Tech tour insights 2024: What's next for AI investing

Jennifer Martin

Welcome to the T. Rowe Price technology tour insights discussion. I'm Jennifer Martin, a global equity portfolio specialist. In my role, I get to engage with clients around the world. And one of the top trending topics of discussion is what is going on in the technology sector and what is our outlook for the sector? Fortunately, for me, I work in an organization with multiple thought leaders and investors with proximity to that sector. Three of them are here with me today.

Please let me introduce to my left Dom Rizzo, the portfolio manager for our Global Technology Strategy; Tony Wang, the portfolio manager for our Science and Technology Strategy; and, finally, Jim Stillwagon, the portfolio manager for our Communications and Technology Strategy.

We have an exciting discussion planned today, but please submit any questions that you have through the Q and A function on your console. If we don't get to them today, we will follow up with you afterwards.

So, to set the stage today, as many of you know, at T. Rowe Price, we explore distinct perspectives around the globe. One of those includes our annual trip to Silicon Valley. We've been doing this exclusive trip for nearly two decades. During our recent December visit, we had over 40 portfolio managers and analysts travel to meet nearly 40 public and private companies where we engaged with top executives of the world's leading and emerging technology firms.

Today's discussion, we will address three broad topics. One year post the ChatGPT launch, our evaluation of artificial intelligence is that it is a technology innovation bigger than even our initial expectations. And as you'll probably hear from Dom Rizzo, I have the potential to be the biggest productivity enhancer since electricity. We are betting the monetization emerges from artificial intelligence. We see signs of this in the consumer internet and enterprise software. And, finally, the importance of identifying early for our clients these companies that sell mission-critical technology crucial to secular growth markets.

So, with that, let's get started. And, Dom, you volunteered to go first.

Dom Rizzo

Yes, I did.

Jennifer Martin

You and a large group of investment team just returned from our annual trip to Silicon Valley. What did you learn?

Dom Rizzo

Well, it's so funny to think, Jennifer, we were sitting here in this seat almost exactly a year ago. ChatGPT was launched on November 30, 2022. I actually took over our Global Technology Equity Strategy on December 1, 2022. So, if you're going to take over a strategy, you should usually do it the day after ChatGPT is launched.

But if I think about what's happened in the past year, it's really incredible. Right?

Last year, the AI gun got fired. This year is the year AI comes to fruition. And what we learned on this tech trip is that AI is probably bigger than our initial bullish views, like you said. I mean, while we were out there AMD actually increased the size of what they thought the AI chip accelerator market would do.

Historically, they used to say it would grow at roughly a 50% CAGR between 2023 and 2027. They took that up to a 70% CAGR, and to put some numbers around that, they've said that roughly the AI semiconductor market will be $45 billion in 2023. And that it's set to go to $400 billion by 2027. I mean, $400 billion, Tony. It, the entire global semiconductor market is $550 billion in 2023, right? This is really incredible numbers.

So the questions that we have to ask ourselves is, what is the ROI on these investments? Are we going to see really strong capabilities come out of these companies because of artificial intelligence?

And we'll talk about it more throughout the day, but I think the answer is, clearly, yes. And I would just say once again to our entire team, this is clearly the best use of my time every single year. We meet with 40 of the leading-edge submission-critical lynchpin technologies in Silicon Valley, the CEOs and the CFOs of the likes of Google, Meta, NVIDIA, Netflix. I mean you name it, we meet them, right?

And it's really just a uniquely T. Rowe event where we all get together, we're all able to exchange ideas and meet with some of the great thought leaders in the technology sphere.

Jennifer Martin

Well, it is an advantage at T. Rowe, the sector teams, how really that team leader connects analysts and the portfolio managers globally. It also really, after 40 meetings, gives you a good sense of the dynamics happening in that sector, and it becomes really the statement the sum the whole is greater than the sum of the parts.

And so maybe, Tony, if you could expand a little bit more on what Dom said about the addressable market of AI and some of the insights from your coverage of NVIDIA, the ultimate, lynchpin technology in in AI today.

Tony Wang

Yeah, absolutely, Jennifer, and just really exciting what's going on in the space right now. I think it's a real paradigm shift in how compute is happening. So, you think about traditional compute versus celebrated compute. This is a completely big, this is a complete shift in what's going on. That's why I think the market size has been consistently underestimated in many ways.

Just as simplistically how I think about it is that previously in order to run your data science program or your machine learning, you had to have essentially a lot of really talented data scientists and AI engineers, programming the model. And so now with the breakthrough of generative AI through ChatGPT, what you're seeing is that all you need to do is really feed the program more data and more and have more compute by more GPUs, it actually generates a lot more content, it just keeps going.

