Navigating the AI boom: A framework for investing

Dive into a world of AI investing opportunities that go beyond the expected. Join Fidelity global SMID portfolio managers, James Abela and Maroun Younes, along with technology sector lead and analyst, Terence Tsai, for an engaging virtual fireside chat on this rapidly changing landscape.

Get under the hood of our proprietary AI investment framework as our experts share with investors how they are leveraging Fidelity’s research platform to uncover new and diverse opportunities. Specifically, discover the action behind the Fidelity Global Future Leader Fund’s alpha, and explore China’s burgeoning tech eco-system to find out why software titans are likely to face stock market jitters.  

Discover the Fidelity Global Future Leaders Fund

 

Lukasz de Pourbaix (LDP): Today we're diving into the rapidly evolving world of AI (artificial intelligence) investing opportunities beyond the expected. Joining me today are portfolio managers of the Fidelity Global Future Leaders strategy, James Abela and Maroun Younes, together with Terence Tsai, who joins us from Hong Kong and is the technology sector lead and analyst. Together, we'll get under the hood of our AI investment framework as our experts share how they are leveraging Fidelity's research platform to uncover new and diverse opportunities.

James, let's start with yourself. If we think about the broader global investing landscape, particularly as it relates to AI and the large cap giants in that part of the market, what are some of the key observations you have? 

James Abela (JA): I think the AI trend is here to stay. Technology has been a huge trend for 20 or 30 years and it's going to be a trend for several decades. But the AI thematic today is going to transform market structures. It'll transform industries. It'll allow for high productivity. It'll lead to innovation. Also, the speed of innovation is also going to change - it’s called the next industrial revolution. Which is what I think it will be. 

So, the opportunities in that context are big. They are broad and they are large. The focus of the headlines has been mainly in the mega caps in the US, which have dominated the headlines, but there are opportunities below that. That's why we look at global small- to mid-caps. 

[The AI thematic is being] called the next industrial revolution. Which is what I think it will be.

Globally, there is energy, there is software, there are enablers, facilitators, semiconductors. So there's a number of opportunities that are available in that entire universe. 

 

LDP: Maroun, it is an exciting part of the market and an evolving part of the market. When you think about the AI theme and those large mega-cap tech names, what are some of the challenges you think exist? 

Maroun Younes (MY): Yes, there's a few. The first will be valuations. These predominantly US mega-cap names are trading at valuation levels in excess of their long-term history. If you look at a 20-year history, they're trading at levels in excess of that as a group. Now, why that's important is valuation levels give you an insight into what the market is pricing in terms of its expectations for those companies. So, the bar is very high. It's one thing to be able to grow quite rapidly when you're small, but as you get larger and larger, maintaining that same percentage growth rate gets a lot harder. You've got some of these companies now that have got market valuation levels larger than developed economies. There's a handful of companies in the US that have market valuations bigger than the entire Australian economy. So, it's staggering that these companies of this size need to maintain such an exceptional growth rate going forward to keep investors in the broader market happy. 

The second thing I would say is the nature of their business model is changing. One of the factors that investors loved about these companies historically is that they were relatively capital-like. You could grow without investing a huge amount of money. If you think about Microsoft, for example, once you created a version of Windows, you could just sell it and sell it and sell it for relatively low cost, incremental low cost. Google could just add users to its platform without a huge amount of additional cost. It was quite capital-like, you didn't need to invest a lot of money to get larger. AI is now a game that is quite capital intensive. You need to buy chips, which are very expensive. You need a huge amount of storage or cloud, whether it's your own data centres or you rent that out from third parties. You need complex LLMs (Large Language Models) etc. So, it's quite capital intensive. If you just took four companies, for example, Amazon, Google, Meta, and Microsoft, and looked at their capex (capital expenditure) profile 10 years ago, collectively, those companies did about US$23 billion in capex back in 2014. That number today is more like US$240 billion, so they've grown their capex profile 10 times in 10 years, which is quite a large growth rate. So, they now need to earn the same rates of return or very high rates of return on a much larger invested capital base, which again makes it very hard. 

