Inside Coatue's AI Public Market Update With CIO Jaimin Rangwalla | Edited Transcript

A professionally copyedited transcript of Molly O’Shea’s Sourcery interview with Coatue CIO Jaimin Rangwalla.

This is a professionally copyedited transcript of Molly O’Shea’s Sourcery interview with Jaimin Rangwalla, Coatue’s CIO of Public Investments, on the public-market implications of the AI buildout.

Chapter Timestamps

00:00 Why AI private companies now enter public-market scale before IPO

00:56 Inside Coatue HQ and the culture behind the investing process

02:48 Why the AI cycle keeps needing a new investor update

04:36 Mapping the AI stack by gigawatts, suppliers, buyers, and token users

06:02 Why supply tightness has lasted longer than prior semiconductor cycles

07:03 Jaimin’s path from semiconductors analyst to Coatue public-markets CIO

10:43 OpenAI, Anthropic, SpaceX, and private giants entering the top-company map

12:40 Market breadth, Mag 7 displacement, and where new winners come from

15:24 Where AI revenue actually comes from: enterprises, prosumers, and IT budgets

17:04 Tokens as the unit of AI work and the economics of agent usage

19:43 Why agents launching agents changes demand for compute

21:58 OpenClaw, harnesses, and phones as remote controls for agents

24:49 Why memory demand expands when each person runs hundreds of agents

27:12 Persistent memory, CPU demand, and the architecture shift beyond chatbots

30:21 Key chip players, Intel’s comeback case, and semis becoming mainstream

33:24 Nvidia mania, GTC, and how Coatue tracks data-center buildouts

35:21 Jobs lost, jobs created, and why the answer depends on productivity absorption

37:30 Sellers versus buyers of shortage in the AI supply chain

40:54 Optical breakouts, bottlenecks everywhere, and why one fix is not enough

44:48 Sentiment versus fundamentals in a market with strong earnings growth

47:10 Handling volatility by re-grounding in fundamentals and management quality

49:17 How Coatue hunts for new leaders at the edge of technical change

51:18 Trillion-dollar IPOs and what happens when OpenAI and Anthropic go public

52:48 Risks: efficiency breakthroughs, regulation, corrections, and Jevons paradox

55:00 Coatue’s growth story and why curiosity keeps compounding

Made with: The Transcript Desk Chrome Extension: https://thetranscriptdesk.two-lambda-ai.com/

Full video: Inside Coatue’s AI Public Market Update With CIO Jaimin Rangwalla:

The conversation uses Coatue’s spring AI public-market update as the backbone: private AI companies reaching public-market scale, “follow the gigawatts” as an investing frame, agents multiplying demand for semiconductors and power, and the difference between companies selling into shortages and companies buying through them.

Transcript

00:01-00:44

Jaimin Rangwalla: You now have private companies breaking into the world’s top 25 even before they go public. It’s an unprecedented time. The biggest of the MAG 7 back in 2012 was Meta—Facebook’s IPO was the largest ever at that time. Everybody wanted in on the most hyped IPO in history, and it was valued around $100 billion. But today, OpenAI’s most recent funding round was over $800 billion, Anthropic’s last round was in the high $300 billions, and SpaceX was valued at $1.25 trillion. These companies are hitting $25 billion in annualized revenue at a third or half the speed of hyperscalers and the big MAG 7. They’re adding around $10 billion a month or about $2.5 billion a week—that’s more than many SaaS companies generate in a whole year.

00:58-02:00

Molly O’Shea: Jaimin, welcome to Sourcery!

Jaimin Rangwalla: Thanks for having me.

Molly O’Shea: So, we’re in the beautiful new Coatue office. How does it feel?

Jaimin Rangwalla: You know, it feels like the first day of school again. We all walked in Monday morning, not exactly sure what to expect since the team had been setting everything up while we were still working downstairs. It’s a much more open, bullpen-style setup where everyone—AI folks, data scientists, analysts—are sitting right next to each other. It really pushes deep insights and collaboration. It’s exciting—much better than having to shout across the room to ask, “What does this mean? What’s going on here?” It actually reminds me a bit of my investment banking days.

Molly O’Shea: I heard Philippe has a desk right in the middle of the bullpen. Is that true?

Jaimin Rangwalla: He does. He showed up the first day, put his jacket on his desk, and said, “Why am I so far from the center of the action?” So we’re probably going to move his desk a little closer to the middle.

Molly O’Shea: That’s great! I was also chatting with some folks, and they mentioned you have a secret soccer group that meets at the pier at 6:30 a.m. twice a week.

Jaimin Rangwalla: It’s not really secret—it’s more selective! We’ve been playing soccer for about eight years now. It started as two-on-two in our gym, and now we’re playing eight or nine against eight or nine. We keep score, have weekly rankings, track goals, and the ranks go up and down based on performance. It adds a nice bit of competitive spirit to the culture, which we don’t necessarily need, but it’s a fun extra boost.

Molly O’Shea: That sounds like a blast!

02:50-04:14

Molly O’Shea: So, today we’re here to talk about your Spring Coatue investor update. You just finished it, right? There’s a ton of data in there. I’ve talked to Michael Barton, Thomas Laffont—they’ve walked through different tech cycles like internet, mobile, cloud. Now we’re in the AI era. What’s happening? It feels like every six weeks we need a new update because things move so fast. What are you seeing?