Dom Rizzo

The more you spend, the more you save.

Tony Wang

That's right. That's right, Jensen's favorite tagline right there. And so I think that's where the market is really underestimating the ROI here, the size of the market because you can't just compare it to the old Intel TAM or the CPU TAM, right? Because it's a totally different paradigm where these are new use cases in drug discovery.

And molecule simulation in ad targeting search engine optimization. So in terms of just the market size, I think

that we're, we continue to underestimate it every quarter because this is there is a big, pretty big paradigm shift.

I think the next forefront of innovation is going to be something called digital simulation, digital twins. So what that means is really instead of, right now, gen AI essentially giving you a recommendation on what will be probably the best, right?

The next wave of innovation that NVIDIA is really powering is something called digital simulation in that it will digitally replicate the real world. And then you can simulate a car factory, you can simulate a drug discovery process. You can simulate rebuilding like a new molecule or something.

And so that will save us billions of dollars from trial and error. And we're seeing a lot of green shoots there. And so once that happens, I think that the ROI, especially of GPUs, are really gonna prove out and like AI in general. So I think last year you look at two years ago, the tech trip was about, oh this AI craze, ChatGPT.

I think this year the tech craze that the tech trip was really about proving out the ROI and making it real and versus just a meme. And so that's what's really exciting about the space. I think that and, and for NVIDIA, they are at the tip of the spear and providing the compute and the acceleration platform.

Jennifer Martin

Well, I think both you and Dom have done a nice job highlighting the confidence of our direct investments and what I know both of you have talked about picks and shovels and really how active management is going to help navigate this trend very responsibly.

Dom Rizzo

That's the key with the AI bubble. It's, everyone is always thinking about, you almost said you said the meme stock phenomenon, right?

And what we're always trying to do is make sure that we navigate these difficult environments responsibly via our investment frameworks.

Jennifer Martin

And so, Jim, you've done a lot of really good research highlighting how the consumer internet mega-caps are leveraging artificial intelligence. And so maybe if you could share some of your findings, that would be very welcome.

Jim Stillwagon

Sure. Well, the companies funding this AI arms race, primarily the mega-caps spending billions of dollars on GPUs, networking equipment, new data center buildouts. Eventually, they need to earn a return on that investment. And that means we need to start seeing graceful handoffs between AI hype cycle enthusiasm and tangible AI monetization events. And, thankfully, those handoffs are happening in real time across our consumer internet majors, and we can run through a few examples.

First, AI is driving engagement across social media. Look at the rise of short-form video paired with AI-powered recommendation algorithms from TikTok to Instagram reels to YouTube shorts. If you think about a traditional social media feed, it actually sorts against a pretty narrow slice of content. It's primarily the posts of your friends and family. And then think about that step-changing complexity as you shift to high-velocity short-form video.

The algorithm needs to select from billions of pieces of content across the entire global creator economy. And what we've learned is that that's not only a massive machine learning and AI exercise to execute well. When it's done well, short-form video will meaningfully outperform a traditional social feed in isolation.

The average daily TikTok user is spending over an hour and a half per day inside the app at this point. Instagram has seen a 40% increase in engagement since they launched their TikTok clone reels. When it comes to social media, all else equal, more consumption equals more monetizable engagement.

The second example of how AI is driving monetization will be the introduction of chat bots in the messaging apps. So you think about WhatsApp, Instagram, Snapchat, all of those are natural distribution channels for AI-powered assistance. But what's exciting here is the potential to drive commercial activity with those AI assistants. Meta already has a $10 billion plus advertising business and click to message ads across WhatsApp and Messenger.

Now those are ads sitting inside Facebook and Instagram. But when you click on them, you open up into a 1-to-1 conversation between that business and the consumer. And if you have a question about a product or service, if you have a question about refund policies or apparel sizing, that chatbot can work you down the conversion funnel.

And what's interesting is we've already seen e-commerce activity taking off in messaging apps in emerging markets. So WhatsApp in India, 60% of users at this point are chatting with a business account on a weekly basis. And so we know that if AI-powered assistants can drive those incremental conversions, marketers will spend more on click-to-message ads.

Then a final example of how we see this link between AI and monetization playing out just be the ad tech stack, right, the engine behind the vast majority of consumer internet. The old targeting approach of individual user tracking based on device or user IDs, that deter deterministic model, it's gradually on its way out in part due to evolving privacy norms. And in its place, we're going to see the most successful platforms used cohort-based probabilistic modeling.

They're going to build privacy-safe campaigns with AI. Google has what they call performance max, Meta has Advantage

Plus. They're all building AI functionality deeper and deeper into the ad tech stack. And this goes beyond ad targeting and attribution. Think about what gen AI could do for ad creative, in terms of those basic audiovisual elements for a campaign.

