Hello and welcome to The Download. I'm your host, Dave Richardson, and it is Tech Day. Or maybe we're going to take a little turn in a second. I've got our good friends Rob Cavallo and Marcello Montanari here to talk life science and technology, not just tech. The big news this morning in the tech world is I have my phishing and cybersecurity score here at our beloved employer up to over 900 after being a loser and almost getting my job taken away because I could not spot phishing emails. But now, I've become a cybersecurity expert. We'll get into cybersecurity later on. You're both probably pretty good at those tests, right?

I'm on a good monthly streak right now.

I don't remember the last time I saw one.

Okay, you need to work on your awareness. I just did a couple of awareness courses. It improves your score. I know you to win.

Yeah, yeah, I've done those tests.

There we go. So on most of the podcasts we do with Rob and Marcello, we tend to focus on the exciting things around artificial intelligence, semiconductor software, quant computing, nuclear energy, cool stuff. But today, Marcello is purely here for eye candy. Because we are on YouTube, by the way. Subscribe to us on YouTube. Because the hot area right now— and you would know this if you listen to their last appearance here and you get all that downloaded. If you subscribe to The Download wherever you get your podcasts, And we'd love some reviews because we're a little low on reviews. Again, my mom's getting tired. The arthritis is hard. Makes it hard for her to type in those reviews. But Rob, healthcare, life sciences, just on a great run. And not only that, a great run from a societal perspective as well. There's some really incredible breakthroughs happening that have been just announced over the last couple of weeks. So, why don't you take up what's going on in that space? And what you're most excited about and what should investors be excited about?

Yeah, it's been quite the run for healthcare. If we would have been doing this a few months ago, we'd be saying, well, healthcare once again is just bringing up the rear in the overall market. But actually now, year to date, the healthcare sector is ahead of the S&P 500 as of last year. So it's been a remarkable turnaround over the last 3 months. I think initially what happened was, as some of the AI trade was getting a little bit fatigued, and the market was looking to broaden out and look into new areas where to go to, healthcare was really a bit of an anti-AI trade— and I'll get into why I think that's the wrong way to think about it long term— but it benefited a little bit from those flows out of technology sector and from the most AI-exposed names. And what happened is not only did you get the initial flow of money back into that sector, but then you had good sets of results. Lots of M&A happening in the space. Regulatory issues are relatively quiet for that whole sector, which is a very good thing because historically that’s something that tends to becomes a headwind. But it's actually been a neutral and a positive right now for the group. And then we've had some really good data come out from some exciting trials across the pharma and biotech space. And this has really reignited this fear that was once just an anti-AI trade. There's actually probably more legs to why the sector could work for a more extended period of time for some of the reasons I mentioned, but also because— and we've talked about this on this podcast before— is over the very long term, I would argue healthcare is probably the biggest beneficiary from the deployment of AI. So as we get closer to more benefits from that stage of the value chain emerging, I think healthcare will continue to benefit.

Rob, let me stop you there because I'm constantly out and about talking to investors and I use that line as likely the, at least the earliest, but one of the biggest beneficiaries is going to be healthcare from artificial intelligence. And people go, well, why is that? And I say, well, Rob told me. And so, Rob, why is that?

Oh, so now I have to elaborate on it?

Well, it could just be «Rob says». I take that as good to go.

But I can give you a few thoughts why— and this is both in therapeutic development, but also just in how patients are served. One of our theses behind the power of AI over the long term is having the ability to take the tremendous amounts of data we have in different fields and actually be able to use it more effectively, use more than just the 10% superficial data or whatever that number might be in a lot of cases. Healthcare has so much data, whether it's in drug discovery, whether it's in patient management, life management, that you cannot really even scrape. And AI has the ability to take that data and create real informed decision-making and so forth that really help the patient. So when I look at the ways that lives of patients can be managed, it's better understanding the right therapy for patients. Because right now you take lots of blanket approaches to how patients are treated. But as you're able to comb data and really characterize a patient's individual circumstances more effectively, you're going to define and be able to give better treatments, better outcomes to these patients. From a drug development perspective, it takes a long time and a lot of money to develop drugs. Now, I'm not this AI hysteric that says that you're going to replace inhuman trials. That's going to remain the core part, and that takes time. But the ability to define upfront the right molecules that have the highest likelihood of success in getting them into the clinic earlier, that time is going to shrink dramatically. That cost is going to shrink dramatically. And you're going to see way more effective and higher success rates start to come out of this over the next 5 to 10 years. And that's what really makes me excited, just the fact that there's all this data and all this opportunity across wide swaths of the sector and the industry that they're going to be able to benefit from. That's just another level relative to some of the other places across the economy.