Jaimin Rangwalla: You nailed it—the reality is AI is massive, and everyone’s making grand statements about how huge this is. We’ve tried to size the total addressable market a bit in our slides, but the most exciting part to me is how fast innovation is happening. You can see it by how quickly companies are reaching $10 billion, $30 billion, $50 billion in ARR. For example, OpenAI hit almost a billion users faster than anything in history. All the growth curves—user adoption, enterprise usage, revenue—are so much steeper and faster.

We always focus on the rate of change—that’s how you really describe technology adoption and how big something can get. We often see fast growth then a plateau, like apps hitting 10 million users and then flattening. But here, these metrics keep climbing at hyper rates. Sometimes we even see what looks like a pause, then another big growth spurt. That tells us one thing: We’re still early in the adoption curve, and the market is far bigger than even the biggest optimists imagine.

04:36-06:02

Molly O’Shea: As the CIO of public investments, how do you map out all the different categories in this AI era?

Jaimin Rangwalla: We’re constantly rethinking how we cover stocks, themes, and subsectors inside this big theme. Two or three years ago, when we first invested in Nvidia, we mainly focused on GPUs. But betting on Nvidia alone wasn’t enough. When you think about selling billions of GPUs, you start asking, “What infrastructure supports all this?”

Now, we’ve shifted even further. We talk about following the gigawatts. Gigawatts are basically the atomic unit of AI growth—they represent the real power demand driving this expansion. It’s also one of the biggest supply shortages right now. So, we’re mapping out who supplies gigawatts, who’s helping reduce lead times for them, who buys them, and who uses the tokens produced by those gigawatts.

We’re slicing this differently than before. Historically, you’d think in terms of sectors—semiconductors, internet, aerospace and defense. Now, it’s more about the AI supply chain slices and assigning people to cover those slices deeply.

06:04-07:02

Molly O’Shea: What’s been the biggest surprise for you so far?

Jaimin Rangwalla: Honestly, I’m surprised by how tight the supply has stayed for so long. Historically, memory—the sector I know well—has had many periods of tightness. But this time, the shortage seems to be getting worse week by week. You’re now hearing about supply agreements locking in commitments through 2029 or even 2030. I’ve never seen that kind of tightness extend this far out.

That tells me that some of the largest companies in the world—the trillion-dollar ones—are the buyers, and they’re not foolishly investing hundreds of billions growing at high rates into something they think will just peak and then fade. Their behavior shows confidence in the long-term strength of demand. The persistence and length of this tightness has surprised me the most.

07:04-07:39

Molly O’Shea: You started your career in semiconductors, right?

Jaimin Rangwalla: Yeah, I started in 2007, right when the iPhone came out.

Molly O’Shea: Wow, that’s incredible.

Jaimin Rangwalla: I kind of hate myself a little for that! [laughs] But it’s pretty cool because here we are in 2024, and so much of the innovation we see now ties back to that moment. I remember when I first started, we sent four IT staff to wait in line for the iPhone. We all got it, and we were like, “What is this?” Back then, we were really entrenched in the BlackBerry world. That was my first real taste that innovation can happen unexpectedly—then you see the power of what follows.

07:41-10:43

Jaimin Rangwalla: The power of what happened from those first million iPhones to what they’re doing now is incredible. It was honestly the best lesson I could have ever gotten.

Molly O’Shea: I don’t think many people know much about you, so I’d love to hear a bit more about your background and how you became CIO.

Jaimin Rangwalla: Sure. I grew up just outside of Philadelphia. I went to NYU for undergrad.

Molly O’Shea: So did I!

Jaimin Rangwalla: Oh, really? Go Violets! I might have graduated a little earlier than you though.

Laughter.

After NYU, I spent two years in investment banking at Merrill Lynch—back when Merrill Lynch was still independent. Then I started thinking about my next steps. A lot of my peers were planning to go into private equity, attend business school, and then figure out their path from there.

But I wasn’t particularly excited about business school, so I began looking for other options. Someone suggested public markets investing. It’s different from private equity—less about hand-holding consultants and more action-oriented. You really only need to find a couple of big ideas a year to be successful.

Molly O’Shea: Is that true?

Jaimin Rangwalla: At an early stage, yes. If you can find one strong, big idea and convince your firm to back it, that alone can change a fund’s entire year—and even an analyst’s career.

Back in the summer of 2007, I met with five firms. It was a market peak at the time, with a strong run. Four of those firms didn’t offer me a job, but Philippe did. So my decision was almost made for me—since I knew I wanted to do public markets investing. We managed about $700 million at the time, a very different setup, occupying half a floor on 56th Street.

The first big lesson I learned there was that you never know what kind of market you’re stepping into. We all remember what happened in 2008, right? The other four firms that passed on me ended up going out of business during that crisis. There’s some luck involved in life, but there’s also what you make of your opportunities.

Even though the other firms didn’t hire me, Philippe did—and they taught me a valuable lesson: what goes up sometimes comes down, at least temporarily, especially in technology. But macro factors do matter. Learning that early in my career was priceless.

I started covering semiconductors, then expanded to other sectors, gradually taking on more responsibility. About two years ago, I got the chance to run all our public investments alongside Philippe.

Every day at Coatue feels like sprinting a marathon—you have to sprint all day with no time to rest, but still keep your eyes on the long term. It’s been an amazing experience.