But the end game here is just to abstract away complexity for small businesses, make it that much easier for a small business to show up with some basic creative elements, the return thresholds, and then hand the keys over to Google or to Meta. If you can lower the bar to experiment in performance marketing, you will increase ad auction density. It's actually one of the main lessons that we learned from Meta's success over the past decade. They had less than a million advertisers back in 2011, 2012. They have over 12 million active advertisers today.

Small social media platforms like a Snapchat, like a Pinterest, less than a million active advertisers. Your largest broadcast networks, a couple 100. So we think AI has the potential to be that next major unlock for ad auction density and overall monetization.

Jennifer Martin

Oh, that was excellent. I mean, my mind went so many different places.

One, thinking about everyone's kids who are probably on TikTok for an hour and a half every day, but also just the engagement and then really the use cases that you are seeing on the internet because I think you've said, very clearly, NVIDIA's revenue is someone else's capex, which needs to be someone's revenue. And I think you've highlighted some of that at least in large-cap internet.

So I think a follow-up to that, Jim, is, you know, we're excited about the AI opportunities in large-cap internet, but how are you thinking about the potential risks?

Jim Stillwagon

Sure. I mean, any time you have an investment cycle of this magnitude, there are risks in terms of margin compression and in terms of capital misallocation. I would argue those risks are even more pronounced today given that we just completed tech's year of efficiency in 2023. Google, Meta, Amazon, all those companies let go between 10 and 25% of their workforce last year. And it's been almost all upside to date, better operating efficiency, leaner and faster engineering teams, the remaining employees, pleased with the share price going up.

As investors, we want those cost savings to stick. Now, I don't think we're going to go back to the pre-pandemic or pandemic-era peak of over-hiring when you had that period where Apple, Google, Microsoft, and Meta were collectively onboarding, I want to say 30,000 net new employees a quarter, but AI and machine learning engineers are in high demand right now, and they are absolute top of the food chain pay scale-wise in the valley. And beyond headcount, you have this debate over rising capital intensity across the tech sector.

Yes, it's great to see NVIDIA's data center revenue triple this past year, that they generated more revenue out of that segment in a single year than in the preceding decade. But to your point, NVIDIA's data center revenue is somebody else's capex. And so in the early 2010s, Google and Meta would spend less than $5 billion a year collectively on capex. That number is over $60 billion today. These are no longer asset-light advertising businesses, and that's before we see the full impact of an AI investment supercycle.

Now, the good news is that the mega-caps do have offsets to these margin pressures. You still have sizable cost savings opportunities to address. Meta spending over $15 billion a year on VRAR metaverse bets. Amazon spending $78 billion a year on Alexa Echo smart speakers. To date, we would estimate those two projects alone cumulative losses in excess of $100 billion. Neither of those companies have product market fit to justify that level of investment.

I think there's also the potential for AI-driven productivity gains. Remember, the single largest cost bucket for the mega-cap consumer internet names, It's tech talent, and so it's early, but we're seeing evidence of engineers with AI-powered copilots delivering 30% plus type of efficiency gains on code documentation, code generation relative to unaided task duration.

And then, finally, these businesses are just incredibly cash flow generative. You look at Google, they're buying back 55, 60 billion a stock a year. That's not even making a dent in their $110 billion net cash balance sheet. So there's flexibility to protect EPS growth algorithms just by adopting more shareholder-friendly capital allocation policies.

Jennifer Martin

Really helpful. I love the word flexibility because I feel like that might be a good lead in, Dom is, you already highlighted in this discussion. AI requires a lot of compute, and what applications are you evaluating that will justify that incredible amount of spend right now? Where where's the flexibility in other parts of the market?

Dom Rizzo

Well, Jim did a great job highlighting the potential consumer applications. You know, I'm so excited for the day that I can just whatsapp with United when they're late, when, when I'm late on my flight next and I could say, hey, please put me on the next flight. I mean, that day is coming, right? And that's gonna be really exciting.

But in the short term, I think the real clear ROIs are actually in the enterprise, right? The enterprise software companies are the conduits in which AI will flow into your day-to-day life. And similar to how smartphone usage actually, you know, blackberry usage started in the in the enterprise and went to the consumer space over time. I think actually that's what's gonna happen with artificial intelligence as well.

I mean, think about Microsoft copilots. We've talked about them on these webinars, webinars before. But Jim highlighted the 20 to 30% productivity gain that you get from developers who develop on using github copilot technology. That's one clear use case today that will justify the very high ROI of these GPUs.

Another very clear use case today is embedding AI into the enterprise applications that you use today on a day-to-day basis. Think salesforce.com, think Service Now, think Workday. You can put so much different artificial intelligence wrappers in those different applications, and for salesforce is the easiest to understand, right?

