Deconstructing the investment philosophy of Gavin Baker, an early investor in Nvidia: Going long on AI infrastructure bottlenecks and shorting overall market risks.

Compiled & Organized by: Deep Tide TechFlow

Hosts: Ejaaz Ahamadeen (EJ), Josh Kale (Josh)

Original title: What The Best AI Investors Are Buying Right Now

Podcast source: Limitless Podcast

Broadcast Date: May 28, 2026

Editor's Note

This podcast episode primarily discusses the investment philosophy of Gavin Baker, founder of Atreides Management and a long-time bettor on Nvidia and Cerebras. His core judgment is that AI is not a bubble, but rather a supercycle of infrastructure driven by electricity, wafers, and computing power; the real excess returns are not in large models or chatbots, but in the "shovel sellers"—GPU connectivity, memory, inference chips, advanced manufacturing processes, and power supply.

While using QQQ puts to hedge against a market pullback, Gavin Baker is concentrating his bets on AI physical bottleneck assets such as Astera Labs, Unity, Micron, Nvidia, Cerebras, and Positron. He shifts the "AI bubble" debate from an emotional level to supply and demand constraints, arguing that as long as TSMC, ASML, high-bandwidth memory, and the power grid do not experience rapid oversupply, AI capital expenditure may not be a repeat of the 2000 dot-com bubble.

Essential Quotes

AI Bubble or Supercycle?

"AI is not in a bubble; on the contrary, it is in a supercycle."
"The biggest rewards are not in SaaS, nor in chatbots like OpenAI or Anthropic, but in electricity, computing power, and silicon wafer manufacturing."
"This is not an internet bubble because the buyers are mainly the world's smartest and cash-rich companies, and they are not buying computing power with debt leverage."
"If the market cannot be oversupplied, it is difficult for it to burst suddenly like a traditional bubble."

The real bottlenecks: electricity, wafers, and tokens.

“Gavin’s theory is simple: just look at the bottlenecks in the AI infrastructure layer. Whoever can improve performance per watt and reduce token costs will have value.”
"AI labs are increasingly concerned with one thing: how many tokens can be generated per watt of electricity."
"Electricity and wafers are two brick walls, and also two key constraints that limit the rapid acceleration of AI."

From pre-training to inference and post-training

“A model’s pre-training is not a guarantee that it will be a genius forever; it still needs to absorb new information in the post-training stage.”
"Inference inherently requires a lot of computation, which is why inference chips and inference infrastructure will be the focus of the next stage."
"The cost or revenue opportunities from inference alone could be 5 to 10 times the investment in pre-training computing power."

Vertical mini-models, edge models, and sovereign infrastructure

“In the future, you may not need to interact with Claude every day; what you may really need is a personalized AI agent trained based on your own data.”
"The speed of infrastructure deployment itself is a moat; the iteration speed of the digital world is far faster than the construction speed of physical infrastructure."

"Whoever can compress physical deployments that would take months or years to complete into weeks will be able to sell their AI infrastructure at a very high price."

Gavin's investment strategy: Go long on bottlenecks, short on overall market risk.

"He strongly believes that AI winners will emerge, but that doesn't mean he's optimistic about the entire market; the QQQ put is his hedge against overall downside risk."
"TSMC actually limited the speed at which the bubble accelerated; as long as chip production capacity cannot expand instantly, capital expenditure is less likely to get out of control."
“Gavin is like an older, more stable, and more seasoned Leopold: the former’s success is measured in decades, while the latter is currently measured more in quarters.”

Assets worth betting on during the AI supercycle

EJ: Gavin Baker is an extremely prolific AI investor, yet virtually unknown to the general public. Over the past 20 years, he has been investing in some AI companies that would later become household names, long before they gained widespread recognition. He made early bets on Nvidia (a supplier of AI GPUs and accelerated computing cores) and Cerebras (an AI chip company), and he holds a very clear view that AI is not a bubble; on the contrary, it is a supercycle.

He believes that by observing watts (electricity), wafers (silicon wafers), and tokens (model generation and computation units)—the underlying infrastructure of AI—key bottlenecks and constraints can be identified. His conclusion is simple: the biggest returns in AI come from electricity, energy, and silicon wafer manufacturing, and have little to do with SaaS (Software as a Service) or chatbots like Anthropic and OpenAI. The entire industry will ultimately flow downstream to semiconductors, which are the "picks and shovels" (the assets that support the entire AI industry).

While many say the AI industry is already a bubble, he believes it presents a generational buying opportunity, especially for AI infrastructure. He uses approximately $4.1 billion in his fund to express this view.