Molly O’Shea: Wow. Congratulations!

Jaimin Rangwalla: Thanks—and thanks for taking the time despite the nonstop pace.

Laughter.

10:43-15:18

Molly O’Shea: I know it sounds repetitive talking about the size of the market today, but it’s truly astronomical. Could you compare the size of the Mag 7 at their IPOs to the scale of today’s AI leaders?

Jaimin Rangwalla: Yeah, it’s wild. The biggest of the Mag 7 was Meta, which IPO’d in 2012 at about $100 billion. The rest of that group went public in the ’90s, typically between $10 and $50 billion, which was huge for that time.

Today, OpenAI’s most recent funding round valued them at over $800 billion. SpaceX, after its transaction with xAI, hit $1.25 trillion. Anthropic’s last round was in the high $300 billions, and there’s talk their next round could be even higher.

Having private companies valued among the top 25 globally before even going public—that’s never happened before. That’s what makes AI so exciting. These private firms are capturing massive value and growth while staying private, something unprecedented.

We’ve had to reshuffle parts of our team around this. For example, we brought in Frank, who’s our Claude expert.

He joined early this year and—in official terms—might be called our Chief AI Officer, but I think of him as our AI mad scientist. He’s trying out everything new and cutting-edge, figuring out how to implement these technologies and how we can utilize 20 years of data to give us an edge against the competition and help us succeed.

Another interesting chart I want to mention is from Bloomberg. Focus on the bold red line—it shows the Nasdaq’s performance since its peak on October 31st. It’s basically ended up flat, though it dipped 10% at one point and is now up 10%.

Then, when we look at the Mag 7, five companies plotted there, you notice that outside of Alphabet, most have underperformed the Nasdaq by quite a bit—and three are actually down year-to-date. That tells me the pecking order within the Mag 7 is rapidly shifting.

One wild metric was the annualized revenue growth of these new players like OpenAI and Anthropic. They’re hitting $25 billion in annualized recurring revenue at a third or half the time it took hyperscalers and the big Mag 7 firms to get there.

How do I interpret that? It goes back to how huge the market is becoming. These two companies alone are north of $50 to $60 billion in ARR, growing incredibly fast.

Just looking at Anthropic’s public disclosures: They were at $9 billion ARR at the end of December, and by now they’re at $30 billion. That means they added $20 billion in just a few months—that’s about $2.5 billion per week.

Most SaaS companies don’t even have $2.5 billion in ARR annually—they add that in just one week. The opportunity is enormous. If you combine those two companies, they’re bigger than ServiceNow, Salesforce, or any major SaaS platform you can think of—all in just a few years compared to the 15 to 20 years those other companies took.

So, the excitement around AI comes not only from the scale but also these extraordinary growth rates. It makes us confident the market size is far larger than anyone has imagined.

Even just thinking about current use cases, breakthroughs keep coming, but we’re nowhere near realizing the true potential.

15:17-15:47

Jaimin Rangwalla: We’re definitely nowhere near where I see this going three or five years down the road. So, I still believe we’re very early in this whole adoption phase.

15:47-16:12

Molly O’Shea: For those who don’t quite understand where that revenue is coming from, could you break it down? I know some of it’s promised revenue or future contracts, but how do you actually think about that kind of revenue?

16:12-17:05

Jaimin Rangwalla: Sure. So, when you think about revenue for a company like Anthropic, it comes from a few pools: enterprises, individuals, prosumers—basically people who want to use these tools to boost their efficiency. It’s not a simple 1-for-1 equation where spending $100 on Anthropic means you reduce labor costs by exactly $100. It’s about productivity gains. Even the top coders and developers say their productivity has easily doubled or tripled with these tools.

In tech companies, the limitation isn’t the number of projects—they have plenty—but the number of people who can get those projects done. AI lets those people work faster, handle more projects, and be more productive overall.

AI also makes you rethink your overall IT budget. Traditionally, a lot goes to hardware, consultants managing systems, and software. Now companies are asking themselves if they need all that in the same amounts, or if some can be reduced or reallocated. The budget impact comes from both the natural ROI of increased productivity and from offsetting existing expenses.

17:05-17:27

Jaimin Rangwalla: Also, it’s a lot about tokens—token use and the token economy.

17:27-18:13

Molly O’Shea: Tokens have been a hot topic recently, especially access to models like Mythos and how tokens work economically. Could you explain what tokens are and how the token economy functions?

Jaimin Rangwalla: Absolutely. A token is basically the unit of thought for any AI model. Whenever you ask an AI a question and it answers, what you see is made up of tokens. It’s like a unit of intelligence in AI—similar to how humans think about making decisions or taking actions. If you imagine an AI agent, each of those thought processes is a token.

The token economy is really about the explosion in the number of AI agents and the decisions they make. Unlike human intelligence, which is hard to quantify, AI intelligence can be measured by how many tokens it generates. This new way of quantifying AI activity is pushing us to rethink how we frame business opportunities for many companies.

18:13-19:13

Sponsor Read: Sourcery is brought to you by Brex, the financial stack trusted by more than 30,000 companies, including one in three venture-backed startups in the U.S. Brex is built to help founders avoid running out of cash by combining checking, treasury, FDIC protection, global money movement, and liquidity tools in one account. Companies including Scale AI, DoorDash, ServiceTitan, Hims, Anthropic, Flexport, Robinhood, and Plaid trust and use Brex. Start today at brex.com/sourcery.

