materials, all the tools for fabricating the actual silicon wafers themselves, all the design pieces, all the fab houses that actually turn this into finished products. There are so many different moving parts to this that in fact disruptions anywhere along here start to create significant problems.

John Abbott

I've seen the sort of big divide in the industry as between the integrated device manufacturers who make their own chips from start to finish. Then on the other side, the fabless companies who outsource their manufacturing quite a few years ago to specialty foundries. The trend in the industry has been towards fabless companies for years now because there's a lower barrier for market entry.

You don't have to invest in all those in hugely expensive fabs. It's hard to imagine any new chip companies out there with a business plan that involves actually building their own foundries. That market has just gone away. Then there's an even further refinement, that with people who just do the designs of the chips and then they license them to other companies who can then pick and choose designs from different designers and then build them into their own products.

That's had a huge impact on the chip industry. The U.S. still designs the majority of processes worldwide. But in terms of manufacturing, that's just fallen off a cliff. In 1990, it was something like 37% of chips globally were made in the U.S. And in 2023, that was down to 12%, and that really caused the U.S. chips and sides active we passed in 2022, which is when they committed something like $52 billion in funding to start repatriating that manufacturing back to domestic U.S. factories.

Eric Hanselman

Is it changes in the manufacturing technologies themselves that have caused that shift? Is this something where we've now got such specialized capabilities that's really what's driving it?

John Abbott

Yes. It's such an expense to retool your fab for every process change from 10 nanometers to 7 nanometer to 5 nanometers and below, just 3 nanometers, every shrink requires a hugely expensive refit. An advanced packaging is much the same, and that things like stacking up the transistors into vertical fins structures to get more of them on more single chip or connecting up multiple chips side by side.

So that complexity also has a huge effect on supply. So GPUs that we're relying on for AI model training and that relies on a process called CoWoS or Chip-on-Wafer-on-Substrate which enables multiple individual dice we integrated on top of a single wafer, and that makes chips potentially much smaller, more power efficient and they generate less heat. But delays in that process caused a bottleneck in GPU availability recently.

This is something where we've been able to disconnect the design and the manufacturing technologies. It's meant that you've got manufacturers who've been able to specialize, but the problem is we end up being that now, we're so focused and concentrated the manufacturing technologies that winds up being a significantly limiting factor.

Chris Rogers

I think just to add to what John is saying, there’s this general view that the semiconductor process is like a single thing. You make a chip in the same way no matter what it's used for. I think when we talk to supply chain specialists who are not tech specialists, they're often surprised by how much those production approaches vary because the end result varies.

The chips that you want to control your smart lighting in your house at home is much simpler and older form of technology than the latest AI graphics processes that John was talking about. A lot of the supply chain disruptions we saw over the past 3 or 4 years weren't necessarily generated by the machine not working or the wrong type of machines being there. It was because it was demand for specific types of chip.

Actually, whilst AI demand is very strong at the moment. Over the past couple of years, it's been more basic processes for using regular laptops or for smart televisions, where we saw a lot of demand during the pandemic era and effectively the capacity for those specific types of chip weren't there.

The production technology used for one sort of chip isn't fungible, isn't convertible to another type of chip. That makes it really difficult for both the chip makers and the firms ordering the chips to know that what they need is going to be there.

I think as another point as well, John talked about the need that to have the different stages of production from sand to finished chips, particularly that packaging layer. Turning a basic wafer and a chip into a completed system. Those aren't always in the same place. They're not necessarily co-located.

So you do see multiple parts of the supply chain across multiple parts of the world and where we're seeing countries be particularly successful in building up a semiconductor business, it's where they've got that full level of integration. So whilst we talk about the semiconductor supply chain, we should really be talking about semiconductor supply chains -- plural for all these different types of chip in all of these different countries.

Eric Hanselman

Yes, that's a really critical point, Chris, because I mean, we think, what, not that long ago, we were having critical shortages of automotive devices. And there was a thought that, hey, we've got lots of manufacturing capacity. But the 3-nanometer fab that's cranking out GPUs can't build the broad temperature range, more rugged semiconductor processes that automotive needs.

