Why the AI Boom Is Just Getting Started | Edited Transcript
A professionally copyedited transcript of Patrick O’Shaughnessy’s Invest Like the Best conversation with Alex Sacerdote on AI S-curves, Anthropic, infrastructure scarcity, software disruption, hardware, and Whale Rock’s research process.
Chapter Timestamps
00:00 Intro: Anthropic, AI infrastructure, and why Alex Sacerdote thinks this S-curve is still early
09:55 AI’s L-curve: enterprise adoption, compute scarcity, and why demand is already outrunning capacity
19:31 Whale Rock’s S-curve playbook: inflection points, moats, and underappreciated earnings power
26:14 Spotting inflection points: when intuition, anecdotes, and customer pull matter more than clean data
32:02 Finding AI winners: backing private companies, public-market analogies, and where Anthropic fits
40:04 AI vs. software: why agentic coding pressures SaaS margins, pricing, and seat-based models
48:13 Hardware renaissance: chips, memory, networking, power, cooling, and the capital cycle behind AI
58:04 Why investors miss AI: valuation anchoring, exponential earnings, and skepticism during platform shifts
1:05:18 Whale Rock’s research machine: data sources, expert networks, company calls, and portfolio sizing
Made with: The Transcript Desk Chrome Extension
Full video:
Alex Sacerdote is the Founder and Portfolio Manager of Whale Rock Capital Management, where he has spent the past two decades investing through major technology platform shifts. In this episode, Alex walks through Whale Rock’s framework for finding the most important companies in technology: S-curves, competitive advantage, and underappreciated earnings power. He explains why AI may be the biggest S-curve yet, how Whale Rock built conviction in Anthropic, why code has become the first major unlock for AI, what AI means for the software market, and why the hardware industry powering AI is entering a new renaissance.
Transcript
00:01-02:14
Alex Sacerdote: When you hit the right part of the S-curve, you experience exponential unit growth. If you have a strong business model, your earnings don’t just grow linearly—they grow exponentially. The world generally doesn’t think in exponential terms. Very few people believe you can accurately predict outcomes three or four years out. However, if you follow and understand the S-curve, recognize the competitive moats, and know how to model them, you really can predict these major shifts. The enterprise AI application market is currently less than 1% penetrated. We often talk about S-curves, but we call this an “L-curve” because the growth is just straight up.
Patrick O’Shaughnessy: Alex, you mentioned that Anthropic is currently your highest conviction position. Can you tell the story of how you discovered it and made the investment? I’d like to use that anecdote as a way to dive into the topics we’re both interested in: public investors moving into private markets, the business of Anthropic itself, and the broader AI landscape. It’s a great way to zoom in. Why is it your highest conviction, and how did you get started?
Alex Sacerdote: Well, when the starting gun went off with OpenAI’s ChatGPT in November 2022, we immediately pivoted the firm to do a massive deep dive with our ten-person team. Whenever a new computing paradigm emerges, a new tech stack is created, which produces new winners and losers relative to the old stack. In this particular stack—as Jensen Huang often discusses—you have power and chips at the bottom, followed by the cloud providers, then the foundational models, and finally the applications on top. In early 2023, we decided we wanted to be in chips and infrastructure first. Not only do those sectors see the demand first, but we already knew who the winners were. No matter who wins in the layers above...
Alex Sacerdote: Regardless of who wins at the top—which we weren’t certain of at the time—we knew we were going to need massive amounts of compute. We did a deep dive into that, which we can discuss later.
02:14-04:11
Alex Sacerdote: Over the next two or three years, we began to get more clarity on how the foundational model layer would evolve. At that time, there were about 60 different companies pursuing it. OpenAI was clearly in the lead.
Alex Sacerdote: We hosted a webinar in April 2023 where we laid out the possibilities: it could be a winner-take-all market, it could become a total commodity due to open-source players, it could be a race to zero, or it could become an oligopoly with three or four leading players.
Alex Sacerdote: What we observed over the following years was that almost all the startups fell away and died. Even some of the largest companies in the world struggled. Amazon never really showed up. Meta came in strong initially, but their efforts faltered, and they eventually had to do a total reboot.
Alex Sacerdote: In the meantime, Anthropic emerged as a dark horse candidate. As a startup, they focused almost purely on the enterprise market, while OpenAI had essentially won the consumer side.
Alex Sacerdote: Then there is Gemini. You can never count out Google; we love the company, and it remains one of our largest positions. It really started to look like a three-horse race—an oligopoly.
Alex Sacerdote: This is very similar to how the cloud market evolved, where three companies underpin the entire SaaS world and maintain excellent businesses.
Alex Sacerdote: We were also aware of the risks from open-source and Chinese competitors. However, we began to feel comfortable that the quality of the tokens from the leading edge was superior. If you are only 80%...
04:15-06:11
Alex Sacerdote: The proprietary models are superior because when you’re already at 80% of the benchmark, moving from 80% to 85% is a massive unlock. Open-source developers don’t have access to the same level of compute, so while they can get close to the leading edge, they can’t quite leapfrog it; they eventually falter. Meanwhile, scaling laws and other methods of model improvement—like feedback loops—showed us there was a very strong runway ahead. Everyone we spoke to who was close to the industry believed these scaling laws would continue to hold.
This led us to develop a thesis that this would be a three-horse race. But the real game-changer was code. Coding is the true unlock for AI. In the first few years, we knew AI would be significant, but we remained skeptical. We made large investments because we knew the demand for training would be there, but we weren’t certain about the revenue potential or if it could truly replace labor. If you remember, early versions of these models were decent, but there was a lot of negative feedback from corporations regarding whether they could truly be agentic.
We realized that in 2023 and 2024, coding tools really began to explode. The first generation, like Microsoft Copilot, cost about $20 a month. It could improve your coding “grammar,” find a bug, or perhaps generate a block of code—the equivalent of a paragraph. Then, Anthropic released a model in the middle of the year that could do so much more. It reached a point where it could function agentically, and we saw the coding market just explode. We started hearing from people who were using it without restrictions—even within Anthropic itself—that the productivity gains were massive.
06:14-08:18
Alex Sacerdote: At that time, even within Anthropic, people were spending $100 a day on tokens. If you do the math, that comes out to $20,000 or $30,000 a year. Considering there are 20 million coders in the world, you have a half-trillion-dollar market just from coding alone. Mind you, that was based on technology that was already seven to nine months old. We could see from the coding market alone that Anthropic had a tremendous opportunity ahead of it.
It’s actually pretty funny—we wrote in our investor letter that we made the investment at a $1.8 billion valuation. At the time, I think they were hoping to reach a $9 billion valuation.
Patrick O’Shaughnessy: So, from one to nine?
Alex Sacerdote: Yeah. And then the numbers were like nothing we’d ever seen before—going from $100 million to $1 billion on the way to $9 billion. But when we did that in August of 2023, nobody had any idea what 2024 could look like.
The second big unlock lately is that Claude Code has become almost completely agentic. You have people like Andrej Karpathy and Linus Torvalds—two of the smartest people in coding—who have completely flip-flopped. Karpathy said that with last year’s tools, the AI could write 20% of the code while 80% was handwritten. That flipped when the latest model came out. Now, he hasn’t written a line of code except in English. Not to mention the pure unlock we’re going to see for people who never knew how to code in the first place.
So, coding alone has completely taken off, and Anthropic has been able to stay ahead in that space. One major difference between the cloud providers—like GCP and AWS—and these AI companies is that cloud services are generally a commodity. They are selling you servers and storage.
08:19-10:18
Alex Sacerdote: While storage and traditional software have a layer of stickiness, many people initially thought AI models would become pure commodities. However, we’re seeing tremendous differentiation within them. There are different training methods and specialized skills that each model excels at. While some users use routers to switch between models—which makes them seem like commodities—Anthropic is actually very strong in private equity and finance, while Google is excellent at ingesting PDFs. There is a significant amount of critical IP involved, which creates a great competitive advantage.