The other part, I would say, is an increasing amount of circularity that's building into these companies with OpenAI seemingly sort of at the middle of this web. Some of these circular transactions like OpenAI are pledging to buy a huge amount of chips of NVIDIA for the next decade and in return, NVIDIA is investing in OpenAI. Similarly, OpenAI is pledging to rent a huge amount of cloud capacity from Oracle, and in return, Oracle is going to invest in OpenAI. OpenAI has a similar relationship with AMD (advanced micro devices), where they're going to buy a huge amount of chips from AMD going forward. So now these companies are becoming tied to one another in a way that's deeper than just the pure custom relationship way and that has some similarities, some shades to what we saw during the dot-com bubble. Back then, we had this concept known as vendor financing. If you were a cable company and you want to build out your cable network, you could go to someone like a Nokia or Alcatel-Lucent, and then you could get vendor financing, where they would lend you the money that you would then use to buy a product off them, so those companies became intertwined. We're seeing elements of that coming in more recently, literally in the last two or three months. So that, I think, whilst not necessarily bad, it's increasing the level of systemic risk, because you can have a scenario where if one domino falls, it has a cascading impact on a whole bunch of other companies. Very quickly, you can have contagion feeding throughout the system there. So, the risks are becoming a little bit more elevated.

 

LDP: Terence, if we think about this part of the market, which is very dynamic and complex to some of the points outlined there, from an investment perspective, you need to have a kind of framework. You, along with the team, developed a broader AI framework that outlines how we look at these companies in this part of the market. Can you take us through some of those key elements of that framework and some of the thinking in the development?

Terence Tsai (TT): Technology is really a lot about looking forward, but you actually learn a lot looking at the past. History doesn't always repeat, but it does rhyme. Every 10 to 20 years, we’ve seen a big technology shift or an era change. We've seen it in the mainframe era, the 50s, 60s, the PC era, and then we moved to the mobile era, the cloud era, and now probably very early innings in the generative (gen) AI era. 

The commonality for companies that participate in these technology shifts throughout these eras, can be roughly grouped into one of three buckets. The first bucket is the enablers. These are the pick and shovel plays. They provide the hardware, the infrastructure that's required for the entire technology to run on. In the middle, you have the networks, the companies that diffuse and cascade these technologies to the masses and improve the adoption of that particular technology. Finally, you have the innovators. These are your creative geniuses that think of use cases and applications for each technology era.

In the current AI era, the enablers are quite obvious. These are your mega-cap tech companies like NVIDIA, Broadcom, maybe TSMC, that are perceived winners of this gen AI era in the early innings. What is less clear is the networks and the innovators. History is only a guide. The winners of the past eras are not guaranteed a spot in the next generation of the next technology era. 

That's where it gets really exciting for stock pickers to look for opportunities in this space. In slide two, if you look at how the values extracted throughout each era, there is a timing to it. In the early innings, the enablers are the ones that extract the most value. Because think of it as laying down the railway tracks. You must lay the infrastructure down before the town and the economy is built around the train tracks. These are the ones that outperform the market early on. And we're probably seeing it right now with NVIDIA, AMD and TSMC. As the technology progresses, the networks will start to extract more value and gain more of that bargaining power because you need them to diffuse the technology to the masses, to enterprises and to consumers. Finally, you need innovators to be able to really make the most profit such that it can feed the money or the value down the chain to the networks and the enablers. You need each level and each layer of the stack to be profitable for the entire era and the generation of technology to work. 

AI isn't new, it's been around for a long time. But the reason it's become impressive is because they threw more compute at the problem.

 

LDP: If you take that framework that you described, this concept of enablers, networks and innovators, there’s a range of sub-themes that sit underneath those three broad categories. Can you give some colour to some of those sub-themes in the enabler category?

TT: For the enablers right now, it's all about compute, right? AI isn't new, it's been around for a long time. But the reason it's become impressive is because they threw more compute at the problem. So, compute is driven a lot in the headlines by NVIDIA, because they're able to produce the most powerful and performant GPUs (graphic processing units) out there. These are extremely costly, like Maroun said, they cost between US$40,000-$50,000 a chip. In technology, the game is always about bringing down the cost. Because when you get costs down, adoption goes up and as these chips get more and more expensive, the cost of doing business is getting higher. That stands in the way of AI being more widely adopted. 

One of the themes that's come up is around advanced packaging. Packaging historically has been a very low margin, labour intensive business. It's the lowest margin business across the semiconductor ecosystem. But because Moore's Law is slowing, Moore's Law has been the tailwind for technology over the last 40, 50 years, it means is you get twice the performance for the same cost. That's why our iPhones are 100,000 times faster than the rocket that sent Neil Armstrong to the moon. But it doesn't cost, we don't have to spend a defence budget to pay for that iPhone, again because of Moore's Law. But Moore's Law is slowing. Moore's Law is not as effective in bringing the cost down. It can still push performance, but the cost is not coming down.