So you've got this issue of matching that supply in the individual supply chain to the type of semiconductor that needs to get made. Although that also brings up the question of it's not just the basic issues, but Kristen, these are areas that you work in their geopolitical aspects to this as well.

Kristen Hallam

Absolutely, Eric. Both John and Chris touched on this, but location is an important factor, the where question. My question for Chris would be what are the geopolitical factors that come into play with semiconductor supply chains based on where these different processes that John described and that you described are happening?

Chris Rogers

Yes. I think I often say that the semiconductor industry sits at the nexus of national security concerns and economic concerns. So if you want a strong modern manufacturing industry in your country to provide jobs for the people who vote for you, then you need to be sure that the semiconductors that drive a lot of that manufacturing, whether it's computers or televisions or cars, pretty much everything these days, has a chip in it of some sort. So you need to know that from an economic perspective, you've got access there.

At the same time, the national security element plays out in kind of 2 big product areas. One is very directly for national defense. So there were concerns in the U.S. about 5 or 6 years ago that a lot of the chips that are used in head-up displays that are used in jet fighters are only available from a couple of producers in countries who are natural U.S. adversaries. So whilst the jet was made in the U.S., the head-up display was made in the U.S., some of the semiconductors that are in it were bought from adversaries.

So ensuring that end-to-end is obviously attractive from a national security perspective. The other element is then the future state of warfare, particularly around cyber warfare. And do you have access to the latest chips? And can you deny a potential future adversary access to the latest technology as well?

So what we see as a result of that kind of combination of economic and national security issues, is this desire to bring onshore, also bring back to a specific country, whether you're in the United States, whether you're in the European Union, whether you're in Japan, whether you're in China, you want that full supply chain available for yourself for those reasons. And that's why we've seen the emergence of so many different types of policy approach designed to ensure that's happening.

Kristen Hallam

Now you mentioned reshoring there, Chris, and I know you and I have talked about that topic many a time on the Economics & Country Risk podcast. But I'm curious how it applies to the semiconductor supply chain. Is it really a case of reshoring? Or is it more of a case of -- I don't know, maybe we can coin another term, upshoring, actually building up capacity on your own shores?

Chris Rogers

We're not going to publicize another form of shoring. I think that's a...

Eric Hanselman

Come on. It's analyst's favorite game. It's neologism, right? They need to keep coming up with new one.

Chris Rogers

Yes. So I think as an important point you make, actually, Eric, it's very easy to simplify the decision as to where you base your manufacturing, whether it's a semiconductor or an electric vehicle or a fighter jet or just basic stuff, chairs or clothing or whatever. If you say it's nearshoring, okay, it might be partly about location, but it's never solely about location. It might be friendshoring because you want to be with a natural ally because maybe as a country, you're too small to have an entire supply chain in there.

However, it's not government that makes the decision as to where manufacturing is based, it's companies. And those companies have a profit motive. They have to. Most directors of firms are legally required to look after shareholder interest. And so when you're choosing where to make something, all of those things go in. Have I got the right location in terms of, "Can I actually get my product delivered? Am I in a free trade area that allows me to trade without tariffs? Am I in an area that has low geopolitical risk? But also does it have low operational risk?"

You don't want to move somewhere that's got low geopolitical risk and find that everyone goes on strike all the time. You've got to have the right cost base. I don't want to spend a ton of money on my staff. Specifically, for the semiconductor industry, actually, the labor element also includes, "Do I have the staff that can actually run my fabrication facility?"

And actually, one of the challenges the U.S. has found is, "Okay, we can firehose money at the industry and say, 'Come build your fab here.'" But if you haven't got the workers to work there, if they don't have the skills that are available, then that's a separate challenge and that needs to be addressed. And frankly, there's not a lot of easy answers there.