19:19-19:41

Sponsor Read: Turing is training the next generation of AI with tasks that require real expertise and real-world judgment. Companies like Nvidia, Anthropic, Salesforce, and Gemini partner with Turing to build realistic reinforcement-learning environments and data systems based on real operational traces—the kind of infrastructure frontier labs need to train superintelligence.

19:48-19:53

Molly O’Shea: You’ve mentioned agents as a huge unlock in AI. What makes agents so different or transformative?

19:53-21:51

Jaimin Rangwalla: Originally, with ChatGPT, you’d ask a question, it would take some time to think depending on complexity, then it’d give you an answer—or maybe intermediate steps, checking in with you, like, “Am I on the right track?” So, humans were involved in the loop a lot.

The big breakthrough came with models like OpenAI’s GPT-4.5, released late last year. Now, an agent can spawn other agents. This multiplies the depth and quality of work. You can give an agent a project, walk away, and come back to find much of the work done, sometimes almost entirely. What’s even cooler is you can tell the agent to spawn as many sub-agents as it wants, and even create agents to check on those agents’ work.

It’s really collaborative and iterative, with humans almost completely out of the loop—which used to be the bottleneck. Before, one chatbot interacted with one human and kept stopping to check progress. Now, the agent system scales the work exponentially without constant human intervention.

For example, I use Claude too and don’t just give it one prompt or problem. I launch multiple agents concurrently across different tasks, and each set can spawn even more agents underneath them. It’s an exponential growth in how many agents are working and how much work is being done.

21:52-21:57

Molly O’Shea: Yeah, we definitely saw that with the OpenClaw slide in your update – that was wild. What kind of growth rate did OpenClaw show?

Jaimin Rangwalla: The OpenClaw phenomenon is really fascinating and maybe underappreciated. A harness is basically the interface or way a person interacts with an AI model. Before, you’d interact through ChatGPT or a code terminal for Claude. But with harnesses, you can control AI remotely on different machines or virtual environments and have it perform actions on your behalf.

For instance, someone might message their home OpenClaw on WhatsApp saying, “Book me a reservation this Friday at this restaurant, send the calendar invite to Jaimin, and also make sure that —”

22:49-23:29

Jaimin Rangwalla: So, imagine you say, ‘Jaimin, make sure we have a car service set up to take us there and back,’ and then send him the confirmation email, make a calendar appointment, and handle all those details.’ The system takes care of all that in the background. It just comes back with a message: ‘Okay, I did this and here’s the confirmation.’ It’s almost like how the way you interact has shifted with the claw setup. Now you can use a terminal to interact, or have it in your pocket like your phone. We actually have a slide in our deck that describes the phone as a remote control—just a remote for your agents. Which agents do you have working on what? You can launch multiple agents at once.

What you’re seeing with OpenClaw is an incredible growth trajectory: more people programming for it, trying to incorporate it everywhere—even in China, where all the big internet companies have their own claw concepts. ByteDance has one, Tencent has one, Alibaba has one. It’s amazing how quickly this has taken off, and honestly, I don’t think most people fully appreciate it yet. To me, this is just an intermediate stage. I don’t think it’s going to be like today, where you have to get home and plug in all these connections to your claw because you still need to give it access. But there’s a major opportunity for someone to say, ‘Here’s your phone, here’s your super app.’ Now you can connect everything you have and just ask that app questions—it becomes your super app.”

23:30-24:07

Jaimin Rangwalla: What you’re seeing with OpenClaw is an incredible growth trajectory: more people programming for it, trying to incorporate it everywhere, even in China, where the big internet companies have their own claw concepts. ByteDance has one, Tencent has one, Alibaba has one. It is amazing how quickly this has taken off, and I don’t think people fully appreciate it yet. To me, this is an intermediate state. Eventually someone can say, “Here’s your phone, here’s the app, this is your super app,” and it connects to everything you have.

24:07-24:12

Molly O’Shea: Did you see the clip of Nat Friedman at Stripe Sessions?

24:12-24:44

Jaimin Rangwalla: I did not.

24:12-24:44

Molly O’Shea: He had OpenClaw deployed on his phone and maybe his laptop too. It noticed he was dehydrated and told him to grab water. He went to the kitchen, drank it, and OpenClaw took a picture and said, “Nice job.” Then while he was driving, it told him he needed magnesium and rerouted the car to Whole Foods.

24:44-24:49

Jaimin Rangwalla: That is pretty crazy. We are just at the beginning of what is going to be possible.

24:49-24:56

Molly O’Shea: So you mentioned memory per person is about to explode. Can you walk through the total addressable market for that?

24:56-26:28

Jaimin Rangwalla: Sure. Historically, memory was tied to smartphones and computers. Everyone has a phone and a computer, so naturally, over time, as more people got devices and those devices’ memory increased—like eight gigs of RAM on a computer, 8 to 16 gigs on a phone—memory capacity grew steadily.

But now, think about this: in a year or two, you or I will have claws running hundreds of agents all the time. Similar things will happen at work—hundreds or even thousands of agents running continuously. Each of those agents requires CPUs, GPUs, and memory. It’s not persistent memory since they aren’t running 24/7; some run for shorter times than others but still, across the board, your semiconductor and tech footprint as an individual is about to multiply tenfold.