Yeah. And with an aging population certainly in the Western world— Western Europe, US, Canada, but throw in Japan. China now has an aging population. And this is important stuff for people for whom, as you age, little things happen. I've talked a lot about the little things that are happening to me on the podcast. And this is going to help a lot.

For sure. The demographic tailwinds have been an argument for a long time, and that's only going to become a bigger driver of the sector over the next decade. But we also have to think about the cost of managing that. And if we can do it all more effectively, you'll see the demand for services go up and you'll be able to find the cost offsets where things are not working or where waste is happening to fund that. And it's going to be better not only for the individual companies that we invest in, but just for society in general.

And these stocks had been ignored for a long time?

They had been ignored for a long time. Healthcare is a lot of specialist investors. It hasn't really been ignored from the specialist community per say, but the big pools of money that have been able to just ignore the sector for years have actually started to percolate again. That's probably the next leg, is when these big pots of money that have just ignored the sector for a long time, start to come back in because the earnings picture is better. There's excitement around AI and just the headlines continue to be much better. As long as the regulatory environment remains tame, I think that the sector could be set up again to continue to be a very good performer.

And there's got to be an advantage for regulators or governments, if you can reduce costs in healthcare. Then you can solve some of the debt issues that we're seeing really creep up, and I'm sure we'll get into that with you, Marcello, in terms of these higher rates and the issues that they're causing in terms of stock valuations, particularly in the technology space. But this could reduce costs for governments if we can manage healthcare more effectively.

For sure. I'm not going to go as far as to say the government's going to actually reduce spending. That feels a little ambitious. But at least from a sector perspective, if you're finding cost offsets and better cost management within the sector, the pressure of trying to find cost has probably reduced. So if you're able to find the cost offset, you're probably going to be in a better position in terms of how the regulators and government come after you from a spending perspective. Again, that easing regulatory environment is generally a good setup for the sector.

And the news over the last week. You were going to get into some of the things that you think are particularly exciting and I cut you off. What are you most excited about?

So last week there was a big cancer vaccine announcement from Moderna— which many of the listeners might remember from the COVID vaccine days— with their partner Merck. Both stocks were up tremendously. I think Merck added over $40 billion of market cap on the day of the announcement. Moderna, not dissimilar. We can argue how much is maybe a little bit overdone, but in general it was a great excitement because there's this push forward in this cancer vaccine. Now when I say vaccine, don't think of it as a vaccine you would get for the flu or COVID. It's just a way that they're able to use this therapy to treat cancer patients. Now, this was done in a cancer called melanoma, which is essentially skin cancer, relatively earlier stage patients. But the results were very good working in combination with an immunotherapy from Merck. And it just created this excitement that from Moderna's perspective, this is a new, potentially large opportunity. But just in general, it opens up a new avenue for this type of personalized vaccine treatment within cancer, and maybe opportunities beyond just melanoma into other important tumor types non-small cell lung cancer, etc. So just another win for the group, another win for just sentiment around the space. And it was great for the stocks, great for patients, and just a win-win.

Yeah, it's kept some of the momentum going, which has been very strong since 2 times ago when you were on and we checked in last time and things were moving well and it's continued to go in that direction.

I try not to talk about it. Every time I say I get excited about healthcare as a sector, something happens. So, I'm going to try to temper my excitement a little bit, but I feel optimistic here.