But there's a lot of different elements that go into that reshoring discussion and where you go to, but clearly, the one thing governments can do is provide the cash to try and bring people, bring companies to their country to build out their silicon, right, wafer manufacturing, their chip fabrication, their packaging operations to a specific country.

Kristen Hallam

So we've talked a fair amount about the CHIPS Act so far. John, maybe you could give us a little color on who's been getting the money so far.

John Abbott

Right. Well, we've been in the process since 2022, but we've just started to see, at the end of last year, some of the first awards from the CHIPS Act. And we – and it started off with some fairly small awards to, as Chris was saying about, legacy chips still being very important.

So GlobalFoundries got a tranche of money at the end of last year. And then just in the last few months, we've seen some big awards for Intel. I think it's $8.5 billion, and Samsung -- $6.6 billion for TSMC for an Arizona project. SK Hynix has getting some awards, Samsung on for Texas. These are multibillion-dollar awards and may include the provision for additional loans as well because even if you're getting $6.6 billion, that's not very much money in terms of these huge projects. So there's still an awful lot of money to be found there.

I'd just like to also mention partly the reason that this is so important is that the vendor concentration has become so pronounced in certain areas. So that's why we're looking at some of this repatriation to try and get some of that capacity back, and it involves multiple forms of investment in ecosystems, not in just one particular aspect.

Eric Hanselman

It's facilities to people, everything all rolling up into this.

John Abbott

Related technologies related packaging and assembly ecosystem, which is why places like Shenzhen in China are so dynamic. Really, if you go around the markets, you can see all these companies geographically located next to each other, which really helps for efficiency. So that's sort of got to be duplicated to a certain extent when repatriation is being considered.

Chris Rogers

I think I'd also add that important point about the ecosystem. I'm always amazed by what's needed to produce chips. Just recently, a company who produce photoresist etching, so it's a type of kind of acid material. I'm sure John has got the more accurate description of it. But effectively, they have supply capacity in Japan. They've now decided to expand that because the investment environment for it in Japan is preferable to elsewhere.

But what it means is that the etch chemicals that are required in other countries, whether that's South Korea, the U.S. or elsewhere, are effectively relying on Japanese government willingness to allow the export of that very critical chemical. And there's been instances in the past where there's been restrictions on those very detailed chemical areas that have been pulled into other areas of trade disputes. So Japan and South Korea had a trade dispute a few years ago. And those chemicals, the export of those chemicals were restricted.

And so when we see these investment plans that a lot of the governments have put in place, whether it's the CHIPS Act in America or the equivalent that EU have got in Japan and South Korea and China have got long track records of using government funding to help develop those ecosystems. It goes all the way back to some of those specific chemicals.

So when we're looking at the potential success of some of these investments, that question of, "How complete is the upstream ecosystem?" or, "Are you just building in fab?" is great for the press release from the government and the cutting of the ribbon. Is it actually going to be a successful asset economically and strategically? Depends on all those other areas being built up as well.

Kristen Hallam

Good points there. Chris, there's another aspect of the CHIPS Act that I wanted to ask you about. You were talking about national security considerations earlier. Can you tell us how the CHIPS Act seeks to address those national security considerations?

Chris Rogers

Yes. So I think the CHIPS Act specifically isn't just providing big gobs of cash for building facilities. It's also providing functional funding for research and development and for educational areas as well because there's a recognition that you need to build up the intellectual property aspect of the industry and not just the physical production.

The U.S. has a long and storied history across government of supporting research and development, everything from the space program leading to the creation of Teflon through to kind of cold wall security considerations leading to the creation of the Internet.

And so there's also a desire on the side to make sure that a lot of these intellectual property areas coming to the benefit of strategic adversaries as well. And that's why we've seen kind of on the side separate to the CHIPS Act the build-out of a lot of export restrictions in relation to the very highest technologies.

So that's around the GPUs themselves, the graphic processes themselves used in AI, but also these big machines that John was talking about. The bad thing about concentration in a supply chain from a government perspective is you have a point failure risk. The good thing is it makes it much easier to control the flow of these highest technologies if you've only got to work with a handful of other countries and a handful of companies.