Alex Sacerdote: Many companies have tried to challenge the coding franchise, but Anthropic has managed to stay ahead. Another advantage for foundational models like Anthropic is that they aren’t just providing an API or a model; they are building an entire ecosystem of products around that API. This includes their SDK, Claude for orchestration, and various other tools. They refer to this as a “harness”—the software surrounding the API that allows users to get the most out of the model.
Alex Sacerdote: We saw a similar dynamic with AWS back in 2013. At the time, people dismissed it as just a commodity server in a warehouse. What they missed was that it represented a new way of doing computing. Amazon developed a suite of products before anyone else, which gradually built customer lock-in. We also look at where we are on the S-curve. We believe the infrastructure layer is only about 10% penetrated. In fact, we think it remains one of the best ways to play the AI trend, especially when you consider how that value feeds back through the system. Even though...
10:20-12:24
Alex Sacerdote: Even though hundreds of millions of people are using AI, they are mostly just using “AI 1.0,” which is essentially a search engine on steroids. But now, with these new primitives—like having Claude integrated directly into your computer—you can start building skills. Individuals and companies are going to start developing these skills, leading to the creation of true AI bots, while large corporations will build even more massive systems.
Where are we currently in terms of adoption? Sundar Pichai mentioned that only about 10 basis points of the world’s knowledge workers are truly utilizing this. Anthropic has around 14 or 15 million daily active users, but likely only a small fraction of them are using AI to its full potential. That 10-basis-point figure represents the classic start of an S-curve; these are the tinkerers. Next, it will move to early adopters, then to the early mainstream. We are going to go from 10 basis points to 3%, 5%, and eventually 15% over the next four years. A light switch went off in the enterprise sector this year; everyone realizes they need to implement this now and they need to do it fast.
Alex Sacerdote: It’s still reminiscent of Internet 1.0. Back in 1998, you knew you needed a website, but it was actually quite difficult to build one. Today, the technology is coming together rapidly. We believe the enterprise AI application market is currently less than 1% penetrated. While we usually talk about S-curves, we call this an “L-curve” because it’s just going straight up.
If you look at the infrastructure side, it’s even more intense. Even with only 10 basis points of people truly using AI, we are already sold out of capacity; there simply isn’t enough compute in the world. Anthropic, for example, has only half of what they currently need, and that’s before this massive wave of adoption truly hits.
Alex Sacerdote: Right now, we are seeing massive demand even before the full take-up. Marc Andreessen said that over the next four years, the one thing he’s certain of is that there simply won’t be enough compute to go around.
12:24-14:16
Patrick O’Shaughnessy: Most software companies try to maximize the time you spend on their app to juice engagement. Ramp does the exact opposite. Ramp understands that no one wants to spend hours filing or reviewing expense reports and checking for policy violations. So, they built their tools to give that time back, using AI to automate 85% of expense reviews with 99% accuracy. Since Ramp saves companies an average of 5%, it’s no wonder that Shopify, Stripe, and my own business run on Ramp. To see what happens when you eliminate the busy work, check out ramp.com/invest.
Patrick O’Shaughnessy: OpenAI, Cursor, Anthropic, Perplexity, and Vercel all have something in common: they all use WorkOS. Here’s why. To achieve enterprise adoption at scale, you have to deliver on core capabilities like SSO, SCIM, RBAC, and audit logs. That’s where WorkOS comes in. Instead of spending months building these mission-critical capabilities yourself, you can use WorkOS APIs to have them all ready on day zero. That’s why so many of the top AI teams you hear about already run on WorkOS. It is the fastest way to become enterprise-ready and stay focused on what matters most—your product. Visit workos.com to get started.
Patrick O’Shaughnessy: Every investor should know about Rogo. Rogo AI’s platform isn’t just another generic chatbot; it was designed to support how Wall Street bankers and investors actually work—from sourcing, diligence, and modeling to turning analysis into deliverables. For me, three key things differentiate Rogo. First, it connects directly to your systems so it can work with your actual data. Second, it understands your workflows and how work really happens.
Patrick O’Shaughnessy: Rogo understands how work actually happens across a deal or an investment. It runs end-to-end and produces real outputs just like your best employees do—auditable spreadsheets, investment memos, diligence materials, and slide decks that match your specific standards. This is possible because Rogo was built by finance professionals, for finance professionals. It is already being adopted by some of the most demanding institutions in the world. To learn more, visit rogo.ai/invest.
14:16-17:15
Patrick O’Shaughnessy: I’m curious about your transition. Historically, you were a public markets investor where you could simply hit “buy” on whatever you wanted. Now, you are operating in many of the most important private companies, such as Stripe, Databricks, OpenAI, and Anthropic. How do you secure positions at the size you want, given that you can no longer just buy shares on the open market? How much of that involves direct creativity with the company? Since private deals require a double opt-in—meaning they have to choose to let you in—what have you learned about securing the allocation or equity you want, especially since this wasn’t your original background?
Alex Sacerdote: In that case, we really took the time to get to know the company. One of our analysts had connections within their finance group. We actually looked at the $60 billion round but decided not to participate at that time. We didn’t know the company well enough yet, their gross margins were negative, and frankly, we hadn’t yet seen the explosion in AI coding that eventually took place. One advantage of the public markets is that you can get to know companies over a long period of time before committing.
Alex Sacerdote: Over a long period of time, you can essentially invest on your own schedule. I had the opportunity to spend some time with Dario Amodei, and I’ve obviously listened to him on various podcasts. I began to realize that their management team is excellent. Their focus and dedication are impressive; they’ve had almost no turnover, and the quality of their code is top-tier.
Furthermore, their business plan was really starting to play out. It’s one thing to grow from $100 million to $1 billion, but it’s another thing entirely to reach $9 billion. We reached out to the company as much as we could, and they eventually took a meeting with us. We prepared a 90-page PowerPoint deck where we actually used Claude to scour the internet for feedback regarding the coding market—identifying what their products were good at and where they might need to improve. We also presented our comprehensive overview of what the coding market would eventually become.
They welcomed us into that funding round, and we’ve stayed close with the CFO ever since. It’s been great to build a relationship with them, and I think we punched above our weight in terms of our allocation. That investment was a total home run.
Looking at the broader landscape, we are currently in a period where the unicorn market is larger than most stock markets in Europe—perhaps even combined. It’s definitely bigger than the markets in Germany or the UK. Even before we started investing in private companies back in 2020, we made it a point to meet with them. You have to know these companies now because, in many cases, they are the dominant players in their space and have a massive impact. We conduct about 2,000 to 3,000 face-to-face meetings with management teams every year. About 10% to 15% of those are with private companies. We focus on the ones we really want to learn about and find ways to get involved.
17:18-19:26
Alex Sacerdote: We wanted to meet with them and get involved in their funding rounds. Our first major private investment was Stripe. At the time—this was around 2017 to 2019—we had a large position in Adyen, which is a fantastic next-gen cloud payments company. They were taking market share from incumbents like Worldpay. Even though modern cloud payments only represented about 5% of a massive $80 trillion market, you couldn’t truly understand Adyen unless you knew Stripe like the back of your hand.
We did a tremendous amount of due diligence, speaking to 200 Adyen customers. Every time we asked about Adyen, we also asked about Stripe. We realized it was a “Coke and Pepsi” situation. We knew we had to find a way to invest, and I finally had the opportunity to meet the Collison brothers in 2019. That was our first foray into that space.
Alex Sacerdote: At the time, we weren’t really known for private equity. I have a friend involved with a venture firm that held a significant stake, and I told him, “Let me know if you ever want to sell some.” I eventually got a call from him in April 2020, during the height of COVID.
By then, we knew a lot about Stripe. We didn’t have their full financials, but we knew enough to realize that at a $35 billion valuation, it was a steal. They had disclosed over half a trillion dollars in Total Payment Volume (TPV). We knew Adyen’s take rate was around 25 to 30 basis points, while Stripe’s was closer to 40 or 50. We also knew their headcount, so we could estimate their profitability.
It turned out the take rate was even higher than we thought, and they were being modest about their TPV; it was actually closer to $1 trillion than $550 billion. We underwrote the investment based on our assumptions, and the reality was much better. We were eventually able to upsize that into a $100 million block from the seller. Sometimes, the market just gives you those opportunities.