Yeah, and it's one of the reasons why I think you manage your portfolio the way you do. Obviously, it's tech in all senses but that layering in of the life sciences as a particular focus in your portfolio. Sometimes you say, ah, geez, when it's underperforming, it's a bit of a drag. But again, you see the connection, you see how it's all intertwined, and you can see the value of having the life sciences attached to it.

Yeah, absolutely. I mean, they work together sometimes, but a lot of times the power is that when one is not working, the other sector probably is. So you get a nice balance between the two. And really, they're both innovation. And that's the message we try to get across with Life Science and Tech Fund, is that it's about innovation across all parts of the market. And it just happens to be that healthcare, tech, life sciences tend to be the biggest contributors to innovation. A nice balance, but at the same time it can also work together as just an innovation theme in itself.

Sounds you're taking a shot at Marcello. Marcello's not working right now. Or are you working? You're just working in different areas, right, Marcello?

We work together, but we have our sectors delineated. But he does all the healthcare. I'm just not smart enough to speak on that stuff.

Healthcare is for smart people. There we go. So, there's the tip of the day Marcello always gives whenever we have him on. But things are changing in the broader tech space. The AI buildout is still happening, but just what the market is rewarding seems to be moving around. It's a fairly volatile area. And then you've got the concerns around debt. So what are you thinking about right now? What's working? What's not working in that space?

The market's taking a period of consolidation in terms of the picks and shovels. They've slowed down a bit and given up some ground. And that's everything from memory to semiconductors to some of the optical names. And we've just lived through this period where there was a view that all software was dead and they'd just been selling off just on the belief that AI was just going to negate the need for software going forward. Those started selling off last August, I'd say, and really took off. When Anthropic came out with the coding, I think it was Opus 4.5 or 4.8— I forget which one. There's been so many of them. But the one in the middle of December, basically. That really just accelerated the sell-off in software. So it sold off into the beginning of this year. And since then, we've seen quite a bounce. The entire sector, whether you're looking at it through the IGV, which is the ETF that has cap-weighted software companies, that is above both a 50-day and 200-day moving average. And then there's another one. I think it's the IWM, which is basically an equal-weighted software index. That one's actually been outperforming. So every name in there is 1%, and that's been outperforming. So we've had a bounce.

And those stocks were down a lot. 30 to 50% in some cases.

Yeah. Even big bellwethers SAP from Germany, things that you'd think are rock solid, you don't really need to worry about. A big hit, and it's since bounced back about 50% from the lows. What's driven that is a number of things. You can cut software into a whole bunch of different subsectors. You’ve got security software, you've got what they call consumption software, and that's comprised of a whole bunch of infrastructure names; Snowflake, MongoDB, those type of names. They've been showing really strong results. So that's helped. The security space. At first, when the Mythos model came out from Anthropic, everybody was freaking out. But then, take a step back, think about it, and go, actually, this is really good for security software companies. So they've been on a tear on top of that.

Just coincidentally, sometimes the listeners think that my comments at the start of each podcast are superfluous. There was a reason why I brought up cybersecurity and the significance of it, because those stocks have been on fire.

Yeah, absolutely. And then, on the margin, the bull case for software is that they can use some of the AI tools to basically create agents that work in conjunction with the software. And so you've had a few things on the margin where companies have actually demonstrated that. One that comes to mind is Manhattan Associates. They're in the warehouse automation business, but they created all these agents to basically layer on top of their software. And the customers started using it. They're not paying for it yet. They go through a trial period. And then if they're happy with it, they sign up and that's when they start paying subscription fees. About 10% of their base has actually started using these agents and 100% of the clients that have gotten to the end of the trial period have signed up. So that's a really good sign because there was a belief that AI would do all of this. Instead, the software companies say, we understand this space, we have the domain knowledge, we know what we're doing here. We'll put the agents on top of the software.

And just before you go on, the first place that I heard of Agentic AI and AI agents, that was a conversation with you on this podcast. And when you had initially talked about how this could work and could enhance value and efficiency and productivity, it was around layering it on top of software. And then just all of a sudden that disappeared. Was it just simply that the AI was going to be able to do everything? And now we've learned that it's 1+1=3, when you layer AI onto the software.