Right now, because we are in the early innings of the broader adoption of AI, the enablers, they're having their moment.

Enter then advanced packaging. This helps you get better bang for your buck. It's another alternative, it's another method to get performance at a reasonable cost. So, there are companies that have been ignored or sectors that have been ignored that are seeing the value shift to them as we look for higher compute at a reasonable cost.

 

LDP: Maroun, if you think about what Terence just described, there's a lot of moving parts, and being a portfolio manager, you're trying to distil a lot of information. How are you leveraging this AI framework that we use at Fidelity in terms of identifying opportunities and maybe touch on some examples?

MY: Terence made some wonderful points there. I think the first is that sequential nature of the timing where it's the enablers followed by the networks, followed by the innovators. Right now, because we are in the early innings of the broader adoption of AI, the enablers, they're having their moment. There's a couple of companies in our portfolio right now that play into that space.

Terence mentioned advanced packaging. One of those companies that we own is a European company called BE Semiconductor Industries, which involves a technique called hybrid bonding. It used to be quite costly to do, but now, as Terence mentioned, with Moore's Law slowing down, that's becoming a cost-effective solution so we're seeing increased adoption of that.

Another one is STMicroelectronics, which is a joint venture between Italy and France. They provide analog chips that get used in industrial end markets, which is predominantly autos, very power efficient way of doing it. It solves a lot of the cost issues that Terence alluded to.

Another is NRG Energy, which is a US Texas based utility. We're seeing a huge build-out of data centre capacity across the US, particularly with the Stargate project. Some of that is going into Texas. You've got a huge amount of land in Texas and it's relatively flat, so it makes it ideal for data centre capacity. Those data centres are enormously power-hungry, so you need utilities to provide more and more power. In addition to powering households, you need more and more power just to power these data centres that are going to come online over the next 5-10 years. 

Historically, we've also owned a couple of other companies. One was Arista. They make network switches. If you think about a cloud or a network and you have different devices that talk to each other and transfer information from one to another, there's an enabler there called a network switch. So, Arista, along with Cisco, are the two market leaders, but Arista has historically been the leader in the hyperscale segment. Their two largest customers are Microsoft and Facebook (Meta) so they're leveraged into that. We unfortunately had to sell that because it was too large for us and a success story for us. 

Another company we've owned in the past is called Vertiv. They do liquid cooling. If you think about data centres, they create a huge amount of heat. Just think about your normal PC when you use it or your phone, if you use it for an extended period, it can get quite hot. Imagine an entire shed of just computer equipment working 24/7. You need to come up with innovative ways of cooling these things down, cost-efficient ways of cooling things down. Vertiv was a company involved in liquid cooling. 

The innovators will likely come further down the road. As Terence said, they extract quite a bit of value, but they build on top of everything that comes before them. So, we're keeping a keen eye on the innovators that come through over time. But right now, we're definitely playing a lot in that enabler space.

 

LDP: Terence, are there other observations from your side?

TT: I think a lot of the innovators have not really come to light and that's normal because a lot of the innovating companies were founded after the dot-com bubble burst. Who knew that we used to have to get on the internet at a specific physical spot in our homes? Who knew that you could get into a stranger's car and have them take you to your destination, or have food delivered to your homes, or check out other people's vacations on the internet?

I think these creative use cases may not have come to light yet and that's why it's important for us to have our research team on the ground to locate and to discover these innovative companies, which start off as small- and mid-cap companies and eventually become a winner-takes-all in the gen-AI era.

The part that's interesting and maybe underappreciated or overlooked by the market is China.

 

LDP: Terence, staying with you. One of the interesting things in recent years is that a lot of the discussion around AI has been very US-centric. A lot of advisors may be familiar with the Mag7 stocks, stocks like NVIDIA etc., which really got the headlines. But outside of the US, Asia is an interesting part of the market in relation to AI and it has made a lot of progress in that part of the market. Can you give us a bit of an overview of the Asian market and what are some of the opportunities you're seeing and some of the risks in that part of the market?

TT: You're right, and James mentioned in his comment earlier that the companies that have been getting the headlines in this AI era have been mostly the US mega-cap companies. If you look at an index like the MSCI World IT Index, you have the US companies that are 82% of the benchmark. You have Apple, Nvidia, Microsoft, that's half the benchmark. The US is not 82% of the world's innovation. That's not a good representation of what the global tech industry is. 