Alex Sacerdote: Sometimes they appreciate that the VCs are going to own a block and then most of them will eventually sell. They like the fact that we will hold our position into the public market, which is exactly what we did with NuBank. We owned it for a long period of time as a public company as well.
19:26-21:32
Patrick O’Shaughnessy: Maybe now is the right time to lay out everything you’ve learned about S-curves. Obviously, your firm is predicated on the idea of technology adoption life cycles and investing in companies at the right time amidst a specific platform shift or S-curve change. I think everyone knows the basic concept of an S-curve and the stages you mentioned—tinkerers, early adopters, and the early majority. But I’d love for you to go into the granular detail of what you’ve learned, since this is the lens through which you’ve viewed markets and stocks for a long time. Bring us into the nitty-gritty, nuanced detail of why S-curves are so useful for investing.
Alex Sacerdote: We have a three-part investment framework: the S-curve, competitive advantage, and underappreciated earnings power. We can dive into each one. When you catch the right part of the S-curve, you get exponential unit growth. If you have a very strong business model—and in tech, there are so many different types of moats—your earnings don’t just grow linearly; they grow exponentially. That leads to the final piece: investing when there is underappreciated long-term earnings power. Very often, earnings can grow from $1 to $10, then to $50, and eventually to $100. It happens far more often than you’d think, and it allows you to buy some of the best companies in the world.
Alex Sacerdote: We’ve been able to buy the best companies in the world at extremely low P/E ratios. When we were buying Nvidia in 2023, we were paying four times forward earnings. When we bought Tesla in 2019 for the electric vehicle S-curve, we were paying five times earnings. When we owned Apple, we were paying four times earnings. When we bought Amazon for AWS, we were essentially getting that business for free.
The world doesn’t tend to think exponentially; people are so focused on the next year or the next quarter. Very few people believe you can accurately predict two, three, or four years out. But if you follow and understand the S-curve, identify the moats, and know how to model them, you really can predict these great outcomes.
21:32-23:25
Alex Sacerdote: Let’s dive into the S-curve. It’s a crucial concept because every technology follows this specific pattern. A technology is introduced, but it often lingers for years before hitting an inflection point. For example, smartphones existed for ten years before the iPhone. The internet was around for twenty years before Netscape. AI has been hidden inside companies for a long time, but it wasn’t until ChatGPT launched that it was truly “taken public” and ignited.
Similarly, Tesla went public 15 years before 2019, which is when the stock finally went vertical. This delay happens because there are usually significant barriers to adoption. The first smartphones were clunky, lacked touchscreens, and didn’t have a robust wireless data system. They were also too expensive, costing $500 or $600. Steve Jobs eventually got the price down to $200, AT&T launched a 3G network, and the touchscreen made the device so intuitive your grandmother could use it. Apple built an ecosystem and made it simple. Once those barriers to adoption are eliminated, the technology rockets.
Alex Sacerdote: Growth really rockets once those initial barriers are removed. That creates a “tornado of demand” where everyone globally realizes they need the product immediately. We saw this shift happen with electric vehicles. Initially, the price was too high and range anxiety was a major deterrent. Then Elon Musk brought the price down to $40,000, increased the range to 300 miles, and finally got the supply chain in place to churn out millions of units. Those factors triggered the inflection point.
Alex Sacerdote: The other nuance is that it’s not just about identifying that a trend is taking off; you have to understand the height and scale of the S-curve. You need to know how tall it is so you can determine when to sell and how long to hold. Since we are underwriting our investments two or three years out, we have to understand what the growth trajectory looks like beyond that horizon.
23:25-25:25
Alex Sacerdote: These S-curves can be quite dynamic. Take Amazon’s AWS, for example. Back when it was just a hidden line item within Amazon, it was being covered by retail internet analysts rather than hardware or chip analysts. It was a completely new business model. However, we realized the Total Addressable Market (TAM) for AWS was the largest in the history of enterprise IT. Previously, the TAM was fragmented across routers, memory, storage, Dell, and EMC, but AWS was consolidating all of it. We calculated that they were directly addressing a $600 billion IT systems market.
Alex Sacerdote: Initially, we thought the technology might be 50% deflationary, which would have meant we were only 1% or 2% penetrated. But over time, we realized it wasn’t actually deflationary. If you talk to anyone today, they’ll tell you that building it yourself costs about the same as using the cloud. This meant the TAM was significantly larger than we first thought. There are mega S-curves and there are sub-S-curves. We’ve been fortunate to invest through Internet 1.0, mobile, cloud, and e-commerce. Now we have AI, which we can confidently say is the biggest S-curve of them all.
Alex Sacerdote: The biggest S-curves often build upon one another. Take the electric vehicle S-curve, for example. You have to pay close attention because, at the time, we estimated that perhaps 40% to 50% of cars would go electric, but the trend hit a major wall at around 10% or 15% penetration. Usually, S-curves go all the way to the top, but in this case, for a variety of reasons, it didn’t. You have to be ready to adjust and stay on top of the data. Generally, once a technology reaches 30% to 40% penetration, you lose that exponential growth. That’s when the sell-side analysts catch up, and you stop seeing those massive earnings beats.
Patrick O’Shaughnessy: Is that typically when you decide to sell?
25:25-27:30
Alex Sacerdote: Generally, yes. We prefer high-growth periods. We actually made a mistake with Apple in this regard. For the first five or six years, Apple was an incredible investment for us; it was our largest position and would go up 50% to 70% a year, with the exception of 2008. We sold in 2012 when smartphone penetration in the US reached about 50%. Now, Apple maintained its leadership position, but it had a couple of years of underperformance. Eventually, the multiple compressed, they added several ancillary services, and they benefited from the app ecosystem by taking a 30% cut of app sales. They were able to continue compounding nicely at around 20%, but the massive gains occurred during the move from 0% to 50% penetration on the curve.
Patrick O’Shaughnessy: I’m fascinated by that long flatline that can last a decade or more at the beginning of these curves. What have you learned about the right moment to start paying attention or to actually buy? How do you measure that? Is it always—
Patrick O’Shaughnessy: How do you measure that? Is it always different? What are the pitfalls you’ve fallen into? We talked about when to sell, but how do you know when to start thinking about buying into one of these things?
Alex Sacerdote: Yeah. You know, Andy Grove says that when you hit strategic inflection points, you can’t trust the data. Strategic inflection points are about intuition and anecdotal evidence. I love this book called *The Tao Jones Averages*, which is a guide to “whole-brain” investing—using both the right and left brain. The best investors utilize the creative side, where it’s visual and about connecting the dots.
For example, we invested in the mobile video game S-curve for a long time. Initially, mobile games were limited because phone screens were small and processing power was weak, so you only had casual games. But then I was in China and saw a twelve-year-old boy with a huge phone playing an incredible video game. I realized, “Oh my god, high-end gaming is coming to the phone.” It was a visual realization.
27:30-29:57
Alex Sacerdote: Enterprise is harder because you can’t see it as easily. We go to the Gartner IT Symposium, where 30,000 CIOs gather. We saw this happen with Splunk back when they were an amazing database company; their presentation room was standing room only. We saw it with VMware maybe twenty years ago when they virtualized the server—again, standing room only. You could just see the corporate demand beginning to swell. It was the same with AWS. We went there and the grand ballroom was completely packed at nine o’clock, ten o’clock, and eleven o’clock.
Alex Sacerdote: The ballroom was completely packed by 11:00 AM. You could actually see the demand exploding before it even happened. We look for all kinds of clues, and there is a specific pattern recognition that takes place. By the way, it is okay to be late. In many cases, it is fine to miss the first one, two, or three years because if the top of the S-curve represents half a trillion dollars, the growth can continue for a long time. You don’t always have to be there at the very start; it is okay to miss the first 100% gain. I started my career at Fidelity, and Peter Lynch loved to mentor the young kids, so I got to spend some time with him. He used to say, “White out the chart. It’s all about the future.”