Well, there's nuances here. Some companies are better prepared for this than others. And we're just at the initial outset of all this and we're seeing how it's actually playing out. So it just so happens that the ones that I'm mentioning, they're a little bit more advanced. There's another company in the CAD space called Autodesk. They're on their third iteration of AI models that were specifically designed off of designs, basically. It's basically all computer-aided design stuff. So their models are based on all of that. This is not stuff that you can scrape off of the internet. This is very proprietary to the companies that have all of this data. So they're on their third iteration. Company that's in the semiconductor space in terms of design, like Cadence, they're adding agents as well, and that's starting to take off as well. And then there was a big story about how a chip was designed. I forget with who they designed the chip. Maybe you can fill it in after, but they said that they used AI to design it, and then once you dug down into it, okay, the AI helped frame what they wanted to do, but then it went into the tool and built it on the tool. Like Cadence and Synopsys, they're the tools for this type of stuff. So again, you get a confirmation that AI plus the underlying— basically you're turning some of these existing software platforms or applications into tools for the AI agents to use— and then, the AI agents interact with human beings and so on and so forth. Anyway, the market's starting to discern the winners and losers within the software space. And so we've seen some interesting things. I forgot the most important one, which was Microsoft. Their Copilot subscriptions is just flying up and to the right and much faster than anyone expected. So all these things are adding up to an environment where people are discerning the winners and losers.

Yes. As Rob said, they're showing they can make money, that there's money here. Which was one of the big questions. How much money and when it's going to come and who's going to make it. But as I'm out talking to investors— I spent most of the last 3 weeks out doing things directly with investors and advisors— and then I sit down with you guys and this just moves so fast and it's so hard to keep up with. And then as I would say to people— we had David Tron on last week— you just realize the amount of effort and the work and the expertise you have to have to be involved in this sector. You can play around with it and you can buy stocks that are going up— we had the conversation about speculation and gambling versus investing last time— and it just seems that a lot of people are doing more of that, the speculation and gambling, and they ride a wave and then they're surprised when they're riding the wave and they get knocked off the surfboard. The surfboard whacks them on the head and the undertow takes them out to sea. And that's just something that your expertise helps avoid because the portfolio you're building makes sense in terms of the way it fits together, the risk management. And risk management is the key.

Yeah, that's why we always say diversification versus concentration. We know things change. As you pointed out, especially in this sector, they've been changing so quickly. So it's important to be diversified, not get too negative on different sectors and not put all your eggs into one side of the boat versus the other. And so we've seen that with, like I said earlier, the picks and shovels, the optical, the memory, all that stuff has come under pressure recently. After having amazing runs— some of these names, just up in spectacular fashion. So they're checking back a little bit and marking time to the next leg up. Because we don't think that this buildout is over by any means. We have not seen any evidence that this is slowing down. In fact, there's times where we sit there and go, wow, this just keeps accelerating. Anyway.

So let's do a little pulse check. Because we generally do that, not as specifically as I'm calling it out right here, but versus even 6 months ago, what are you seeing in terms of the amount of money that's going to be spent on this buildout and how long it's going to continue? And then how much money is there? Have we gotten to the point where we understand the scale and the scope of what's going to be made in this space?