Now the problem with any export restriction, if you're trying to stop it going to your strategic adversaries, is to make sure that you are covering the right technologies. And because this space moves so quickly, what we've seen is that the U.S. has actually had to put in place rules that have a gray zone.

So it's not just you cannot export this particular chip, a particular company, a particular brand. If your chips have these characteristics around processing capability, power usage, even the scale of the chip itself, if you lock those down, all you will find is that a manufacturer will create a China-specific version so that it doesn't have to deal with the U.S. rules. So we've seen a move towards these types of more adaptive restrictions.

Eric Hanselman

Looking ahead, what should we be thinking about in terms of other aspects that are coming into play?

Chris Rogers

If we're looking ahead on the national security side, we talked a second ago about the kind of the gray zone of export restrictions. And I think we're going to see more of that. 2023 was the year really where the EU, Japan and U.S. were adding restrictions. 2024 and beyond, we're going to start to see, I think, firstly, more of those kind of restrictions. But secondly, we're going to start to see a degree of retaliation.

So we've already seen the government in China start to restrict the use in government settings of technologies produced by U.S. companies, whether that's chips within telecoms networks or the use of software by government agencies. So the use of Windows-based technologies specific to U.S. companies. So I think that age of retaliation is really just beginning to pick up. And John mentioned much earlier, the use of some of these critical minerals, rare earths, particularly like neodymium where China really dominates those supply chains.

Thus far, what we have seen is China being willing to apply restrictions to specific elements of critical minerals, particularly around gallium, germanium and graphite. And I think there's a clear risk those restrictions will be spread out to other areas as well, other critical minerals, particularly around rare earths. And that's going to continue to cause challenges for the semiconductor supply chain as we go through the coming months and into the next year or so.

John Abbott

Yes. And I'd just like to add that this is a bit of a problem for companies, both in China and in the U.S. because they potentially halves their market. And so depending on how far these restrictions go, it becomes difficult for those companies to keep their huge investments realistically sustaining their business models, but when half the market is disappearing. And also it means you're doing the same job in multiple places, basically definitely puts costs up for the entire industry.

Eric Hanselman

And these are manufacturing processes that really focus on scale. And if you've got to bifurcate your approach and your production capacity, it adds even more stress to what clearly is a complex-enough situation.

Chris Rogers

I think I'd add as well that these restrictions can be counterproductive. We saw when Japan restricted the exports of photoresist chemicals, South Korea started building up its own photoresist chemical business. We've seen China already start to develop its own high-tech semiconductors. And the problem is the longer those restrictions on exports go through, the more the governments in other countries are incentivized to build up their own industry.

After all, the CHIPS Act isn't the only game in town. You run the risk that if you restrict access to another market, to John's point, you restrict your company's ability to raise money to pay off the R&D costs, the cost of building these big machines, and that restricts their ability to then go to the next round of development.

So governments have to strike a really careful balance between the national security consideration and the long-term economic consideration. My concern at the moment is that focus on the short-term national security consideration could prove count productive for the construction of longer-term supply chains.

Eric Hanselman

We've got more than enough history in the effects of boycotts, limitations, blockades and native development. But again, one more complex world to look in on. But I guess more to discuss on this front, yes, Kristen?

Kristen Hallam

Absolutely, Eric. As Chris often says, I will quote you, Chris, "Supply chains never sleep." So we'll be doing another one of these twin episodes in the near future?

Eric Hanselman

Well, I certainly hope so and look forward to it. But this has been a great one, but we are at time for today.

Kristen Hallam

Thank you for having me as your copilot today, Eric.

Eric Hanselman

And vice versa. It's been a pleasure to tag along and join this discussion. And of course, John and Chris, it's been great having you on.

John Abbott

Thanks very much.

Chris Rogers

Yes, thanks for having us.

Kristen Hallam

Thank you to Eric, John and Chris for our discussion today. We will continue visiting some of our key themes for 2024 in future episodes of the Economics and Country Risk podcasts. So stay tuned.