Alex Sacerdote: So, it is okay to miss the beginning, but what helps with the S-curve is understanding its duration. Then there is the slope of the S-curve, which is also important. A lot of people think that because we live in a modern world, everything moves fast, but there are many factors that determine the actual pace of adoption.
Alex Sacerdote: We commissioned a gentleman named Horace Dediu, who used to work with Clayton Christensen, to look back through history. We actually have the major S-curves of the last 100 years displayed on our wall. The radio S-curve was one of the fastest ever; it took only seven years to reach nearly 100% penetration. However, the dishwasher S-curve is much slower because the product needs to be integrated into the home’s infrastructure.
Patrick O’Shaughnessy: That’s fascinating. What else did you learn from that study?
Alex Sacerdote: B2B technology can take a long time because it has to be plugged into existing systems. Much like the dishwasher, it has to be installed inside the house and integrated.
30:00-32:27
Alex Sacerdote: Consumers generally tend to adopt new technologies much faster than enterprises.
Patrick O’Shaughnessy: I love that the radio and the dishwasher are your two models for adoption.
Alex Sacerdote: Yeah. I covered the internet at Fidelity. My first stock was Amazon, which is a fun story for another time. But I also covered B2B internet, and back then, there was a massive bull case for it. However, the underlying infrastructure simply wasn’t in place for B2B to happen yet. It ultimately took another 20 years to materialize with SaaS.
That is a potential risk with AI. Large companies are very security-conscious and can be slow to move. There are cultural issues with AI where you really need internal evangelists and top management to push it through. Meanwhile, IT departments are often saying, “This is too risky.” We saw the same thing happen with the cloud. Initially, everyone was afraid it was insecure to have data in the cloud. Then we saw the CIA adopt it, followed by Capital One. We actually spoke to the Capital One CIO, who argued that data is actually more secure in the cloud. That’s when it really started to take off.
Alex Sacerdote: But those takeoffs were gradual. Perhaps because SaaS and cloud are like the dishwasher—they have to be “plugged in” to the existing infrastructure—they grew at a steady 30% to 50% rate. What’s amazing about AI is that, at least for consumers and even some businesses, you don’t need a complex installation. You just open up...
Patrick O’Shaughnessy: The browser, and it’s right there.
Alex Sacerdote: Exactly. And that’s why we’re seeing this vertical, “straight up” adoption curve.
Patrick O’Shaughnessy: And I think there’s enough runway in the near term, moving from a tiny fraction of people really using it to maybe 2% or 5% of the population, which is going to cause it to explode.
Alex Sacerdote: This momentum is going to cause the growth to keep heading straight up. We call this a “backwards L-curve,” and it’s a really exciting development.
Patrick O’Shaughnessy: What have you learned about the moment a leader separates themselves from the competitive pack? You’ve discussed the overall growth of the S-curve and the demand behind it, but there are always multiple players fighting for dominance. It seems like your strategy is to invest after someone has already pulled away from the pack, rather than trying to pick a winner from the middle of the fray. Is that a fair assessment?
32:27-34:27
Alex Sacerdote: Well, we definitely look for the S-curve first. Then, we conduct an exhaustive study of every company with exposure in that area to find the one with a truly powerful competitive advantage. Historically, many people—including Warren Buffett—didn’t like tech because things moved too fast to predict the future. For us, the S-curve serves as our map for looking ahead.
People were also worried that tech was so disruptive that you could never trust a company to be a long-lived asset. However, what we’ve found over the years is that competitive advantages in the digital world are often just as powerful, if not more so, than those in the offline world.
You have the network effect, which was incredibly powerful for companies like LinkedIn, Facebook, and Alibaba. Then, you have companies that become the industry standard, like Oracle or Bloomberg. Take Oracle, for example: they charge a premium, and while there are free or open-source alternatives, Oracle owned the ecosystem. They had all the database administrators and all the software tuned to work specifically with them. They essentially became the bedrock of the industry.
Alex Sacerdote: Oracle, for example, basically had a stranglehold on the relational database market forever. In this industry, you can reach massive scale very quickly because these S-curves grow so fast. Suddenly, a company like Anthropic is doing billions in sales, or Amazon achieves Walmart-sized scale in just five years compared to the 40 years it took Walmart.
When you have that kind of growth, you can leverage network effects and become the industry standard—a platform that everyone else builds on top of. You can also secure critical intellectual property. Qualcomm is a great example; for a long time, you couldn’t make a phone without paying them. Similarly, ASML has critical IP today; you simply cannot manufacture a high-end chip without their lithography machines.
34:27-36:28
Alex Sacerdote: What’s interesting is that these AI foundational companies already have that scale. Another key factor is brand. Brand is vital because companies like Google and Amazon were able to grow without ever having to advertise. Elon Musk has never had to advertise for anything. When your customer acquisition cost is zero, it completely changes the lifetime value equation and the entire business model. Most of the successful companies I’ve mentioned, like Apple, have all of these advantages rolled into one.
Sometimes we can identify these advantages before the rest of the market does. One of our highlights was pitching Amazon specifically for AWS back in 2013 at the Robin Hood Investors Conference. We argued that even the bulls had no idea what they were sitting on. We said Amazon had won the war before it even started. At the time, we described it as “Coke with no Pepsi.” It turned out there was a Pepsi eventually, but the market was big enough for both. We could see they had a seven-year lead. Being a first mover is important, but they also became an entire ecosystem and platform. They reached a scale ten times larger than anyone else, meaning no competitor could afford the R&D necessary to catch them.
Alex Sacerdote: You’re right that if you don’t have a competitive advantage, you can be in the best S-curve of all time and still lose out.
Patrick O’Shaughnessy: Exactly. If your name was RIM, Palm, Nokia, HTC, LG, or Motorola—I could go on forever—it was just negative after negative. We saw the same thing at the foundational model layer, where there were fifty companies trying to compete. Most have fallen away, and two or three have emerged at the top. There are plenty of reasons to think they will continue to hold their positions.
Alex Sacerdote: Google is a little trickier because they have that massive, complex legacy business attached to Gemini. But if you look at Anthropic and OpenAI as pure plays and reason through their competitive advantages, why aren’t they susceptible to the erosion of those advantages over time?
36:28-38:11
Patrick O’Shaughnessy: Of all the S-curves we’ve analyzed, AI is by far the most complex and the fastest-changing. We have to keep in mind that while there are risks, the rewards are also the highest because we’re talking about a market in the trillions. We previously estimated the cloud market at maybe $800 billion; we now think this could be $3 to $5 trillion. It’s higher risk, but higher reward.
Patrick O’Shaughnessy: Take Anthropic, for example. It appears they have critical intellectual property, and they’ve been able to maintain a high market share in coding applications. Secondly, they’ve built a strong brand within the enterprise. If you talk to any CIO, the first thing they mention is Claude. They are reaching escape velocity.
Alex Sacerdote: These companies are going to achieve escape velocity and scale. What was initially daunting for OpenAI and Anthropic in their fight against giants like Google was that the incumbents had these massive cash cows. To the credit of both management teams, OpenAI and Anthropic were able to operate in these incredibly capital-intensive industries and find ways to raise the necessary funds.
With Anthropic specifically, given their 10x sales growth and fundraising prowess, it looks like they’ve reached escape velocity. They now have scale. Another advantage Anthropic and OpenAI possess is the concept of recursive improvement. For instance, now that Anthropic is leading in code, they can feed that code back into their own models. If you look at the pace of their innovation, it’s actually accelerating.
38:11-40:04
Alex Sacerdote: They may be entering a true liftoff stage. OpenAI was focused on many different sectors, but they’re starting to perform better in the enterprise space, their coding tools are strong, and they’re seeing accelerating growth there.
Right now, the enterprise market looks much more promising than consumer because customers are willing to pay a premium when the technology replaces human labor. On the consumer side, you might rely on advertising, though people might pay for a high-quality assistant—like a “Claude-bot”—if it worked perfectly. Regardless, they already have a massive number of eyeballs on those platforms.