If you hear him talk about these constraints, especially AI infrastructure, you'll find the theory very familiar. We've talked about an investor, Leopold Aschenbrenner, several times in the program before; he also made many investments in similar areas. The difference is that Leopold has only been doing this for about 3 years, while Gavin has been doing it for over 20 years.

Leopold manages roughly three times the assets of Gavin Baker, but show producer Luke once pointed out a good point: you might outperform Warren Buffett for a year, but can you outperform him for decades? Gavin Baker's track record suggests he may have a different perspective on this investment theory.

For those unfamiliar with Gavin Baker, it's worth noting that he's the founder of Atreides Management, an investment fund that has been investing in Nvidia for the past 20 years. The fact that you could hold Nvidia shares for 20 years and still be able to work is already incredible, as it should yield phenomenal returns.

Some of his recent wins include Cerebras and Astera Labs (an AI data center connectivity chip company). Cerebras is an AI chip company, and the show mentioned that its post-IPO valuation was incredibly high. There are also some companies you may not have heard of; in this episode, we'll follow his portfolio and judgments to see where he sees the AI investment opportunities.

So the question becomes, what exactly did he invest in, and why? Looking at Atreides Management's recent 13F filing (a quarterly disclosure of holdings by US institutional investors), this fund has approximately $4 billion in AUM (assets under management). Breaking down some of its largest holdings reveals that these companies all point to the bottlenecks in AI development that Gavin has repeatedly mentioned.

He has significant holdings in companies that aren't exactly glamorous, and many people haven't even heard of them. For example, Astera Labs accounts for almost 9% to 10% of the fund. You can think of Astera Labs as a connection layer between GPUs. If you imagine a data center as a system, the GPU is the engine, responsible for model pre-training, post-training, and inference. But for GPUs to work, they must transfer large amounts of data between each other and access the memory chips that store that data.

To achieve this, a "pipeline system" is needed. I'm going into a high-level way here because I don't pretend to understand all the underlying details. Astera Labs solves precisely this problem. When AI clusters scale to hundreds of thousands of chips, the bottleneck is no longer just the GPUs themselves, but the data transmission window—how to send the right data at the right time and access the right data. Astera Labs has built just such a pipeline system.

I hadn't heard of Astera Labs before doing my research for this episode. But I remember Cerebras was in a similar situation. Gavin talked about Cerebras about six months ago, and considering the timescale of AI, six months is a long time. Then it went public, and the show mentioned a valuation of about $60 billion, which then increased by 40% after the IPO. This suggests that Astera Labs could also be an important name in a similar trend.

Josh: Cerebras was one of his very early investments. He entered Cerebras very early in the company's life cycle, meaning he had been betting on this theory for many years. There are several other companies he has invested in long-term, the most flagship of which is, of course, Nvidia.

To have been involved with Nvidia for over 20 years and maintained conviction throughout is truly remarkable. I recently listened to two podcasts by Gavin, where he clearly expressed his belief that Nvidia can maintain its current profit margins and sustain demand. This implies that he sees Nvidia as having the potential to reach a market capitalization close to $10 trillion, and currently it's only about halfway there.

Another company worth mentioning is Micron Technology (a major global memory chip manufacturer). We discussed the AI investment stack and the positions of these companies within it in the previous episode, which I highly recommend watching. Micron is one of the largest memory makers. The program mentioned an astonishing figure: a year ago, its market capitalization was less than $100 billion, while at the time of recording, it had already surpassed $1 trillion, a tenfold increase in just one year. This illustrates just how crucial the memory problem is.

There are also some less conspicuous but very interesting companies. EJ, I especially want to mention one: Unity Software. Anyone familiar with games knows Unity; it's a game engine, and many popular games are made using this 3D rendering software.

So why would an AI investor invest in Unity, a company "that makes video games"? The answer is a 3D game engine. Unity is a world model builder with a deep understanding of physics, how the world works, materials, and lighting. When AI companies want to build AGI (Artificial General Intelligence) and humanoid robots, a crucial step is simulating virtual environments and datasets for training. Unity happens to be one of the most powerful tools for this. So, as a world model enthusiast, you should appreciate this example: a company known for its game engines has a clear path to becoming a major player in the AI world.

Gavin's investment theories and strategies

EJ: The theory behind world models is simple: current AI models or LLMs (Large Language Models) primarily understand the world through text and books, like a student sitting in a library, but they lack real-world experience. World models aim to unlock this: placing a game character into a simulated environment and allowing it to understand how physical reality works. For example, what happens if I drop my phone or kick a ball? What are the next steps? What should you do? World models solve this problem.