You used to have two or three devices—maybe a phone, a computer, an iPad. Now that will grow exponentially to thousands of virtual devices, which are really agents running on top. Each of those agents has semiconductor content. It’s hard to fully grasp, but the future is a massive increase in individual semiconductor and power usage footprints.

For instance, on a recent call, we had Boris Cherny, head of Claude Code at Anthropic and creator of Claude Code, who explained that during the day he runs some number of agents, but when he goes home, he might run a thousand agents overnight to finish tasks before morning. Just imagine if all of us do this: running hundreds to thousands of agents consistently, handling simple tasks or complex ones.

The impact is huge—semiconductor content, power consumption—all going up significantly because of these behavioral changes.

27:12-27:17

Molly O’Shea: Because of those behavioral changes, how is that affecting the architecture of AI?

27:17-28:05

Jaimin Rangwalla: We’re moving beyond chatbots with amnesia—where you feed it info, it does a great job analyzing it, but when you come back the next day, it doesn’t remember anything. It’s like the movie *50 First Dates* where the main character has to re-explain everything every day.

What’s changing now is the introduction of persistent memory using the next generation of accelerators, so agents won’t forget anymore. Because agents launch other agents and handle increasingly complex and simple tasks, they need offloading from CPUs too. Historically, CPUs did serial processing—one instruction after another—while GPUs handled the heavy math and parallel computing.

Now, the ratio is shifting: it used to be one CPU to eight or even sixteen GPUs doing the computation, with GPUs doing almost all the math. But today, the ratio is moving toward one CPU to four GPUs, and some believe it might flip entirely to one GPU to four or even eight CPUs. This depends on how many tasks are serial—simple, step-by-step work—versus massively parallel work.

Think about 7 billion people on the planet, and if each person effectively needs 1,000 virtual agents, that’s 7 trillion virtual entities—massively multiplying the compute demand and semiconductor needs. That’s the kind of scale we’re facing.”

30:01-31:40

Jaimin Rangwalla: Right now, all these digital agents we’re talking about will need processing power—serial processing, parallel processing, memory—the works, just like a human brain does. So, the analogy is almost like a population boom, but digital. If we multiply the population by a thousand, we’re talking about a digital population that needs all the resources to support it.

When it comes to the main players in this tech stack on the CPU side, it’s Intel, AMD, and ARM. Amazon has also done a great job—they actually have one of the most efficient CPUs around, but they keep it mostly in-house. They acquired a private company that powers this, which has helped drive the success of AWS, especially as CPUs become more critical. That’s a strong tailwind for their business.

Intel, for a long time, was kind of the “forgotten” semiconductor company. They faced missteps—technical issues, product issues—and had a lot of leadership changes. But now, they’re led by an amazing CEO. Sometimes, the best investment ideas are the simple ones. For example, there’s this shift happening from a 4:1 ratio to a 1:4 ratio in chip architecture, which mathematically means a 16x increase in market size.

Add Elon Musk giving Intel his stamp of approval and pushing innovation faster, and you have a compelling thesis. Sure, Intel’s stock has made a big move this year, but if you look back over the last four years across semiconductor companies in the AI boom, many have gone up 5x, 7x, or 10x. Intel’s still a laggard beyond a 12-month timeframe, so there’s plenty of room for catch-up.

31:41-33:15

Molly O’Shea: That’s exciting to hear, especially since you started your career in semiconductors—though back then it was more focused on gaming. How do you feel now, looking back?

Jaimin Rangwalla: Oh, absolutely. The first seven or eight years of my career, I mostly worked on short positions in semiconductor stocks because the sector was seen as super cyclical. It was common to overbuild capacity at the top of the cycle and burn through huge cash flows. The sector was fragmented—companies made one product and sold to Apple, but then Apple would double or triple source, and those companies would drop 90% overnight. So, I spent most of my early career working on shorts. We didn’t own semiconductor stocks; we focused on internet and software companies instead.

Now, it’s amazing—my career has basically come full circle. I get to make portfolio decisions at a time when semiconductors are the most profitable sector and generating the most cash flow. I never thought I’d say that Samsung and Hynix together generate more cash flow than all the hyperscalers combined. It’s wild but totally justified given the times. It really shows how important it is to constantly evolve your thinking. If I was still stuck on the old idea that semiconductors are shorts, it would be the biggest mistake.

We were like the nerdy kid who suddenly became cool.

Molly O’Shea: [laughs] Jaimin Rangwalla: Exactly. Semiconductors’ CEOs used to be the ones no one wanted to meet at conferences. Everyone wanted to meet the Facebook and Google CEOs back in the 2010s. Now it’s flipped. Today, at a bank conference, you’ll find thousands of people eagerly waiting to hear Jensen Huang speak. It’s just unreal.

33:16-34:00

Molly O’Shea: GTC feels like a Taylor Swift concert these days.

Jaimin Rangwalla: It really is. It’s an event now. I’ve been going since 2016, and it used to be a small hall in San Jose with 100 people. Now it’s grown unbelievably. And Jensen does his whole two-hour keynote without a teleprompter—that’s incredible.

Molly O’Shea: Absolutely wild.

34:01-35:00

Molly O’Shea: Switching gears, can you talk about how data centers fit into all this? How do you monitor these build-outs and whether capacity is keeping pace?