You’re right that things shift, but in the charts we use for almost all of our pitches, the trend is clear: on the internet, the leader grows bigger and faster, and ultimately wins. Most of the time, the leader maintains that position. Look at Shopify—once they became the leader, they just kept going.
Alex Sacerdote: The leaders just keep going. Amazon, as the leader, continues to dominate. With SaaS companies, once you establish a lead, it compounds. On the internet, success compounds on itself. Another critical factor is scale. You need massive amounts of compute, and you have to be able to pay for it. There are only a few players who can actually do that.
Those are some of the moats we see emerging now. Of course, there are exceptions. Usually, during paradigm shifts, some leaders fail to adapt—like when AOL couldn’t make the transition from dial-up to broadband, or when Netscape arrived early but lacked a sustainable business model. However, if you talk to anyone in Silicon Valley or any startup founder, they’ll tell you they are building on top of the “Big Three” foundation models. The global economy is vast enough that they will find ways to differentiate within those ecosystems.
40:04-42:12
Patrick O’Shaughnessy: I’m curious what you think all of this means for the software sector. When I look through your portfolio, I don’t see many large enterprise software companies. I’m not sure if you held them previously and sold out, but it’s hard to use these incredibly useful AI tools—even if they still feel like toys—and not think, “Wow.” Even a non-technical person could potentially build an ERP replacement for their own company if they spent enough time with these tools.
There doesn’t seem to be a fundamental reason why that isn’t possible, which suggests traditional software companies could be in a lot of trouble. Everyone seems to have a strong opinion on this one way or the other. Given that you don’t own many of them, how are you approaching those types of companies?
Alex Sacerdote: At certain points, perhaps five years ago...
Alex Sacerdote: At certain points, maybe five years ago, we might have had 40% or 50% of our portfolio in software. Early on, in our April 2023 seminar, we advised investing in chips first. At the time, we thought the application layer would be huge. These companies have massive sales forces, they can leverage AI APIs to build products, and they own the data. We thought this was going to be amazing for software.
However, we quickly realized that their AI products weren’t very good. They weren’t moving the needle, and no one could figure out how to charge for them. Consequently, we sold almost all of our application software holdings. We still have one or two small positions, but entering this year, we were actually net short on the sector, which really helped our performance in the first quarter.
There are so many layers to this. The old way of building software is like using a pen and paper, or a horse and buggy. The new way is like a jet engine—or frankly, like the transporter from Star Trek. It is such a revolutionary change that it feels like it has to be disruptive.
42:14-44:26
Alex Sacerdote: Even if it isn’t disruptive immediately, software companies face another problem: they have fallen significantly down the priority list for CIOs. Even if AI isn’t directly disrupting a specific software tool yet, companies are shifting their spending toward things like Anthropic tokens because the ROI is faster there.
Second, because they are spending so much money on AI, it puts pressure on the rest of the budget, which hurts traditional software vendors. Third, many software companies used to raise prices every year, but now they are likely nervous about doing so. Fourth, we have to see what happens with jobs. There are smart people on both sides of that debate, but we are...
Alex Sacerdote: We are seeing people on both sides of this issue. Some companies are really gutting their staff or freezing hiring, which hurts the seat-based revenue model. Perhaps they are focusing on building their own internal applications instead.
Alex Sacerdote: If you want to be optimistic, it’s simply taking them a while to execute. We discussed how early the primitives of AI still are, so it might just be taking time to develop something they can truly commercialize.
Alex Sacerdote: However, they might not have the right talent. Selling a fixed system is a completely different motion than installing something that performs human-level work. To ensure the work is actually getting done, you need forward-deployed engineers (FDEs), and many of these companies may not have the right internal teams for that.
Alex Sacerdote: Then, of course, there is the risk that customers will just build it themselves. The bulls will argue that a company is never going to build its own ERP system, and they are probably right. Historically, old technology is very sticky.
Alex Sacerdote: Mobile video games didn’t kill console games; tablets didn’t kill the PC; and the smartphone didn’t kill the PC either. There is a massive amount of integration and work that goes into these software suites. It’s also true that companies generally prefer to buy solutions rather than build them from scratch.
44:27-46:52
Alex Sacerdote: That is all true, but you can certainly imagine a world where, in the next few years, a brand-new AI-native company goes after each of these strong incumbents. Their data advantage could be obviated, and AI might make it much easier to rip out the old system and plug in a new one. So, if you like software, the current valuations are...
Alex Sacerdote: Valuations are very high right now, and everyone knows they are under pressure. Some people are tempted to buy into these companies, but AI coding tools are just getting better and better. We’ll have to wait and see. We are watching these software companies very closely to see if they can generate any revenue that might change their current trajectory. However, it’s difficult. If you’re a company like Salesforce with $40 billion in sales, and you only have $500 million or $700 million in AI-related ARR, you’re dealing with a massive legacy base. While it might eventually start to work, it takes a long time. In software, we have the “Rule of 40,” which is your growth rate plus your operating margin. If you have a 20% growth rate and a 20% margin, that’s considered good.
Alex Sacerdote: For AI, we have a new version of the Rule of 40. We actually use it primarily for chip investing. It looks at what percentage of your sales are AI-driven—say 30%—and what your market share is in that category—say another 30%. That gives you a score of 60. That’s a great place to look because it shows you have both exposure and a strong market position. The problem with software right now is that their AI revenue is only at 1% or 2%, so they have a long way to go.
Alex Sacerdote: One thing we are picking up on lately—and this is still a bit half-baked—is that AI could actually make some of these software platforms more important. For example, what’s the first thing you do with Claude? You plug it into Slack. If Slack can become a key repository for that data, it becomes a permanent fixture within the organization. We may find that the next wave of AI consists of agents that use these tools, operating inside existing incumbent software just like a human being would.
Patrick O’Shaughnessy: To pull on that thread, it seems like the commonality of the tools they might use—the ones that are the most sticky—would be...
46:54-48:38
Patrick O’Shaughnessy: The tools that are the most sticky are often network-based. Slack is a great example. The software within Slack itself might leave something to be desired—it isn’t necessarily the special part. The special part is that everyone is already there. I’m curious what other qualities you look for. Is a network effect the only thing that really matters?
Alex Sacerdote: It’s still early in our thinking, but consider systems like Workday, HR platforms, or the big systems of record. AI agents may end up running on top of them. There is a bear case for CRM where it goes “headless” and gets relegated to being just a database. In that scenario, you lose the human interface, and the AI interface becomes “no interface”—the agents just go straight into the data. You lose that direct customer interaction. However, if the agents are going into the CRM to perform the work, it actually solidifies the CRM’s importance. It ensures the platform isn’t going away.
Patrick O’Shaughnessy: Can we talk about chips? You’ve referenced them a few times—infrastructure, chips, and everything surrounding the data center. Why is this area so interesting to you? I love your modified “Rule of 40” for hardware, looking at the percentage of revenue that is AI-related and the market share within that category. That’s an interesting metric.
Alex Sacerdote: Which companies shine on that metric today, and which are the surprising laggards? For the past 40 years, almost nothing changed in the data center. Even with the move to the cloud, we were basically using Intel x86 architecture. It became the standard data center chip sometime ago.
48:40-50:50
Alex Sacerdote: The CPU became the standard data center chip sometime in the ‘90s. Throughout the cloud era, compute workloads grew by about 25% to 40% every year, but Moore’s Law was improving at that same rate. Consequently, it didn’t require tremendous innovation. For years, there was almost no growth in hardware; the entire industry basically commoditized every component. Every chip, every part of the server—from the printed circuit boards and memory to the enclosures and networking—became a commodity. There was very little innovation. For example, moving from one gigabit to ten gigabits would take seven years. While the initial switch to ten gigabits required some innovation and created a small cycle, it would quickly commoditize again.
Alex Sacerdote: Now, look at AI. Workloads are growing 10x every year, pushing every single aspect of this hardware to its physical limits. This isn’t just creating tremendous unit growth; it’s leading to what we call the “de-commoditization” of the hardware industry. I met with Shaun Maguire about three years ago, and he said, “I wish I could come back and run a hardware hedge fund because all the companies are public and they all have powerful IP.”
