Subscribe to the Alger Podcast in
Spotify | Apple Podcasts
In this conversation, Redwood portfolio managers Mike Mufson and Ezra Samet discuss potential artificial intelligence (AI) investing opportunities outside the U.S.
ALEX BERNSTEIN: Hello, I’m Alex Bernstein and you’re listening to The Alger Podcast, Investing in Growth & Change. Much of the current conversation around AI has focused on U.S. companies — where the best investments may lie and what risks could derail the story. But what about opportunities outside the U.S.? Here to discuss that with me today are Alger Redwood Portfolio Managers Mike Mufson and Ezra Samet. Mike, Ezra, thank you so much for joining me here today.
MIKE MUFSON: Thank you for having us.
EZRA SAMET: Thanks, Alex.
ALEX: So, Mike, let’s start here. What you, Ezra and the Redwood team have done is create a comprehensive thesis around an entire AI equity cycle. Tell me what that means.
MIKE: Sure. I think what we think is this AI cycle is going to be quite significant. This is one of the game-changing technologies of the last decades. And one of the questions is, do the stocks get ahead of the business, and how is the business playing out? And what we’re trying to do is measure the key drivers around the AI cycle. What are the things that we can look at to figure out are AI stocks still attractive? Even if the AI business is good, are the AI stocks attractive? And so we set up a framework with four pillars to build around that.
The first is financial. The financial economics of the business. Are they generating good returns, high profit margins, good growth? Second, is on funding. A lot of these businesses are early stages. They don’t generate cash, they need cash to build out the infrastructure to generate the revenues, and can they get that funding. Third is on infrastructure, and we want to understand whether they can put the infrastructure in place and execute on the business model. And then fourth, is the regulatory and political environment, and whether that environment is constructive to build out the AI environment.
EZRA: I think one of the reasons that we created this framework, we are believers that AI is a transformational technology. We can see it, I think, in our own personal lives as consumers, in our own professional lives, that the power of these tools to accelerate certain processes and automate certain processes is massive.
One of the questions that we kept asking ourselves and we kept getting in every meeting that we hosted was, is this a bubble? And we wanted to create a structure around thinking through that question and how we might address it. And so, we started to ask ourselves, what are the conditions under which such a bubble could burst, right? We think, looking at historical technology cycles that this was not going to be due to valuation. We think the bubble would not burst because the pure play AI stocks were expensive, that valuation is a consideration once a bubble bursts. And so those four pillars that Mike described were the kind of aggregation of all of those different potential sources of downside.
MIKE: And even if you think about some of the big trends in technology or health care or any industry, the stocks don’t go straight up. There are periods where they can be great for a few years, have a correction and then they’re great again. And so we’re trying to invest along that cycle.
ALEX: What drove the need to create this equity cycle framework? How did it develop?
EZRA: We found ourselves discussing the state of the AI trade and the AI theme almost daily and we were moving in circles. We would reference the same potential risks, what if we don’t have enough power, what if there’s not enough water, what if there’s political pushback, what if large language models become commoditized? And so, then all the value that’s being accrued to these private companies, Anthropic and OpenAI and others, could be lost, and then the funding might dry up. And so, you can sort of go in circles endlessly.
We know that hundreds of billions are being spent on this theme, but we think it’s not yet clear, frankly, where the long-term value will reside. We realized, what we needed was a framework. Something that we could build up front, agree to as a group, and then monitor in a disciplined way so that we weren’t blown about by the winds of the market.
ALEX: Do you think U.S. investors are so focused on the mega caps that they’re missing the greater global picture?
MIKE: What we’re finding in the non-U.S. space is two different things I think are worthwhile mentioning. One is sovereign AI, and the second is infrastructure support, which is mostly done outside the United States. So, what I mean by that would be the foundries.
So TSMC produces all of the chips for NVIDIA, and that’s based in Taiwan. And then we have, the DRAM producers are predominantly in Korea. And the flash memory companies are in mostly outside the U.S. as well, in Japan and Korea. So, the foundries are becoming an increasingly important component of the AI infrastructure buildout. And that is one of the areas that’s really dominated by non-U.S. investing.
We think it’s getting much more expensive and much more complex to build the next generation of logic chips or the next generation of DRAM, the next generation of flash memory. And we think that is reducing the number of companies that can compete in each of those businesses.
ALEX: Let’s talk a bit about Sovereign AI. What does that mean and how does that play out in the conversation?
EZRA: Sovereign AI, I think there’s different ways to conceptualize it, but if artificial intelligence is going to be the transformational technology of the next era, we think it will underpin and power much of the most important functions within the global economy, but certainly also within sovereign nations, right.
As we think about the role of AI, we think governments will become increasingly dependent upon it. And the state of play today is that almost all of the large models that are gaining traction have been created by American companies.