Currently, there aren't many players capable of scaling up this type of capability. Google is likely the current leader, with models like Genie 3 (Google's generative interactive world modeling project). The program also mentioned that Google recently released Gemini Omni, but this type of model hasn't truly reached its ChatGPT moment (its breakthrough moment).

What I like about Gavin is that his strategy is very similar to a barbell strategy. On one hand, he's very traditional, recognizing the need for GPUs and storage, so he bets on the biggest players, Micron and Nvidia. On the other hand, he's very forward-thinking, anticipating where the puck will go, so he bets on Cerebras because he believes inference will be crucial; he also bets on Unity because he believes world models will be the way to train robots and the next generation of LLMs.

His team also includes Positron, which makes inference chips. If this sounds similar to Cerebras, yes, they both revolve around inference. Gavin has repeatedly mentioned a trend in recent interviews: the infrastructure stack of AI models, especially the training stack, is shifting from pre-training to placing greater emphasis on post-training.

If you're in the AI community, you'll know this shift has already happened. Gavin is extremely focused on this. A model still needs to understand new information and new data; it needs to update itself. Just because it's pre-trained on a certain dataset doesn't mean it's a genius forever. It still needs to learn new information, which happens in post-training layers, and that requires a lot of computation.

Secondly, if you need AI models to truly think about problems, just like we think after receiving new information, "Is this perspective valid? Is there another theory that can explain it?", that's reasoning. Reasoning also requires a lot of computation. Current estimates suggest that the cost or revenue opportunities from reasoning alone could be 5 to 10 times the computing power invested in pre-training.

Therefore, both AI labs and chip makers are undergoing significant shifts. You've seen Nvidia release many GPUs geared towards inference to support agency applications. Gavin has also expressed his bet on inference through a series of investments.

The last point I found very interesting was Gavin's discussion of China. In the AI race, the narrative has always been China versus the US. China has a unique advantage: relatively abundant energy resources and the ability to expand chip manufacturing. The US is currently struggling in this area, which is why many processes are outsourced to TSMC (Taiwan Semiconductor Manufacturing Company, the world's most important advanced semiconductor foundry) in Taiwan.

His explanation is that China has a unique opportunity to create a very different kind of AI infrastructure or chip compared to the US, because it will be very focused on inference. You could say that Gavin is leading the charge in building US inference infrastructure through his investments in the US. I think this could be a huge opportunity in the future.

Josh: It's worth noting that this bet isn't just on the upside. He also holds a large position in QQQ puts (the put options on the Nasdaq 100 ETF). QQQ is an ETF that tracks the Nasdaq 100, a basket of stocks, and the second most traded ETF in the US. It has performed exceptionally well: up 55% in 2023, 25% in 2024, 20% in 2025, and 17% so far in 2026.

In other words, QQQ performs exceptionally well as an index fund; it's easy to buy because it's a basket of the top 100 stocks. Gavin is hedging against it. He's not saying AI won't win, but rather that he wants to invest in the key manufacturers that are truly solving bottlenecks, but he's not particularly optimistic about overall market sentiment. The QQQ put is downside protection: if the overall market collapses adversely, even if AI still wins in the long run, he has this layer of hedge.

Four types of investment opportunities

Josh: We can break down what he considers the most important investment bottlenecks into several categories. The first category is verticalized small language models. Ordinary LLMs, such as chatbots like Claude and ChatGPT, are generalized LLMs; they have a broad understanding of the world and can answer specific questions. But training a model around a specific vertical domain or a specific problem is another matter.

These specific problems often exist within enterprises, especially those that specialize in a particular problem or have formed a niche in a specific market segment. Verticalized SLMs solve this problem: they are frontier models, but highly optimized to run efficiently on specific enterprise data or locally on the device.

We've previously discussed on-device or locally run models. This is because your phone or other devices contain a large amount of highly personal data that you might not be willing to hand over, and companies might not have access to. Examples include medical records and financial details. I saw that OpenAI released a financial AI agent that can access your bank account, but it can't actually operate on your behalf because it contains a lot of personally identifiable information, such as your Social Security number and bank details.

Native models, or SLMs, can solve these kinds of problems. Gavin is largely betting that they will become very important in the future. There is one company he is very optimistic about: Apple. Although he may not have expressed an explicit investment interest, he believes that Apple will be one of the major device makers that will enable native models to run on devices.

If this is the future, we may no longer think of Claude as a model you interact with every day. What you might need is a personalized AI agent trained on your own data, which is what SLM could ultimately become. A general version could run on your phone, while numerous enterprises would run highly optimized, specialized models trained on their own proprietary data to better sell or market their products.