I can start. I would say we've gone through fits and starts. We came to a period where, how are we going to ever raise the amount of capital that's needed? And things have taken place, whether it's some of the funding mechanisms through the chip suppliers like NVIDIA and AMD and so forth, whether it's some of the private credit structures that have been announced, like the partnership with NVIDIA and the 6 large PE firms. So some of the funding concerns have eased a little bit. In terms of the absolute dollar, it's hard to say with confidence beyond the next 2 years— we feel good about a longer period than that, but for the next 2 years, the visibility is pretty high of what the spend is going to be. And the spend is going higher. Probably at a decelerating pace, not at the same level of growth that we've seen just because you run a little bit into the law of large numbers. But we don't see anything in the next couple of years at least to suggest that that spend is going to be curtailed. I mentioned the funding. The other thing that gets pushed back is, are there returns for the big spenders? Maybe we'll get into that in more detail, but last quarter we saw some strong lines in the sand that the returns are there and some good explanations of how they're going to get better. And the market got comfortable, discerning between winners and losers even within that spectrum. But that has faded a little bit. Although it ebbs and flows how important that is to the market narrative. The other big thing to really watch, the issue that could derail some of the spending in a shorter period of time is the regulatory issues. You've seen a number of states issue moratoriums about slowing the data center build and it's becoming a bit of a platform issue in the US midterms in 2026, with the potential, depending on how that goes, to become a bigger platform to 2028 presidential election. So if I was going to make a call today as to why spending would be worse than our expectations over the next 2 years, it's probably on regulatory issues. That's one of the things we're watching really closely now.

So the money's there to spend today. But it might not even be allowed to spend it or able to spend it because of restrictions around where you could build a data center. If you turn on the US television and you get on an IPTV and you're watching stations from areas where it's a little bit more competitive from an election standpoint and you see the ads being run that are very much anti-data center. I think the governor of Pennsylvania just announced a moratorium— and that’s a fairly large state. So I may be all ready to go, but I'm just not allowed to.

And you've seen Texas, of all places, even Texas is understanding that it's a platform issue potentially and there's concerns. That's a big risk. We've generally been of the view that the funding is there, especially at the hyperscaler level, and some of the other things that are happening are going to be good for some of the non-hyperscaler spending. So, that could change. The market could say we want more returns, we want you to pull back on the level of free cash flow or debt that you're piling on or equity raises that you're doing. So the market whim could change at any time. But I'd say as we're sitting in the seat today, some of the regulatory concerns are probably taking a bit more of a front seat as to what derails that spending in a 12- to 24-month period. But again, it's hard to pull apart what is a talking point into an election versus what's going to have more teeth to it. These moratoriums today are also not that we're stopping building. It's just, let's have a discussion. And discussions can go in a lot of different ways. So we're watching it closely because that's a real risk to think about.

On that note, I saw an interview with Josh Shapiro. And there was a bit of nuance around what he was doing. He basically said, there's just so many players now. And we have this concept called «fill and kill». Whenever you have a new ecosystem, you get speciation. All of a sudden, all these species come to fill it in. We know about all the big hyperscalers and some of the big neoclouds, but there's literally hundreds, if not thousands of other smaller players that want to build data centers, and they're making all sorts of pronouncements and putting their hands up and making claims on power and stuff. For the most part these guys will never see the light of day. But it's out there, and it's adding to some of the concerns around all the data center noise that's out there. So there's the real players who have a long track record of building data centers, working with communities, and then there's the other guys who are just a bunch of wildcatters out there trying to do it.

That's almost the right analogy, isn't it?

Yeah. Let's be clear, there's concerns across the board, but the wildcatters are probably adding more than that needs to be there. But to date, I'm not so sure that any of the big players have said that they've been really curtailed by anything yet.

And I don't know if any of the pronouncements are really going to impact the level of spend in 2027. It's going to be how are we thinking 2028, 29, 30? Some of it's going to be how are they addressing power concerns and how they're filtering power into their projects. So it remains to be seen, but clearly, we're trying to balance out where could we be wrong, where could there be risk. That's going to be an area where it's a possibility. So definitely one that we're spending time just to try to better understand that landscape.

Josh Shapiro, governor of Pennsylvania, just to make the connection. He's been out with some pronouncements. And again, we're seeing this across the US. It's created an opportunity for Canada anyway. We had a big deal last week with Quebec and Newfoundland for hydropower generation in Labrador.

That's massive.

14 gigawatts. I mean, we're cold. We got lots of water, lots of hydroelectric power. Canada can come in and take advantage of this. And we got lots of land too, I heard. So it's a pretty big place. Not too many people. It sounds perfect.

In theory, we should totally be there. In practice, let's see if Canada can build things.