We think that it’s increasingly untenable for these other nations, for our allies, just as much as our adversaries, to be entirely reliant on U.S. technology. And we think AI accelerates that process, which was likely already playing out. So, for that reason, we expect that governments, as a matter of national security and of sovereignty, will need to invest in their own infrastructure. And that doesn’t mean that they’ll replicate every element of the AI stack, but I think to the degree possible, there’s a very strong incentive to drive that investment forward domestically. So, we expect sovereign wealth funds, we expect specific government budgetary items and carve outs to fund the development of many different components of the AI, technology stack across the world.
MIKE: We think there’s four areas that these governments want to influence over in sovereign AI. One would be the compute. They want the data centers to be in country. Two, they want the data sovereignty so the data also stays in country, it doesn’t get transferred across borders. Three would be on the models. And, yes, they’re probably not going to create their own model. They may take an open model and tweak it and make it country specific. And then four is also talent. They need to have domestic skills. They can’t be reliant on another country or on another partner.
ALEX: We had that moment last year with China and DeepSeek. Some investors panicked, while others viewed it as a potential positive for all AI companies. How do you look at it? Do you think a high tide lifts all boats?
MIKE: So, I do, but I think that created a way to do AI with less capital, right. And if you can get more of it for less, it drives the cycle. I think what DeepSeek did is it opened the eyes of governments around the world, smaller countries that said, oh, now we can get into the sovereign AI game. But it doesn’t mean they start in day one. They get the idea that, yes, we should do it. Now they’ve got to pass some legislation, they’ve got to assign some capital. They’ve got to hire people. Plan it out. And we think that probably takes a year. And so that’s one of the reasons we think the timing around Sovereign AI being a 2026 into 2027 acceleration is because of the natural flow of how governments work around the world.
And we already have seen a lot of announcements. The U.K., France, Singapore, Korea, even small countries like the UAE has allocated $44 billion for AI build out. Saudi’s allocated $20 billion for their Hexagon Data Center. So, the projects are getting underway right now.
ALEX: I’d like to take a few minutes to talk about the kind of name that stand out in your framework. Which companies currently excite you?
MIKE: I think SK Hynix has started to separate itself from the crowd in the DRAM space. SK was the first company in the High Bandwidth Memory 3, which was critical for AI. They were the first one to 3E, which was the next generation. The current generation is 4, and they were the first there. I think it’s the first time where SK Hynix has been the leader and not a follower to Samsung. As a result, their capacity is completely sold out for 2026. It looks like it’s mostly sold out for 2027 already. And their gross margins have now reached the 70 percent range. They’re above TSMC and the profitability levels are very high, yet the valuation figures are only 4x earnings.
We think the view of SK Hynix is the old view, that it’s a commodity business, that anybody can get into it, that if the three of us raised total capital, we could compete at the highest level of DRAM. I think that is no longer the case.
ALEX: Ezra?
EZRA: I’ll mention EssilorLuxottica, which is really an 800-pound gorilla of the eyewear industry. They have partnered with Meta to bring the AI Meta Ray-Ban product to market. They’ve now expanded that to Oakley. And so, this is not a digital company by its founding or by its legacy, but we think they have leveraged their position as a critical global supply chain into this to this leadership role in AI wearables.
Another one is Tencent. Tencent is building its own family of models and its own infrastructure and its own chips, et cetera. Tencent is largely video games and WeChat and then a financial business. We think Tencent is positioned really uniquely on a global basis to benefit from AI. AI is already reducing the cost of producing their games. It’s enhancing the game play by turning these otherwise scripted characters that you interact with through the games into interactive LLM powered characters, and so I think significantly improving the way that the game functions.
And so, we think Tencent has a massive scope across which to run AI and to help you book hotels and order meals and drive your car. So we think it has an incredible reach across the consumer life of a Chinese citizen in a way that I think really no U.S. company has. And so, I think that’s another, another exciting beneficiary.
ALEX: Last question. What do you think differentiates how Alger Redwood invests in non-U.S. AI versus other similar managers?
EZRA: I think one important differentiator of our approach is thinking not only about the thematic opportunity but then marrying that with a really detailed fundamental view from the bottom up to try to understand which of these businesses have structural advantages that we think cannot be overcome or undermined by competition, and try to make sure that our exposures are aligned with that fundamental bottom up investment approach.
MIKE: I’d add that we’ve created a culture around trying to develop mis-forecasted earnings or out of consensus projections and it’s the same investment philosophy and process that we started the firm with in 2005, and we continue it today.
ALEX: Mike, Ezra, thanks so much for joining me this afternoon.
MIKE: Thanks, Alex. Great talking with you today.
EZRA: Thanks, Alex. It’s great to talk to you.
ALEX: And thank you for listening. For more information on Redwood and non-U.S. AI investing, and for more of our latest insights, please visit alger.com.