If you're a salesforce user today on the line, it's very, very possible that you have to go in and you have to write out this is who I met, this is their contact information, this is where I think they are from a legion perspective. That's all gonna be done by the artificial intelligence at some point.

If you think about PowerPoint creation or even this, think about how much time we spent writing and thinking about what we were going to say today. That's all going to be AI enabled and AI augmented.

Finally, the last point that I think is really underestimated when it comes to AI and the enterprise is enterprise building their own applications. So for a long long time, all the enterprises in the world would build their own applications because they viewed it as a source of technical talent differentiation, right?

Think about all the different internal applications that T. Rowe developed over time or that large banks have on their own balance sheets, right? Look at that very, very large R and D number that JP Morgan spends annually on their software spend, right? They've [unintelligible] that was like the eighties and the nineties and the early two thousands, right?  Then from the mid -2000s, probably until the 2020s, we've been on this steady pace of outsourcing to these enterprise application software companies and letting them do all the hard work of the code development.

Now with AI, your coders are significantly more efficient. So what does that mean? It means that the enterprises are once again going to start coding their own internal use cases, spinning up all sorts of different internal applications driven by AI. Some of that will be built on top of the platforms like a Service Now, but a lot of that will be built on top of the infrastructure software companies. And that's like some Mongo DB who should see significant database usage or someone like a data bricks, which T. Rowe actually led the private round in September on. And I think that they're building the AI platform of the next generation of software.

Jennifer Martin

And I also know that what kind of came out in a lot of the earnings calls was just kind of the basics. A lot of enterprises have to get their data organized for any of this to start. And so you're seeing that also in some of the enterprise applications.

Dom Rizzo

I think a great example of that is SAP, right? We haven't had a meaningful ERP upgrade cycle in two decades. Cause they're painful. Cause they're painful. And if you're a CIO, you may get fired if you suggest it. You got to do this impossibly difficult thing, and then at the end, it's probably not going to be perfect. And then when it's not perfect, you probably have to move on to the next company.

But what's happening now, you got to get your data in order. In order to get your data in order, you have to put it in a data lake house from the likes of data bricks. You have to do an ERP upgrade cycle probably, which should benefit the likes of SAP. And with that, you once you have clean data and usable data, that's when you can really start building those, those new applications that are AI specific for your organization.

Jennifer Martin

Really helpful. Well, I know, I mean, I think Dom just made the case that data is a strategic asset.

And, Tony, I think maybe what you could expand on, you've done a lot of really good work on AI's impact on privacy and also maybe a little bit in the cybersecurity world and so maybe you could expand a little bit there and kind of complement Dom on some of his excitement on enterprise.

Tony Wang

Yeah. Well, I think there's AI is going to be very disruptive in cybersecurity as well. I think that what you're going to see is that because gen AI can really produce like really tailored messages that can essentially send you a really good phishing email. They can scan your LinkedIn, scan your Instagram, and that just creates like the vulnerability surface to be a lot bigger than it used to be. And so, and then the cost of producing spamware and malware is going to go down a lot. So as a result, this is going to really make the solutions from crowd strike, Scalar, the endpoint players to be really, I think there's going to be greater demand for that.

So I think that it's going to be essentially like it increased the attack surface as a result of that. On, in terms of, I think another aspect when I think about privacy, you know, another company we should probably talk about is Apple.

You know, they are actually, I think, kind of a sleeper in terms of an AI beneficiary here. I don't think companies are, I don't think a lot of people are thinking about Apple as particularly AI forefront in terms of what they're doing. But if you think about their advantage and their competitive ability to control essentially what's in their platform ecosystem, I think there could be a lot of aspects of AI that could be beneficial to Apple. For one like, you know, they control the App store.

So if we have gen AI apps that are gonna be coming to fruition, like they're gonna benefit from their App store. And then a lot of times you're gonna have that gen AI be run on the phone. And so at the device level, so that could create an upgrade cycle. But I think that what Apple will probably do a little bit, be a little bit more careful on is that they are very privacy first, right?

And so they probably don't want to be creating unsettling the kind of conditions for their, their customers to feel weirded out, I suppose, that the Apple AI knows more about them. So I think they're going to tread pretty incrementally, but I think there's a big opportunity for them to essentially have their version of ChatGPT for their phones and deliver probably an upgrade cycle because of the functionality will be so much better, kin of, Dom, your early example of like kind of if I need to book a flight or something, like can I do that easier through what's going on with gen AI?

So, I think there's going to be this real balancing act of privacy and then also convenience and AI first. So, but I think Apple is well positioned to eventually benefit from it.