But I think there's a lot of the same concerns. For sure, Canada should be at the forefront because of that power point that you brought up. But anecdotally, there's a very small data center not far from where I live. And the community's in an uproar for many of the same issues. And I think the problem is the tech companies and the data center, they just have not done a good job evangelizing why we should build these and addressing some of these concerns up front. So in theory Canada should. In theory there's data sovereignty that would be a good reason for Europe to do it and elsewhere, but, I think this whole industry has to do a better job of evangelizing and putting the local communities at peace as to why this is good for them.

And we have no choice because we're not putting this genie back in the bottle. We're going to build it.

This deal between Quebec and Newfoundland and Labrador, they’ll send down energy to the US which is going to be charged at a higher rate. You can discern between what you're charging locally. The concern in your neighborhood where maybe electricity rates go up because of what's being consumed in the data center, that goes away. In today's world, AI is going to help this as well. I can charge different things to different people at different times, and just like tailored health care, I get tailored electricity charges and consumption.

And the thing is, in a lot of those cases, there are already agreements as to why, there's sidecars, how the power comes in or whatever. It's not really an issue, but it's being made an issue politically. Nimbyism is a real thing. So yeah, it's going to be a battle. But I think the messaging has to be better from the whole industry.

I was just going to say coming back full circle, maybe tying some things together, what has changed, especially in the last quarter, was we had a couple of hyperscalers with good results. But Andy Jassy from Amazon basically recognizing that there are concerns around returns on invested capital and all that. He basically came out and said, I'm going to put this to rest. And he basically mapped out why AWS is seeing positive returns on invested capital. And he basically broke up all of the capital spending in terms of the shells, the buildings, the need for power. And then there's the compute that goes in it. There's the initial load of compute that goes in, and that includes memory and networking and all of that stuff. And then over time, as compute deteriorates, there's the ongoing maintenance and replacement of those. And he broke it up into 3 broad buckets and you could map it out. And he gave the investment community a lot of comfort in, okay, this is starting to work. And on top of that, you just saw that the top line growth numbers in terms of revenue from these hyperscalers are just off the charts. So it's just brought a little bit more comfort to the whole ROI.

Yeah. And I guess one of the arguments here is these are some of the best companies on the planet with some of the smartest people on the planet running them. So they're not going to go and dump $1 trillion in a hole. They've got a plan. And generally, when they've had plans in the past, it's worked out fairly well.

Generally. Mark Zuckerberg has spent a lot of money on things that haven't worked out. But he's one of the few guys who has a license to do that. Let's put it that way.

Yeah, exactly. So, anything else? Here's one I'll throw out maybe just to try and wind things up. I was watching videos this weekend of the Chinese Robot Olympics. I don't know if you were watching any of that.

I've seen some clips of it.

And the sprint robot that broke Usain Bolt's record. But they've been showing them folding laundry and doing basic household chores and things that. When you see a video like that, what does that get you thinking as an investment manager?

I keep pointing that out as something that's coming that's exciting. So it's always just refreshing to see that and scary in some ways too. But what really stood out recently: historically, you've had to code all the movements that a robot makes. It takes coders to code specifically each movement. And the promise of AI and large language models is that you could just show a robot a billion hours of an activity and you wouldn't need to code it at that point. It'll self-code just by watching and learning. I forget what the company is called, but we just saw a video of a robot being shown these activities, simple things from opening jars and flipping things open, and these robots had never been taught how to do this, but they're showing videos or they're showing actions and they're immediately doing it and they can do the full range. It doesn't have to be one specific thing. It could be a whole range of different ways of opening a jar and things like that. So I just chalk that up, this is probably going to hit us faster than we think. If you can start to train a robot or machine to do something at that speed, I think it'll just accelerate.

Yeah. And that's what I took away from watching those videos this weekend. I'd watched the same video last year, 2 years ago, and the amount that the robots have improved. Just in terms of what looks more life motion. It's almost they can mimic us and what we're doing. I mean, it's not perfect yet, but like you say, it's coming. And it's likely coming faster than you think. And that's China putting that on display. But I imagine in Europe and North America that we've got some pretty good stuff too that we could show off.