Jaimin Rangwalla: Yeah, it’s a great question—and no, not in space! [laughs] What we really focus on is understanding whether bottlenecks in the supply chain are easing or getting tighter. The key for me is tracking the rate of change—is it moving toward tighter constraints or is there relief? And if there’s relief, what’s causing it? Is the easing enough to be healthy and support growth, or is it becoming a problem?

We’re very focused on data center companies, especially the new “neo-clouds” and hyperscalers. We keep tabs on where build-outs are happening, where power is available, where equipment and labor are accessible. Tracking all these factors in real time is crucial.

I was recently in Miami visiting Exowatt—a renewable energy company powering data centers. Their CEO, Hanan, explained how they’re bringing in new labor to support this growing industry. They manufacture and build out these facilities right here in the U.S. It’s an important point to remember that, while there’s concern about AI replacing jobs, this industry is also creating new jobs.

35:01-36:51

Jaimin Rangwalla: We’re seeing new layers of the economy form. When you look back at previous waves of technological innovation—from the automobile era to all seven major tech waves you can think of—there’s always some job loss, sure. But there’s also a huge wave of new job creation and new businesses springing up because of the core innovation.

Right now, people aren’t talking enough about new business creation, but we are at one of the highest levels of new business formation ever seen. Everyone’s saying, “Hey, I can use AI to rethink how a business operates or build something totally new.”

Sure, some focus on the negatives, but I see enormous opportunities for job creation. The speed of this disruption matters though. Sometimes, innovation happens faster than the new electricians, plumbers, or construction workers can be trained—these jobs take two or three years to prepare workers for. So, there may be a lag before the new workforce catches up.

36:52-37:18

Jaimin Rangwalla: There’s a bit of a time lag, but broadly speaking, we see a huge shortage of labor across many different types of skilled jobs. This usually means salaries for those roles will rise as companies try to attract more talent. So, in the coming years in an AI-driven world, those are the jobs that will be in high demand—maybe quite different from what dominated in the last 20 years.

37:18-37:29

Jaimin Rangwalla: In an interview with Thomas, he talked about the evolution of bank tellers—not really increasing the total number of jobs, but rather spreading them out differently. This is always a hot topic when discussing market dynamics.

37:29-38:02

Jaimin Rangwalla: When we talk about supply and demand in these markets, there’s this framework of “sellers of shortage” versus “buyers of shortage.” Let me break that down for you. Sellers of shortage are like semiconductor companies, power and memory infrastructure providers—basically, those with a fixed capacity but rising demand. This imbalance drives prices up and expands margins. Operating profits often rise multiple times faster than revenue growth because fixed costs don’t scale with price increases. We’ve seen earnings jump three, four, five, even seven times in just a year or two for companies that used to hover in a steady range for decades. These shortages—whether memory, hard drives, or other components critical to AI development—are really fueling earnings power, and the market is rewarding those players.

38:02-38:44

Jaimin Rangwalla: What the market isn’t rewarding as much, though, are the companies buying into these shortages—those investing heavily in CapEx. For example, look at Microsoft, Amazon, and Meta: their valuations have compressed in recent years largely because their capital spending has ramped up. If memory prices double, Microsoft has to spend twice as much just to maintain the same level of benefit. The benefit itself doesn’t improve; it’s a like-for-like price increase, not new technology breakthroughs driving prices higher. So, the market currently penalizes these buyers of shortage a little because their ROI is impacted by inflationary costs rather than genuine advances.

38:44-40:00

Jaimin Rangwalla: On the other hand, stocks in sectors like optical infrastructure, power, and labor are doing really well because those areas face shortages too. Labor costs are rising, and so are prices for memory, CPUs, GPUs. Even TSMC, one of our biggest holdings, is increasing prices because capacity is finite, and they need to ensure the right margins and returns to justify future investments. So, these sellers of shortage are winning.

40:00-40:38

Jaimin Rangwalla: That said, companies like Amazon and Google don’t fit neatly into just one camp. We actually like them a lot. They’re hybrids because, for example, Google sells TPUs and Amazon has openly discussed the value of Trainium chips. So they’re both buying into shortages and also can be sellers. Google, with its vertical integration, has been a standout stock. This hybrid model positions them uniquely.

40:38-41:16

Jaimin Rangwalla: So, in summary: the market today is divided between buyers and sellers of shortages. Buyers are getting punished because their near-term cash flows are impacted by high CapEx, while sellers benefit by capturing profits.

41:16-41:30

Molly O’Shea: Have there been any surprising breakouts in this market?

41:30-41:59

Jaimin Rangwalla: Yeah, optical has surprised me the most. For years, optical was seen as a sector with constant swings between oversupply and undersupply. Now, it’s becoming a crucial part of future infrastructure. We’ve been tracking memory, CPUs, GPUs, accelerators closely, but optical is really where I’m seeing unexpected strength.

41:59-42:20

Jaimin Rangwalla: As for bottlenecks, it’s not just one thing—there are bottlenecks everywhere. Power generation, power transmission, distribution—all are bottlenecks. NAND flash memory, DRAM, optical components, labor to build out data centers—the list goes on. I’ve never seen so many constraints happening simultaneously.