Dom Rizzo

And Tony and I talk about this all the time with Apple.

What's so interesting about Apple is that it's really a secret silicon company, right? It is got the second-best probably, if not, they, they would get mad if we said second best, so we can say one of the best, one of the best, one of the best semiconductor engineering teams in the valley up there with the likes of NVIDIA or an AMD or anybody in digital semiconductors.

And with that, like you said AI happening at the edge, AI happening at the device, that is a real competitive advantage that they have, which is the silicon in your pocket.

Tony Wang

Yeah. Absolutely.

And then, you know, it's, it is a closed ecosystem, right? So they've got the full stack that they can optimize for from the silica into the software across their whole platform. So I think that's, that is really exciting. And then, you know, they can essentially like make sure everything's working really well versus like a relatively open ecosystem as a result.

Jennifer Martin

You just highlighted again why many of those companies are the ultimate mission-critical linchpin technologies. We can't live without some of them.

So I think maybe pivoting a little bit because, you know, Jim, another area of opportunity that you identified from our technology trip was really this group of unprofitable technology companies that really has turned into cash-generative businesses, and they've kind of turned from proof of concept to real companies. And so what, maybe let's hit on some of those because we did talk to other companies outside of AI.

Jim Stillwagon

You're right, we're finding opportunities with companies that were labeled incorrectly as cautionary examples of Zerp-era venture capital excess that have subsequently proven out attractive underlying unit economics for their businesses. And over the past decade, you had this challenge, right, of discerning business quality, given all the distortions at play with a historically low cost of capital environment.

You know, every start-up was incentivized to pursue the most aggressive cash-burning land grab possible. But then over the past two years, the public market slammed on the brakes with these growth-at-any-cost playbooks, and that has filtered down into private market funding behavior as well.

So now we get to see a handful of start-ups from that 2010's vintage really prove out their original pitch, substantial growth plus solid profitability plus share gains from weaker hands. And Uber is, is a great example there.

The global rideshare industry has rationalized into a favorably imbalanced duopoly. Uber has 7030 dominant market share splits in most regions globally. And that core Uber X franchise has turned into a high-growth transportation utility for millions of consumers worldwide. And Uber keeps extending the growth runway for mobility with new products.

Uber reserve, Uber comfort, taxi integrations, two wheelers, and then on the cost side, you've had subscale rideshare competitors retrench as the driver supply dynamics have improved. And that's allowed Uber to dial back some of those really costly incentive payouts that distorted mobility economics for years.

So after the losses piled up through the 2010s, they got worse at the start of the pandemic. Uber has now turned the corner. Uber generated over $3 billion of free cash flow last year. They strung together enough gap profitable quarters to prompt S and P inclusion. We're now debating their capital allocation strategy for 2024. How much stock they can buy back.

Dom Rizzo

They just hired the CFO of Analog Devices, who was a buyback machine while he was there.

Jim Stillwagon

It just goes to show this is a, a comprehensive dismantling of the whole rideshare will never make money trope.

We see the potential for similar business quality reassessments with names like Door Dash, like Shopify, like Spotify. You know, specific to Spotify, they're exiting a deep investment cycle in podcasting now that they've taken category leadership away from Apple, and they're, they're still going to have exclusive talent signings.

But the model going forward will look more like YouTube than like HBO. And when you combine reduced podcast investment with, you know, a reduction in their headcount, they shrank their workforce by over 20%, the path to double-digit margins for the streaming audio leader looks clearer than ever. And this is for a company that has struggled to produce profitability for most of its history in the public market since its direct listed back in 2018.

So we keep looking out for these businesses that were dismissed as profitless byproducts of, you know, zero interest rate largesse. And we're focusing on the ones that have taken real action to not only survive but thrive in a higher-cost-of-capital environment.

Jennifer Martin

Well, you and the team have done a great job of re-underwriting a lot of those businesses, and I feel like you probably just hit on most of the apps that are on everyone's phones right now. So I'm glad that they're making money because I want to keep using them.

So I think pivoting a little bit, Dom, another area that you've been evaluating related to artificial intelligence and helping companies we get organized in terms of the digital revolution brought on about AI is the IT services industry.

So maybe start there and you can expand in the directions you go.

Dom Rizzo

The IT services industry is so interesting because it's really the ultimate use case for whether or not AI is a replacement technology or an augmentation technology. And what do I mean by that?

You know, if AI makes your developers 20% to 30% productive, do you need to rely on the IT services companies as much as you have historically or can that just be done in-house? On the other hand, AI is really, really hard to do. It requires utilizing these GPU parallel processing technologies, which are frankly just different than traditional serial CPU technologies, right?