Yeah, for sure. One of the best companies in this is Boston Dynamics, which is an American company but now is owned by Hyundai. But you just go and look at the iteration of their robots and everyone's seen the crazy dog-looking robots and stuff like that. But their first humanoid robot, I think it was Atlas 1, and it was all hydraulic driven and the thing was massive and it's heavy metal, hydraulics, gears, all sorts of stuff. And then the latest version is Atlas 2, I believe. It's all electric. So you've lost all of that extra weight, all of the hydraulics that can burst on you and things that. Start leaking everywhere. Just there you can just see the advancements as you're going from hydraulics to electricity. With higher energy density in lithium-ion batteries, that helps as well. And coming back to the whole coding and training of these things.

And they're already doing stuff in healthcare.

Yeah, I don't know. I decided to step back and say it's the bipedal stuff. Intuitive Surgical has had a surgical robot for a long time, and they keep iterating on it and adding new surgeries, and it's becoming a standard of care in a lot of indications. So yeah, robotics in that sense are there. A little bit different fashion than what we saw at those games, which were incredible. But yeah, robotics has a place in that space as well.

So, I always end for either of you with: is there anything else that's just popped up and piqued your interest around any areas of technology healthcare that you think that might be particularly interesting? You were only on a couple of months ago. So, I guess that's a lot of pressure to put on you. You have mentioned the robots and agents and all those things before, but anything particular? Maybe even let's just spread it out, any one area you think to watch over the next 2 or 3 months, particularly with rates rising and other pressures on the sector?

Yeah, I think the debate around «open versus closed» weight models is going to continue to come to the forefront. It's going to be important to think about the implications there for the suppliers of the actual compute. So that's going to remain important. We were just talking about robots and physical AI. I think that could become bigger, if we look into '27. And the explosion demand for chips from physical AI is going to be huge. Within healthcare, one area that's really lagged for a while had been tools and diagnostics. Companies that supply bioreactors and reagents, that has started to percolate again. Reshoring is a big driver. As you get into '27, some of the reshoring initiatives are going to be important. I wouldn't be surprised to see that space have some legs and some acceleration into next year. But it's been a tough space for several years. So let's knock on wood on that one. So I'm looking at least.

Rob raises an interesting issue with the open weight models. I wanted to come back to the whole software space. That was another thing that helped the software companies because part of the negative on software was that their costs were going to go up as they needed to basically use all these frontier models for their agents. But now with lower costs in terms of some of the open models, their costs might not go up as much. But in terms of new things that we're looking at, I just keep going back to robotics. One thing that's interesting: what we saw with Anthropic, with all the coding models and how that accelerated— and we just saw the use of tokens, particularly in that domain, just skyrocket— and what made that special was, it's verifiable. The software either works or doesn't, and it can verify itself and then self-correct. I've been thinking about it for a while, but it was recently brought up by a guy named Ben Thompson. In the advertising space, we're at the point where all the major platforms, Google, Meta, they have basically self-serve tools that automate the entire process from creating a campaign to launching it to measuring it and seeing the results and everything. So this has set up itself as a verifiable process. You could actually see where the advertising business accelerates. Because they create a campaign, test it really quick, there's a click-through on the ad, or there's actually a conversion. So if that's a good ad, rerun it. If it's not a good ad, immediately get rid of it and move to the next ad. So we might actually see an acceleration in advertising. And the results out of Google and Meta have been really good. And by extension, Shopify, because Shopify serves so many small businesses and that rides off of Meta as well.

When we get to that level of customized healthcare, you're able to be diagnosed, medicated, worked on from a surgical perspective in a fashion where it's really customized down to the individual. That's the direction advertising obviously headed and AI only helps that. Well guys, that's a fantastic recap. Always great to have you on. I saw Marcello down at the elevator last week and said, we got to get Rob and Marcello on again. So thanks for accepting the invitation as always. And keep doing what you're doing. The results are just fabulous. Always great guests, always such a popular episode whenever we have you on. So thank you.

Thank you.

It's great to be here.