Sequoia made some of their best investments back in the hardware era with companies like Apple and Cisco, and we are now entering a similar renaissance for chips. Not only do you have massive unit growth, but the sector requires tremendous innovation at every level of the server. For instance, memory used to be a pure commodity, but now we have high-bandwidth memory with ten chips stacked on top of each other. The input/output speeds are now 10x what they used to be.
50:52-52:44
Alex Sacerdote: The input-output requirements are now ten times what they used to be. For example, it took Samsung years to master this, and it’s an absolutely critical component that is constantly being upgraded. They have to work with Nvidia three or four generations in advance.
We saw a similar situation with Celestica. Celestica is a contract manufacturer, which has been a disastrous industry to be in since 1999. Most of that work went offshore to China and became a commodity business. However, Celestica hung on and retained their heritage in IBM supercomputing, keeping all that specialized talent and skill.
About three years ago, we noticed they were the sole supplier for the Google TPU server. We thought, “Oh my god,” because the stock was trading at only eight times earnings. They also had a business selling “white box” Ethernet switches—which is essentially a code word for commodity hardware—directly to cloud providers.
It turns out these are excellent businesses. Not only are they seeing tremendous growth, but building an AI server is a different beast entirely. They are liquid-cooled because they run so much hotter. We’re talking about a $200,000 to $300,000 piece of machinery, whereas an old server cost $5,000. With the old ones, if it broke, you just threw it away. If an AI server breaks, the whole system goes down.
Because of that, you become critical infrastructure—it’s like selling a vital part for an airplane. You’re never going to be swapped out. It also turned out that Celestica was quite good at liquid cooling, while many others tried and failed, allowing them to maintain their position. Furthermore, the Ethernet market shifted because...
52:48-54:51
Alex Sacerdote: The Ethernet market has changed dramatically. In the old days, you would transition from 100 gig to 400, then to 800. It used to be a seven-year upgrade cycle. Now, they are upgrading every single year, which is incredibly difficult to execute. Then there is the entire software layer—the open-source SONiC layer. Some of the people at Celestica actually wrote that open-source software, and they work very closely with Broadcom. What we initially thought was just a great growth driver turned out to be a massive competitive advantage. They now hold about a 50% to 60% share of the cloud Ethernet switch market, which is crucial because AI is incredibly network-intensive.
Alex Sacerdote: Even something as seemingly simple as the printed circuit board (PCB) has evolved. A regular server might require 10 layers, but these AI servers require 40 layers. There are very few PCB suppliers capable of manufacturing these, and the complexities involved are significant. We also own Elite Material, which produces the leading ingredient—copper-clad laminate—that goes into these boards. So, you have a situation where PCB units are growing and layer counts are rising.
Alex Sacerdote: You are looking at a 50% to 60% CAGR (compound annual growth rate) just in units, while ASPs (average selling prices) and gross profits are also rising. Your visibility has shifted from “we’ll call you next week if we need you” to “we need you to design this roadmap with us for the next four years.” These companies have gone from 5% growth and low margins to a 35%, 40%, or even 50% top-line CAGR for the next four years with expanding margins. On top of that, there are shortages of everything. Even if a product is technically a commodity, it’s going to be a fantastic cycle. We see this up and down the supply chain with companies like Corning, who make the fiber. They have some ridiculously high...
54:52-57:01
Alex Sacerdote: Corning has a ridiculously high share of the fiber market. I was reading about a Microsoft data center they just built; there is enough fiber in that one facility to circle the world four and a half times. Their fiber is thinner, more bendable, and can be specially manufactured to exact specifications. It’s a higher-margin product and the fastest-growing part of their business.
In networking, you have “scale-out,” which connects all the server racks together, and “scale-across,” which connects different data centers. When you want to build these massive clusters but can’t get all the power in one location for training, you have to wire them together. However, the wiring requirements are ten times greater; the cables have to be much thicker. This is creating massive growth.
The real kicker comes with “scale-up,” which involves connecting every GPU within a rack to the others. Currently, that’s done over copper, but eventually, it will transition to fiber. When that happens, it will double or triple Corning’s total addressable market. At every layer of the rack, the demand is staggering.
Patrick O’Shaughnessy: Everyone is overwhelmed.
Alex Sacerdote: Everyone is overwhelmed. But look at the story with power supplies: every Nvidia chip or rack uses 50% to 125% more power. That literally drives up the average selling prices (ASPs) for companies like Delta and Advanced Energy. I find these stories hard to believe when I first hear them. I’m thinking, “Wait, your ASPs are going to increase by 40% every year for the next four years, and it’s higher margin?”
The broader picture is that if we are right about this “L-curve” of AI demand, we are already facing shortages. We are already 30% short in the DRAM market, the NAND market, and the PCB market. We are hitting supply constraints across all these different components.
Alex Sacerdote: We are currently 30% short on all of those things.
57:01-59:24
Patrick O’Shaughnessy: As your business scales up, everything gets more complex—especially your compliance and security needs. With so many tools offering band-aids and patches, it is unfortunately far too easy for something to slip through the cracks. Fortunately, Vanta is a powerful tool designed to simplify and automate your security work, delivering a single source of truth for compliance and risk. There is a reason that Ramp, Cursor, and Snowflake all use Vanta. It frees them to focus on building amazing, differentiated products, knowing that compliance and security are under control. *Invest Like the Best* listeners get a special offer of $1,000 off Vanta when you go to vanta.com/invest.
Patrick O’Shaughnessy: I know firsthand how complex the tech stack is for asset management firms. Seemingly every new tool and data source makes the problem even worse, adding more complexity, more headcount, and more risk. Ridgeline offers a better way forward: one unified platform that automates away that complexity across portfolio accounting, reconciliation, reporting, trading, compliance, and more—all at scale. Ridgeline is revolutionizing investment management, helping ambitious firms scale faster, operate smarter, and stay ahead of the curve. See what Ridgeline can unlock for your firm. Schedule a demo at ridgeline.ai.
Patrick O’Shaughnessy: Regarding the metrics of AI percentage and market share: do you care more about the absolute numbers or the rate of change?
Alex Sacerdote: That’s a good question. We did a presentation in early 2024 where we actually listed everyone’s market share. I asked Claude to plot it for me, but it didn’t quite get it right because...
Alex Sacerdote: The market actually didn’t get it right because it failed to account for the rate of change. The rate of change is incredibly important. When you go from 10% to 30% penetration, your growth rate accelerates and your margins expand. That shift is a critical factor.
Patrick O’Shaughnessy: Why don’t more people in the public markets get this right? If your entire framework is based on S-curves, competitive advantage, and underappreciated earnings power, it feels like a movie that has played out repeatedly over the last 25 to 30 years.
Alex Sacerdote: My mom once asked me, “Why do you tell everyone your secret?” It’s like asking why a casino teaches people how to play blackjack. The truth is, it’s just really hard to do. You have to be deeply comfortable with this style of investing. I’ve been focused on tech for 20 years at Whale Rock, and we have a team that has covered many different cycles. We understand the nuances. For a long time, almost no one was paying attention to hardware and chips, so now you have all these newcomers entering the space.
Patrick O’Shaughnessy: It’s basically just you and Gavin Baker.
59:24-1:01:11
Alex Sacerdote: Exactly, and Gavin has done a great job. Most people weren’t comfortable with the sector, and it’s much harder than it looks. When you look at the charts for these companies, they’ve already gone up significantly, which makes people scared to buy. You need a holistic view to maintain conviction. For example, with Nvidia over the last four years, the narrative is always: “They had a great year, oh my god, it must be a bubble.” Then they have another great year, followed by six months of the stock treading water, and people again insist, “It’s a bubble; this is getting out of hand.” It’s scary for people because the bear cases aren’t entirely baseless.
Alex Sacerdote: The bear cases aren’t totally without merit, but you have to see the whole picture to understand how these things are unfolding and gain conviction. Frankly, if you’re just a semiconductor analyst, you might have missed it. So many of them did because they didn’t see what was really happening at the foundational model layer. It helps to have that big-picture perspective, informed by decades of looking at scores of S-curves and seeing how they play out across different sectors.