42:20-42:45

Jaimin Rangwalla: If it was just one bottleneck, you could pour money in and solve it, but here, even if memory bottlenecks get resolved, power or optical or labor bottlenecks will persist. The biggest challenge is that there are so many bottlenecks at once.

42:45-43:01

Jaimin Rangwalla: One of my favorite charts shows how the market rewards sellers of shortage right now but not the buyers. The market is focused on capturing short-term profits, and that’s understandable. But eventually, this dynamic will shift, and we have to be very aware and well prepared for when it does. I’ll speak more about how we plan for that later.

43:01-43:44

43:01-43:44

Sponsor Read: VCX by Fundrise is the public ticker for private tech, opening venture capital investing to everyone, no matter the size. Check out the portfolio at getvcx.com.

Announcer: And some of you may not have heard yet, but our sponsor Public just launched something called Generated Assets. It brings AI into investing in a way I honestly haven’t seen before. Here’s how it works: You type in an idea, like ‘AI-powered supply chain companies with positive free cash flow,’ or ‘defense tech companies growing revenue over 25% year-over-year.’

Public’s AI dispatches a swarm of agents that scan every single U.S. stock, evaluates them, and instantly builds a custom index based on your thesis. What really stands out is how clearly it explains...

43:46-44:14

Sponsor Read: What really stands out is how clearly it explains why each stock is included. Before you invest, you can even backtest your idea against the S&P 500, so you’re making decisions with real context—not just guessing. And beyond generated assets, Public lets you invest in stocks, bonds, options, crypto—all in one place. They’ll even give you an uncapped 1% match when you transfer your investments from another platform. If you want to build a portfolio that actually reflects your thesis, visit public.com/sourcery. Full disclosures are in the description.

44:16-44:32

Sponsor Read: Enterprise AI runs on Merge—the AI infrastructure platform for integrations, agent tooling, and model orchestration. This lets your teams ship products instead of plumbing. Mistral, Dropbox, and Drata already trust Merge in production. Start building at merge.dev.

44:34-44:43

Sponsor Read: Founders scale faster on Deel. You can set up payroll for any country in minutes, hire anywhere, get visas handled quickly, and get back to building. Visit deel.com/sourcery—that’s d-e-e-l.com/sourcery.

44:50-45:24

Molly O’Shea: One interesting thing right now is the headlines around AI and the economy are super negative. But public markets are performing better than I think they have since April 2020. So if sentiment is so negative online and in the news, but markets are actually better than April 2020, how do you explain that? What’s really going on?

45:27-46:02

Jaimin Rangwalla: Yeah, that’s a great question. We actually did some analysis a few calls ago on sentiment versus actual market performance. Turns out it’s basically a coin flip. Negative sentiment doesn’t reliably predict market downturns or upswings. People latch onto different stories in the news or social media, but the core underlying fact is that companies are beating earnings. The S&P 500’s earnings growth through 2026 is projected at about 15%, accelerating to 18%. During times of strong economic growth like this, stock markets tend to do well because fundamentals are driving it. The economy and consumers are strong. Even with some near-term issues like higher gas prices, consumers received strong tax refunds, so there’s no panic in spending or enterprise investment. Earnings are really solid.

46:04-46:59

Jaimin Rangwalla: So at the end of the day, fundamentals matter more than sentiment. If you look at earnings revisions for some of these companies facing supply constraints, revisions keep going up. Every quarter, they’re beating earnings by 30-40%. I remember seeing something like SanDisk or Micron—Wall Street would estimate one dollar in earnings per share, and they’d report $2.50. They just keep beating the numbers by a wide margin. The multiples for the market haven’t really increased much either—they’re staying consistent because growth is so strong. So even if the market only goes up 7 or 8% this year, earnings growth is double that. If the market stays flat or dips slightly, the implied multiples actually decline by year-end. Bottom line: we’re in a period of very high growth and that’s what really matters.

47:00-47:25

Molly O’Shea: I think this circles back to the start of our conversation. You’ve been at Coatue through many cycles, especially ones like this. You talked a bit with Thomas on stage about dealing with hot-flash volatility, mixed guidance, and shifting sentiment. How do you stay focused as a firm amidst all this noise?

47:29-48:16

Jaimin Rangwalla: Yeah, Philippe always says we should do these presentations more often because it helps us ground our thinking. It forces us to remember what really matters—the five or seven key things in any moment. It is hard, though. The volatility lately has been much higher than I’ve seen before. Usually, if a bad event happens, the market goes down. You don’t always know how much, but it’s understandable. What’s tricky now is some days our stocks drop by 5 or even 10% for no apparent reason. That’s the tough part. Sometimes it’s because of a cloud release or something random, but you’re always wondering if someone knows something you don’t, or if the company is slipping behind. It’s a big risk if you aren’t tracking everything carefully.

48:19-49:09

Jaimin Rangwalla: So, we try to always re-ground ourselves in fundamentals and the big picture, the overarching theme. We make sure the stocks we own are high-quality businesses, with strong management teams we can understand deeply. That helps us handle the volatility. We’ve also got tools like hedges, puts, and shorts. But ultimately what we’ve concluded is the fundamentals are strong, the market is resilient, and AI is accelerating faster than I would have guessed, even compared to a year or two ago. People say AI’s been talked about for three years now, but it’s actually speeding up.

49:10-49:32

Molly O’Shea: When you’re hunting for new market leaders, what do you look for? How do you evaluate companies?