And so you have specialists at these IT services companies who are gonna be able to help hold your hand if you're a large enterprise through your AI journey.

And finally, we talked about getting your data in line, getting your infrastructure in line. What does that mean for these large organizations? You know, historically, that meant going to an Accenture and asking them to help you with your SAP estate, asking them to help you with your Oracle estate, get it in line, get it clean.

And so when I think about these IT services companies going forward, you know, we just went through a little bit of a cyclical downturn in the IT services markets. Most of these companies probably over-hired. If we look back historically and now they have large benches. And so the question is, does the IT service demand comes back as  result of AI and getting your data house in order? My gut is yes, but it still remains to be seen, and we'll see what happens with the likes of an Accenture, an Epam, a globot, an Indaba.

But my gut is, going forward, that what we're going to see is that these companies are really going to be more helping partners in the AI journey rather than being replacement. And, actually, if you go back and you look at the cloud transition, people thought that you wouldn't need Accenture after you went to the cloud as well. So we'll still, it still remains to be seen still an open question. But it's a really exciting place to continue to study.

Jennifer Martin

Really helpful, and I think it's a good transition because, Tony, as AI expands into the industrial economy, notably autos, why don't you share how Tesla is benefiting from this technology? It's another company we visited on the trip.

Ton Wang

Yeah, absolutely. I think that what's interesting about the Thomas driving space is there's two ways to do it that the companies are pursuing.

One is the kind of HD mapping where you drive the streets, you mark everything manually. And that's kind of where I think a lot of the Waymo cruises of the world have been directing their efforts.

And then Tesla has a very AI-first strategy, which is instead of having a of car where you've seen kind of the cars drive through the streets with a bunch of lidars and spinning disks and stuff. Tesla has like a fleet of cars that are very lightweight in their hardware. So it's all camera driven and they are collecting data driving around and then using that data to train a model.

So I think what's happening with, with AI is that that AI-first strategy could prove to be possible to make up for that worse signal that, kind of the big kind of lidar and radar and all those big hardware brute force mapping strategy is like providing. And so I think that we're seeing that this could really result in step function in Tesla's strategy in terms of improvement.

So I think that and you're also seeing other companies now pivot more for towards AI first, training their models to be kind of like a thing about like truly self-driving autonomous car where you can plop the car down somewhere it's never driven and it can still drive it because it's learned based on models versus having to have a car only drive in roads that have been mapped out in a high-definition way.

So, I think that's what's going on, and it's really exciting for the space. But, you know, a lot with a lot of these things, it's like two steps forward, one step back. And so, you know, we, but that is, I think the, the really exciting takeaway that we've seen in, in with AI.

Jennifer Martin

And it certainly is a better-looking car when it's just camera only, to your point.

Tony Wang

Absolutely. And economically, it's a lot more feasible to have a few $1,000 worth of sensors versus $50,000 of sensors to actually make kind of autonomous driving ROI positive and improve out a business model.

Jennifer Martin

Very good.

Well, I think we'll move towards the idea that every major compute paradigm generates different winners in the market.

And so maybe we should start with you, Dom, on what are the next winners?

Dom Rizzo

Yeah, I mean, I think it comes down to this fundamental question is AI a sustaining innovation or is it a disruptive innovation?

I mean, I guess you could argue NVIDIA is a new winner relative to intel, but NVIDIA has been a winner for a long time. But in general, I think AI is a sustaining innovation. And what does that actually mean for this computing paragraph paradigm relative to historical computer computing paradigms? It actually means that the large companies will probably get larger, that the winners will probably keep winning.

And why is that? It's really because you need four things to be successful in AI. Number one, we talked a lot about it today. You need vast compute resources. This is extraordinarily expensive technology that you're implementing. The good thing about this AI investment cycle versus historical cycles like the internet is that that vast compute resources are being funded by the cash flows of a lot of Jim's companies, which are some of the most profitable companies in human history, right?

And so this is not a debt-driven cycle like the internet cycle was. This is a cash flow-driven cycle by some of the really the most profitable amazing companies that we've ever seen.

The second thing is you need talent, and who has the money to spend for all of the great talent? What does a great AI engineer go for these days? I mean, we're seeing stories of the seven-figure budget just being given out like candy. It's a talent war between deep mind and open AI right now. It's, it's starting at seven figures for the best engineers. Easily, easily. So you need talent, right?

And who can afford consistent seven-figure salaries, bonus spot bonuses for a lot of these guys, it's, it's a large company. And the two things that probably get underestimated that you need for AI. One we talked about data as a strategic, as strategic asset, data is very, very important, and who has all the data, the large companies, right?