Patrick O’Shaughnessy: In this whole picture, I would describe your stance so far in our discussion as very bullish on the impact AI will have and the returns available as a result. What makes you the most concerned or uncertain? Is it just the rate at which all of this changes? What keeps you worried amidst what seems like pretty extreme bullishness?
Alex Sacerdote: One thing that bothers me is the amount of negativity in the general population regarding AI, as well as the negativity coming from certain parts of the government. For example, I think Maine just banned data centers. Only about 20% of people are optimistic about AI, which creates the potential for negative regulation. However, I do think...
1:01:13-1:03:40
Alex Sacerdote: Regulation is a factor, but I think the genie is already out of the bottle. Another risk is the possibility that AI improvements might start to slow down. Even if the models stopped improving today, there is still a massive amount of AI adoption yet to happen.
Jensen Huang said this years ago when he was just talking about graphics chips: “If ‘good enough’ is actually good enough, I won’t have a business.” Every year, he made the graphics slightly better, and people always wanted the best. In AI, if a company like Anthropic or OpenAI hits a wall and stops improving, the open-source models will catch up. At that point, it could become a race to the bottom, which probably wouldn’t be great for those specific stocks. However, it could still be good for the chip companies. The chip companies don’t really care who is winning the token war.
Patrick O’Shaughnessy: Right, they don’t care who wins.
Alex Sacerdote: Exactly. That’s another positive for them; they’ll benefit even if open source takes off. In fact, Jensen kept mentioning open source at the last GTC—he clearly wants it to thrive.
Another risk is if one or two of the major players falter, lose their position, and can no longer compete. That could result in a lot of excess compute that they no longer need. However, if AI is as big as we think it is, someone else will quickly swoop in and suck up that capacity.
Alex Sacerdote: If one player pulls back, someone else will step in to fill that void. We saw that recently when Oracle canceled a major deal and Meta immediately stepped in. But let’s just say Meta decided not to be involved with AI—if they felt they couldn’t keep up and it was a waste of resources. We monitor that very carefully, but in general, we see more and more companies truly going after this. Even Microsoft is trying to build their own chips. I think those are some of the key risks we keep an eye on.
Patrick O’Shaughnessy: It seems like you’ve done very little in the application layer of AI so far. Historically, applications ended up capturing most of the market cap, rather than the infrastructure. There wasn’t really a “model layer” in the past—I guess you could say it was the cloud providers or something similar.
Alex Sacerdote: Yeah.
Patrick O’Shaughnessy: Why focus so much on the bottom layers of Jensen Huang’s “five-layer cake” versus the application layer things that consumers are actually using?
Alex Sacerdote: Well, we do have some exposure. Part of the value of OpenAI is ChatGPT, which is a massive application. However, we believe the application layer always develops later. If you look at the first three or four years of the iPhone, for example...
1:03:41-1:06:05
Alex Sacerdote: It took about four years after the iPhone launched for applications to really take off, so perhaps we are just at the beginning of that cycle for AI. To date, however, we’ve found the application layer to be quite risky. It’s difficult to determine where the foundational model ends and the application begins. The big question is whether these applications can build a sufficient moat to fend off competitors and establish sustainable businesses.
Alex Sacerdote: We expected to see more progress from incumbents like Salesforce, and while they are starting to move, it may just be a matter of time. We haven’t truly seen a breakout in the enterprise world yet. While there are some very promising startup application companies out there, the ecosystem remains murky.
Alex Sacerdote: When we first started looking at this space, the chip ecosystem was clear. The foundational model ecosystem was initially unclear, but it has become much more defined for us now. However, the application layer is still uncertain and a bit dangerous to navigate. That said, great application companies will undoubtedly be built. For instance, we are closely watching Bret Taylor at Sierra. Bret was the co-CEO of Salesforce, he created Google Maps, and he was the CTO of Facebook. He is currently building a fantastic company.
Alex Sacerdote: Bret Taylor is building this fantastic company called Sierra. We aren’t involved, but that is where the rubber hits the road. Will he be able to turn this into a massive company? He’s doing quite well so far, but we’ll see. It’s a matter of timing as to when these things really start to come into their own and prove they are sustainable. Usually, that doesn’t happen in the first three or four years; it comes a little bit later.
Patrick O’Shaughnessy: At your office, you have this giant award wall for research. I can’t remember exactly what it’s called, but it’s for the best research project of the year given to an analyst. I think you might have self-awarded it back when you were starting out by yourself, but you now have a 20-year history of people putting their names on that wall for doing the best job on a research project.
I’m curious about the nature of that research and how it’s changing as a result of all this. Consider the person who is going to win the award this year—what kind of work does that require a human to do now? So much of the work that probably would have won you the award in, say, 2009 could likely be fully automated or completed in an hour with Claude or Code Interpreter today. How has that evolved?
1:06:07-1:07:22
Patrick O’Shaughnessy: Whether it’s using cloud-based code or other modern tools, how is the nature of research changing in real time? What does it take today to get your name on that Whale Rock award wall?
Alex Sacerdote: I’d like to say that our AI systems are so advanced that they’ve already sparked a massive change, but so far, that isn’t quite the case. While AI is helping us get up to speed and we use a handful of great applications, it isn’t supplanting the role of the analyst yet.
So much of what we do involves meeting with as many companies as humanly possible. We focus on developing deep relationships with the management teams we cover and talking extensively to their competitors. Our system is actually pulled straight from *Common Stocks and Uncommon Profits*, which Philip Fisher wrote back in the 1950s. It’s the “scuttlebutt” approach to growth investing: getting out there and talking to suppliers, customers, and competitors to identify the key characteristics of leading companies and build real conviction in them.
Now, if we’re looking at a new, complicated area like ABF substrates or PCBs, AI helps us get up to speed on the technical details quickly. However, it can’t pick stocks for you in any meaningful way. I will say that if you’re an analyst who excels at the “blocking and tackling”—and there is certainly a role for that—the job is evolving.
1:07:24-1:09:54
Alex Sacerdote: There is a role for automation, but you still need to have the insight on top. We are now using AI to write notes or review the quarter, and while those notes are much better, there still needs to be a high-quality paragraph at the top that provides the wisdom. What does this actually mean? How does this impact our thesis? What has changed? You can’t just be a reporter. AI can be a great reporter, but it can’t quite predict the future.
Take the work the team did on AppLovin two years ago. I think we have two of the best ad-tech guys around, and they convinced me to buy. I was already familiar with ad-tech; I actually started my career nearby in New York at an internet advertising startup after I left banking. I knew that internet advertising and ad-tech have historically been terrible industries. However, Michael and Sam really figured out the AppLovin story before anyone else. They followed it while it was still private, they knew all the competitors, and they understood all the intricacies and terminology of the field. Sam even went to the app advertising conference in Las Vegas, and we went to Cannes. We talked to scores and scores of people to build that conviction.
Alex Sacerdote: Our team put in a tremendous amount of work on the model and developed a great relationship with Adam Foroughi. He is truly one of the best managers out there. I just don’t see AI being able to replicate that kind of relationship building and deep analysis.
Patrick O’Shaughnessy: What role does talking to other investors outside of your firm play in your life?
Alex Sacerdote: One of the best parts of this job is the friendships I’ve built with so many smart investors. Philip Fisher actually said that part of his process was getting to know 10 or 15 like-minded people around the country to share ideas with. Many of them are great friends who have actually been guests on your podcast.
You develop these strong friendships and you trade ideas, but it’s crucial that it remains a two-way street. I call it “the tripod.” When I like a company, my analyst likes it, and then someone else I deeply respect also likes it, those three legs of the stool really help build my conviction.
1:09:54-1:11:21
Patrick O’Shaughnessy: What have you learned about shaping the products you offer your investors throughout the history of the firm? It’s no longer just one monolithic structure.