49:32-50:33

Jaimin Rangwalla: What’s worked well for us is focusing on the leading edge of change—identifying companies driving that change. For example, Nvidia pretty much sets the standard for what the accelerator market will look like in two years, in terms of capabilities and performance. Broadcom and TSMC are also important. We watch what they’re changing because they’re all trying to solve their own bottlenecks, reduce costs where technology hasn’t advanced, and optimize profitability. Here’s an example: today, accelerators mostly connect using copper wires to transmit data, but maybe in two or three years, we’ll see less copper and more optical connections—which can transmit data faster over longer distances, though there are limitations. Every new version of Nvidia’s accelerators adds or removes features, and we track those changes closely. We ask: how many players are there in this industry? Is it concentrated or fragmented? That kind of insight helps us understand the competitive landscape.

50:36-51:07

Jaimin Rangwalla: So, is this an industry dominated by just a few consolidated players who have been serving a legacy market, and now suddenly they’re facing a huge unexpected surge in demand they can’t keep up with? If that’s the case, maybe that’s the sector we need to follow closely. We’re really trying to stay ahead of where the change is happening. We have significant investments in OpenAI and Anthropic, so we’re constantly in touch with them—understanding how they’re optimizing their models, the technology they’re using, and other developments. We try to do deep research to stay at the forefront.

51:07-51:19

Molly O’Shea: This is wild because whenever someone mentions OpenAI and Anthropic, I just automatically assume they’re already public companies by now, but they’re not.

51:19-52:48

Jaimin Rangwalla: Right. So how are you thinking about these trillion-dollar IPOs? I think it’s fantastic for the market. Tech leadership changes regularly—I mean, studies show that roughly every five years, about a quarter of the top 25 tech companies are swapped out for new ones. If anything, this just reflects the natural evolution and dynamism in tech. Companies like SpaceX, OpenAI, and Anthropic have already vaulted themselves into the top 30 largest companies in the world by valuation—even while still private. So, when they go public, it’s really just opening the door for retail investors who previously didn’t have access. It won’t force people to sell other big names like the Mag 7 stocks because the total market cap of the S&P is massive—tens of trillions—so adding a few trillion-dollar companies is not a huge displacement.

It’s healthy to see new big names emerge. If tech just recycled the same companies over and over, it wouldn’t reflect the true innovation happening. Fresh players keep the ecosystem vibrant.

52:48-53:29

Molly O’Shea: As Coatue’s public sector CIO, what are the biggest risks you’re watching right now?

Jaimin Rangwalla: The biggest risk is that some breakthrough technology suddenly emerges that changes where current shortages exist—especially around chip supply, memory, or power demand. For example, last year, there was a “deep seek” moment where efficiency improved dramatically. If something like that happens again—say a model that requires far less power or fewer semiconductors—that could disrupt the market. But in the long term, that’s probably positive for AI adoption because it would drastically lower costs. There’s actually a Jevons paradox effect here—the cheaper AI gets, the more creative ways people will find to use it, accelerating demand overall.

Another risk is regulation—something that comes out of left field and could seriously slow growth. But AI is becoming a national security topic, and I think regulators worldwide will want to encourage a cooperative AI environment. So while regulation is a watchpoint, I expect it won’t kill the secular growth trend.

53:29-54:59

Jaimin Rangwalla: We did another study showing that roughly every 10 years, the markets go through a significant crisis—say a 30% drop—and every year you can expect about a 10% correction. Last year gave us a 30% drop related to tariffs. So we’re entering a period where risks will always be present, but the underlying secular trend for AI and tech growth feels solid.

54:59-56:12

Molly O’Shea: Other than being on camera more, what are you most looking forward to this year?

Jaimin Rangwalla: We just moved into a new office, which is exciting. Honestly, Philippe Laffont is an incredible investor and also an amazing CEO. I joined when we were just 12 people managing about $700 million. Today, we’re over 80 billion in assets with over 200 people across multiple offices. Reflecting on that twenty-year trajectory is incredible. Philippe probably still gets to the office before I do and leaves after me—he works harder than anyone, never rests. The journey continues. We’ve talked about one inflection leading to another, and I think as an organization we’re at another big inflection point—a fifth, maybe. We’re on the brink of breaking out.

56:12-56:33

Molly O’Shea: What do you think his secret is?

Jaimin Rangwalla: Lately, he’s been playing a lot of padel tennis. I think that new hobby keeps him energetic and young. He’s endlessly curious and always thinking about what’s changing and how to capitalize on it. That mindset—constantly asking, “What’s changing? How do we catch it?”—is crucial. It’s tough because a lot of people get comfortable with the known, but in tech you have to be relentlessly forward-looking.

56:33-56:56

Molly O’Shea: Amazing place to end. Thank you so much for this great conversation.

Jaimin Rangwalla: Thank you very much.

56:56-57:00

Molly O’Shea: Hey, it’s Molly. If you enjoyed this interview, check out our newsletter at sourcery.vc, where we deliver a weekly roundup of top deals and tech headlines, plus deeper dives into our podcast interviews. Subscribe to Sorcery today, and don’t forget to follow the podcast on YouTube, Spotify, Apple, or wherever you listen. Links are in the description.

Made with The Transcript Desk Chrome Extension: https://thetranscriptdesk.two-lambda-ai.com/.