And then, finally, distribution compute resources spending super high. We need to amortize that cost across a massive user base, and who has all the users, the large tech companies. And so if you look at what it takes to be successful in AI and who has those capabilities to compute resources, the talent, the data, and the distribution, it actually leads you to the conclusion that AI is basically a sustaining innovation and the large companies will continue to get larger.

Jennifer Martin

Very good.

Tony, do you want to add any thoughts on that? That was pretty comprehensive.

Tony Wang

Yeah, I think, I think I'm on a similar page. I think the big do get bigger here.

When you think about kind of like what is really special about gen AI is that if you have an application or a customer base, they're giving massive benefits to that install base. So you think about one example would be Microsoft, you know, like Copilot essentially would be a gam changer and it's not that hard relative to other things that gen AI can do.

And so if it can help you summarize your email faster or craft something Like, I think you would pay $30 more a month and that would be, you know, massive in terms of what the impact on Microsoft's PNL is. And then when I think about just kind of the trust that Microsoft, for example, has in the enterprise, if you're going to be installing kind of a gen AI tool, you probably are gonna go with something you trust.

So like a Microsoft or Service Now is a kind of having that incumbency and trust and relationship with that customer is really important. And I actually think that a lot of the stuff in terms of large language models, if they're pretty good, I think that they will be good enough. And then from there, it'll be more about like your apps and customer base to leverage your cost of AI.

So I actually view that mostly it is going to be something that's going to be pretty democratically rolled across the digitally

enabled companies. And then I think the rest of the, and I think there will be disruptors that will come out, but largely that incumbency is going to be really important, I think for gen AI to benefit.

Jennifer Martin

So maybe, Jim, I could end with you on this question.

If AI ends up being more of a sustaining rather than disruptive innovation, what does that mean for the Magnificent Seven concentration going forward?

Jim Stillwagon

I think everyone here has framed it right. Like AI is going to favor the largest platforms with the most compute resources, the deepest data pools, the strongest engineering teams. All of that benefits the Mag Seven. That said, we do see potential constraints on concentration.

First, the mega-caps have maturing growth runways. Let's take digital advertising as an example. Digital advertising was just over 15% of measured media spend coming out of the GFC, got up to about 50% pre-COVID. Now we're at 70%. So we, we still think that digital advertising will ramp alongside e-commerce growth. Online retailers will continue to pay their digital rent to search and social platforms. But there just isn't as much budget left at this point to take from analog channels. So digital advertising or a 4 to 5 times GDP grower pre-pandemic it's looking more like 2 to 3 times GDP growth prospectively.

Second, the mega-caps aren't playing nice in the sandbox anymore. They are getting into each other's core business lines, and that's partly a function of those maturing growth runways. Think back to the mid-2010s. Google and Meta were taking upwards of 80% share of every incremental digital ad dollar in the U.S. Fast-forward to today and Amazon has effectively gate crashed the duopoly. I mean, everyone here is a Prime subscriber. You go on Amazon, you search for a product, you see those Amazon-sponsored product ads. That is now part of a $45 billion advertising business. It's throwing off more absolute profit dollars than the AWS Cloud unit for Amazon at this point.

Dom Rizzo

It's bigger than YouTube, right?

Jim Stillwagon

It is. And growing faster.

Dom Rizzo

It's bigger than YouTube and growing faster.

Jim Stillwagon

And, finally, we have to talk about regulatory concerns both in the U.S. and the EU. Some of the regulatory headlines are a distraction. You know, when the FTC is going after Amazon's private-label business, when it's less than 2% of their global sales, I mean, if we really want to crack down on retailers favoring in-house brands, we could probably start with Kirkland's signature before we go after Amazon basics.

But some of the recent regulatory action carries real impact. The DOJ's case against Google search for their default deal, distribution deal with Apple, that's a $20 billion per year payment from Google to Apple at risk It has consequences for Google's competitive positioning on IOS devices going forward. It has consequences for Apple's services revenue outlook.

So we aim to balance our AI optimism against these potential constraints, and, and in my view, that's where the power of the T. Rowe platform comes into play, right? The Mag Seven have to earn their portfolio weightings every day because they're getting stacked up on a risk/reward basis against all the idea flow from the broader T. Rowe platform up and down the cap range.

Jennifer Martin

That's a great segue into, really, as sector leaders at our organization, what are you thinking about technology's leadership in 2024 as is what you are being asked to communicate to our, our team?

Dom Rizzo

I'll start. Look, there's four things that I look for on my strategy, right?

The first is what I call linchpin technologies. We've been talking about that a lot today. These are the technologies that are mission critical to the success of their customers or make their users lives dramatically the better better.

The second is that they should be innovating in secular growth markets. That means growing faster than the fast end markets that they serve.