Patrick O’Shaughnessy: Whale Rock is no longer just one monolithic structure. There are now several ways an investor can give you money. How did you arrive at those different offerings? Furthermore, how would you turn that experience into advice for other investors who are trying to provide their LPs with the right set of options?
Alex Sacerdote: For the first 15 years, we were strictly a long-short fund. We wanted to stay focused, as losing that focus can be difficult. We grew that business until it reached the scale we wanted. We’ve been around for 20 years now, and about 10 years in, people started asking for a long-only product. Consequently, we launched the long-only fund in 2020. We are now six years into that, and it has actually become larger than the long-short fund; the bulk of our assets are now in those two products.
Around 2015, we formalized the idea that we might start investing in privates. We gave our investors the option to opt in or out, with the ability to allocate 15% or 25% to private companies. However, we didn’t actually “break the seal” on private investing until 2020. Then, in 2021, we offered a hybrid fund that could be up to 80% invested in privates. It follows a similar approach but is designed for those who want significantly more exposure to the private markets. And then, very recently...
1:11:25-1:13:13
Alex Sacerdote: We also recently launched the Whale Rock Mega-Cap Tech Fund. We believe there is a massive structural underweighting of the world’s largest tech companies. We realized that a significant portion of our performance over the years actually came from some of these giants, whether it was Apple, Amazon, or Tesla.
It’s difficult for many investors to overweight these companies to the degree necessary. Many of the largest pools of capital, such as endowments, have realized they’ve been massively underweight in mega-cap tech for years. This happens because they hold a lot of private equity and not as much in public markets. Furthermore, perhaps half of their public exposure is international.
Within their public equity bucket, there’s often a prevailing belief that there is no alpha to be found in large-cap stocks. Consequently, they underweight large caps and favor small- and mid-cap managers, assuming that’s where the stock-picking opportunities lie. Even in their hedge fund portfolios, even those with a long bias, they aren’t going to hold a 15% position in Nvidia or similar names. We’ve realized there is a huge gap here; people are worried about the size of these companies, but their scale is simply a product of the modern digital economy.
Alex Sacerdote: This is simply a product of the digital economy. In the tech sector, the leader usually grows larger, wins, and develops a very high market share quickly. These companies possess great competitive advantages and sell their products globally. Consequently, this leads to massive profit pools and enormous market caps, a trend that will continue into the future. Most endowments are essentially betting against this because they are completely underweight in the sector.
1:13:13-1:15:11
Alex Sacerdote: Finally, people started coming to us asking what they should do and which index they should look at. I’m on the investment committee board at Hamilton College, and they were trying to figure this out as well. We kept hearing the same concerns, so we eventually told one of our clients that we would manage this for them. There is a lot of alpha to be found here. Whether you call them the “Magnificent Seven” or “FANG,” the leadership group is always changing. In 2022, they all rallied together, but last year their performance was very divergent, and this year some are down.
To address this, we created the Whale Rock Mega Cap Tech Fund. We look at a universe of the top 30 market caps globally and then select the 12 or 13 best companies. I believe there is tremendous alpha to be found in the largest caps because if you think about it...
Alex Sacerdote: Think about it: with a small-cap stock, it only takes one person to figure out it’s good and move the price. But with a large-cap, it takes a hundred people—a hundred diversified portfolio managers—to realize that Google isn’t a loser, it’s a winner. Our goal is to figure that out before 95% of those generalist PMs do. So far, we’ve been able to do exactly that.
Patrick O’Shaughnessy: We like your odds in that pursuit.
Alex Sacerdote: Yeah, we like our odds too. There is definitely alpha to be found there. As an asset category, it’s fantastic because these companies, by definition, have wonderful moats. They might not always be the “super S-curve” companies, but sometimes they are. NVIDIA certainly is, and TSMC is heavily levered to it. SK Hynix is extremely levered to it, as is ASML. We’re only about four months into this new cycle, so it’s a great time to be involved.
Patrick O’Shaughnessy: The right way to think about it is that you’ve essentially built a research machine designed to understand the world through the lens of companies. You are constantly trying to improve that research machine and the way you express those findings through your investments.
1:15:13-1:16:30
Alex Sacerdote: If you want to understand Whale Rock, you have to investigate our research machine first and foremost. We call it the “Whale Rock Learning Machine.” It’s a group of ten highly experienced individuals. While Warren Buffett reads books, we read books and blogs too, but in the tech world, you also have to get out there and talk to people.
We conduct 2,500 to 3,000 face-to-face meetings with management teams every year. Munger and Buffett talk about compounding knowledge, and we’ve been compounding that knowledge for 20 years. While there have been some changes to the team, there is a great deal of consistency. Andrew and Michael have been with me for 19 and 18 years, respectively, and the average experience level on the team is about 10 years, even including some of our newer members.
That research engine supports all of our products, and it’s the same people handling both public and private investments. We aren’t going to scour the world and turn over every single rock, but when we see something that fits our system, we are able to act on it.
Patrick O’Shaughnessy: It’s been so much fun doing this with you. To wrap up, I always ask the same final question.
Patrick O’Shaughnessy: I always end with the same closing question for everyone: What is the kindest thing that anyone has ever done for you?
1:16:30-1:19:00
Alex Sacerdote: I have to say it was definitely my father. I was incredibly lucky. My father graduated from Cornell with a degree in electrical engineering before pivoting to Wall Street, where he had a fantastic career at Goldman Sachs. He ran corporate finance in the ‘80s and then served as the chairman of private equity in the ‘90s. He was whip-smart, but he also possessed such humility and was a true gentleman.
When I started Whale Rock, I reached out to friends and family, and he was my first call. He told me, “I’ve been at Goldman for 41 years. How about I come and join you? I’ll be the ‘gray hair,’ provide oversight, and serve as chairman. You do what you do—build the firm in Boston, build the team, and manage the money—and I’ll help raise capital.”
We were able to work together for six years until he passed away in 2011. I feel so fortunate to have had that time with him. Running a fund isn’t easy, but we never once raised our voices at each other. He was an amazing mentor to so many people. When he passed away, I received so many letters from people who...
Alex Sacerdote: I received so many letters from people saying, “Your father was such an influence on me. He was such a gentleman and a great mentor.” I feel incredibly lucky to have worked with him. If I could be half the person he was, I’d consider that a complete win.
Patrick O’Shaughnessy: How did he do that? What was his method? Why did so many people feel that way about him?
Alex Sacerdote: I’m not entirely sure. He was modest, incredibly smart, and very wise. He was also known as a great investor, which isn’t always the most common trait at investment banks. He served on their commitments committee and helped steer them away from a lot of difficult situations. He was very warm; people could go into his office with any problem—whether it was personal or professional—and he would handle it with grace. He just had a gentle way about him, and he also had a great sense of humor.
Patrick O’Shaughnessy: You were lucky. Alex, thank you so much for your time.
Alex Sacerdote: Thanks so much.
1:19:01-1:20:04
Patrick O’Shaughnessy: Big mistakes often leak through a thousand tiny decisions that nobody is watching. Ramp puts guardrails on spending before it even happens with real-time limits and automatic rules, meaning zero firefighting for you. Try it at ramp.com/invest.
Patrick O’Shaughnessy: As your business grows, Vanta scales with you by automating compliance and providing a single source of truth for security and risk. Learn more at vanta.com/invest.
Patrick O’Shaughnessy: Ridgeline is redefining asset management technology as a true partner, not just a software vendor. They have helped firms scale 5x, enabling faster growth, smarter operations, and a competitive edge. Visit ridgelineapps.com to see what they can unlock for your firm.
Patrick O’Shaughnessy: Every investment firm is unique, and generic AI doesn’t understand your specific process. Rogo does. It is an AI platform built specifically for Wall Street that connects to your data, understands your workflow, and produces real outputs. Check them out at rogo.ai/invest.
Patrick O’Shaughnessy: The best AI and software companies—from OpenAI to Cursor to Perplexity—use WorkOS to become enterprise-ready overnight rather than in months. Visit workos.com to skip the unglamorous infrastructure work and focus on your product.
Made with: The Transcript Desk Chrome Extension