In Asia, I can roughly divide it into two big categories. You have the Taiwan/ Japan, and Korea markets. These are very highly correlated to the US markets, mostly because these are enablers or the pick and shovel plays for the US customers, like NVIDIA. I guess the headline companies are TSMC in Taiwan that manufactures 95% of the world's most advanced chips. You have Japan, which sells the equipment to TSMC, and you have the two largest memory companies in Korea, Samsung and SK Hynix, that sells high bandwidth memory to pair alongside the GPUs (graphics processing units) that NVIDIA creates to sell to the hyperscaler customers. So, those companies have performed really, really well in line with their US peers. 

The part that's interesting and maybe underappreciated or overlooked by the market is China. China's AI and tech ecosystem has largely been left for dead. If we had this conversation a year ago, you would see that the prevailing assumption or the prevailing view on China is that because AI is driven by compute, and because the US government has export controls on the best, latest, and greatest AI chips to China, that China would never be able to catch up in AI. That's why US exceptionalism makes sense. But earlier this year, when our team went to China and we did on the ground research and met with the companies, we found out China was doing AI in a very different fashion from what the US was doing. 

Then we had the DeepSeek moment. Coincidentally, before it became mainstream, we were already discussing this internally at Fidelity. After the moment, the view on China's tech and AI ecosystem has been improving and has been changing. China's doing it cost-effectively. China's pursuing an open-source system, as opposed to the US, which, ironically, open AI is a closed-loop system. So, I think it's the opportunity set and the innovation that's coming up from China that deserves more investor attention.

 

LDP: If we look beyond Asia to some of the other regions, are there any interesting opportunities, interesting regions that you're focused on as well?

TT: The one that we haven't talked about is Europe. Unfortunately, in market circles, the joke has been the US leads in AI, large language models, and Asia leads in the manufacturing, and Europe leads in regulation. But that's probably an exaggeration. There are very outstanding companies in Europe that we monitor, and we look at as bellwethers. ASML is one of them, the monopolistic company that provides lithography, that is used to shrink transistors to make advanced chips. You also have very innovative software companies like SAP that's already been using AI in their ERP (Enterprise Resource Planning) systems and enterprises. So those are areas that we continue to spend time on to look for opportunities. But right now, Europe's been taking a backseat to the US and Asia.

 

LDP: James, how do you think about the Global Future Leaders strategy in terms of a broader advisor portfolio looking to get exposure to that AI thematic?

JA: It's very much what Terence has alluded to. Enablers and facilitators are really where you can get a lot of exposure to AI, especially mid-caps and small-caps. 

It is going to be a very dynamic market. Market structures are going to change. One example of that is NRG Energy and the US energy space, as Maroun mentioned. The mega-tech companies in the US are now signing 10-year contracts above spot price, 20%, 30%, 40%, 50% above current spot price to lock in 10-year contracts with data centres to make sure they get the supply of power. Now, that's changing the whole market structure and the return profiles just in the power space. So, that power space example extends into semiconductors, enablers, facilitators, and eventually, as I've alluded to, into software businesses as well that are going to facilitate the use of AI and the diffusion of AI into much bigger markets and into our lives and our workplaces. 

There's a small number of large companies that have dominated the headlines. But there's a lot more happening below the surface.

So, that breadth and depth of opportunity set is available. It is growing, but it is dynamic and it's changing. So, we need to be checking in with people like Terrence every day, listening to what is happening, because it's a very dynamic market.

 

What's the one key take out you would like advisors to leave with?

MY: For me, there's a small number of large companies that have dominated the headlines. But there's a lot more happening below the surface. So definitely stay diversified beyond just the mega-cap tech names. Because there's a lot of dynamic opportunities building up in the small- and mid-cap space, and they will be attractive over time.

TT: I would just add on to what Maroun has said and say diversification in geography. There are areas outside of the US that have interesting opportunities and attractive investment opportunities for us to monitor. Too, I guess beyond tech, right? We've talked a lot about tech. If history is any guide, companies or industries that are quick to adopt technology in their workflow, in their businesses, are usually the early winners within their respective subsectors. For example, your e-commerce companies within retail. So, within your industries, look for companies that are successfully adopting technology to either increase productivity or drive better customer acquisition and those could be the winners of your industry beyond technology.