The third is that they should have improving fundamentals that comes in the form of revenue that's accelerating, operating margins that are expanding, or free cash flow conversion that's improving.

And the final, and arguably the most important, thing right now is making sure that they have reasonable valuations, and you have to use all four of those things to navigate the AI environment responsibly.

If I look at those first three parts of my framework, broadly speaking, technology looks pretty good. So then the question really comes down to where are evaluations today? And are we past the point of reasonableness? Right? And I, I personally think we're probably still in the range of things being OK from a valuation perspective. I just pulled some numbers.

You know, if we look at the S&P 500 on 2024 earnings is trading at a low twenties earnings multiple. Now, as everyone knows, you really got to pull those Mag Seven out of the S and P 500 and do the Mag Seven versus the X Mag Seven for the S and P 500, and the Mag Seven are trading at a high twenties, maybe a low thirties earnings multiple depending on the day versus the rest of the S&P 500. The S and P 493 trading at a high-teens multiple, right?

Now, if we look at that Mag Seven multiple relative to history, it's actually not that bad. We've seen periods where the Mag Seven has traded 30, 40 even higher time, even 50 times earnings as a group. So, you know, we're in the OK range in the high twenties from a Mag Seven perspective. If I take a step back and I think about Mag Seven performance, these companies really struggled from peak to trough, 2021 to 2022. Right?

As a group, the group was down 40% from its peak and it was trading at 50 plus times earnings at the time. Since then, we're up over north of 70% and we're at much more reasonable high-twenties earnings multiple. So if you think about one way to think about the Mag Seven, is that the PEs have come down roughly 35% and the EPS has increased roughly 35%. These are from peak to trough numbers. And then that's how we kind of get to the kind of hitting all-time highs now period.

Then, finally, the global technology index. That historically peaks around 27 to 28 times earnings. That's when you really got to get a little bit more careful, Right now, we're at 25 times earnings. So at the upper end of the range, certainly, but not quite at that historical peak. So when I look at valuations broadly, for tech, again, they're OK, they're in the range of reasonableness but clearly at the higher end of that range.

Jennifer Martin

Really helpful.

So, Tony, do you want to add to reasonable?

Tony Wang

I'm pretty constructive on tech. I think in two different parts.

One is the AI part, which I think is durable that is a kind of paradigm shift. And so the picks and shovel sellers to the AI revolution as well as the companies that are really benefiting from AI, I think that's like creating a, you know, second kind of act for those companies this year that also did well last year.

And then I think about the rest of tech, which is not as AI exposed, essentially came off of two years of, of pretty slow IT spending. And so our economy is, is pretty much, I think we're looking at where economic activity indicators like ISMs are at, they're pretty low. So I think we're getting a celebrating environment in the next few years.

The rest [unintelligible] will do well. And so perhaps the leadership will change a little bit over the next few years because of the handoff of these kind of AI versus cyclically depressed, maybe IT tech stocks. But I think that there's a good opportunity in both those swim lanes. And so I'm still very excited about 2024 for tech.

Jennifer Martin

Perfect.

All right, Jim, bring us home.

Jim Stillwagon

Sure.

When you have a thematic breakout, like what happened with A I in 2023, you want to be thoughtful about the likelihood of an encore performance. You know, on the macro front, there's a lot of enthusiasm right now around the prospects for a soft landing, still navigating a lot of geopolitical risk with Russia, Ukraine, China, Taiwan. You know, but what I always emphasize particularly with the tech sector is that breakthrough innovations do not wait for benign macro conditions.

Bloomberg had this great article back in 2022. Late 2022. It was titled "Surveyed economists project 100% odds of a recession within one year." One month after that article was written, open AI launched ChatGPT. One year later, we're all sitting at a table talking about the AI arms race.

So I just go back to what we talked about earlier today. Like do we see tangible evidence of customers paying up for AI innovation? Are Google and Meta advertisers spending more for AI-enabled campaigns? Are enterprises increasing their cybersecurity budgets to fend off AI threats? Is Microsoft gaining traction with AI-powered copilots?

Those type of debates will shape the tech industry narrative in 2024, from semis to software to consumer internet. And if we can answer those questions affirmatively with conviction, that's how you get an encore for tech leadership in 2024.

Jennifer Martin

Thank you Dom, Tony, and Jim.

All of you have helped contribute to the clear case for active management. It's clear from our discussion today that there are many durable trends to invest in technology. Secular trends like AI, but we also hit on a lot of other areas of technology.

We also want our audience to consider the amount of expertise that is required to navigate the rapid changes in both the technology sector and the market environment and to consider T. Rowe Price active management approach to investing.

We also want to thank our audience, and we would like to welcome you to our next event, which you can register in your console. Thank you again for your time today.