EJ: Apple is in a great position. I'm really looking forward to WWDC (Apple's Worldwide Developers Conference), it's coming soon.

Josh: Yes.

EJ: With only a few weeks until Apple's Worldwide Developers Conference, they'll be releasing new AI software and how that software integrates with hardware. This will be very important, and we'll continue to cover it. I'm looking forward to discussing this.

Josh: The second pillar is sovereign infrastructure. We often say that bits are much faster than atoms. This is very clear when looking at AI infrastructure: model quality improves almost exponentially, and the intelligence generated per watt, the intelligence corresponding to each token, will only continue to improve.

However, the speed of physical deployment has not increased at a similar pace, which in itself constitutes a moat. The hardware is extremely complex, with transistor precision approaching atomic levels; large-scale deployment in a world where existing infrastructure is already under pressure is not easy. The rapid popularization of electric vehicles has already put greater strain on the power grid, with many areas nearing full capacity. Now, AI brings with it both energy and chip problems.

Gavin is strongly betting on the fact that infrastructure is difficult to build, taking many days, months, or even years. He's betting on those who can compress this cycle into a few weeks. Therefore, the speed of physical deployment is itself a moat. He's narrowing down his search, looking for companies that can deploy as quickly as possible.

The first example that comes to mind is SpaceX (Musk's aerospace company), and the speed at which they built Colossus (xAI's large AI supercomputing cluster) and leased it to Anthropic, and possibly to other companies in the future. This infrastructure pillar is one of the key areas of focus for Gavin.

If you look at Leopold's team, this is also a core part. The reality is: building things is extremely difficult, but those who can build them can command very high prices. The program mentioned that SpaceX's largest source of revenue now comes from renting out data centers, not rockets. This illustrates how crucial this pillar is.

EJ: He cares about speed, but also about cost. He repeatedly mentioned a metric: performance per watt. What he really meant was that AI labs are increasingly concerned with how many tokens can be generated per watt.

If you consider that only about five companies are spending billions or even trillions of dollars this year on GPUs, compute, and the power that drives these systems, you'll certainly want a high bang for buck. This is especially true when hyperscalers expand to this scale, where cost becomes a core issue.

Let's say I ask Claude a question, and the cost of its answer is 2 cents; I ask ChatGPT a question, and the cost of its answer is $1. Even if Claude only has 95% of ChatGPT's intelligence, I would most likely still use Claude. This is because I can ask multiple questions and ultimately get the answer at a lower cost.

Therefore, the cost of accessing this kind of intelligence is very important. Just this week, Microsoft and Uber announced that they are actually reducing their use of Claude Code (Anthropic's AI coding tool for programming scenarios) because their annual budgets were used up in about four months.

You can see this in Gavin's portfolio: Cerebras, Positron, Astera Labs. He identifies very specific infrastructure bottlenecks and then makes a simple bet: if this company solves that bottleneck, achieves a certain level of performance per watt, and reduces token costs to a certain level, then the AI lab will buy more GPUs, more products, or more of that stuff.

So his theory is actually quite simple, despite the complexity of the specific technology: I only focus on the bottlenecks at the AI infrastructure level. If I can find a company that can improve performance per watt and make tokens cheaper, I'll bet on it becoming very valuable in the future, either through an IPO or by being acquired at a high price.

Josh: In this section, if anyone wants to replicate Gavin's deal, they need to know a few names: Astera Labs, Cerebras, SiFive (a RISC-V chip design company), and Positron. These four companies are crucial in this sector.

The fourth and final direction is the combination of energy and space. As we mentioned earlier, the terrestrial grid largely limits energy supply, and building new energy facilities is also very difficult. The program mentioned a statistic that about 40% of new data centers encounter very strong opposition, with people lobbying and protesting against their establishment.

There are two types of solutions. One type is to create out-of-the-box energy, which is portable energy. You can take the data center there and power it with a small energy device. Blue Marble, which Leopold is very optimistic about, belongs to this category.

Another category is orbital computing, which is a direction Gavin is currently focusing on. The largest and most core company in this field is, of course, SpaceX. It is the only company capable of becoming a highway to space, sending payloads into orbit, sending racks and data centers into low Earth orbit, and generating enough intelligence and power to send them back.

I think SpaceX is more significant than SpaceX itself. I'm a bit surprised that Gavin's portfolio doesn't have a larger allocation to space stocks, considering he sees it as a huge industry. Perhaps the reality is that it's still too early, and SpaceX is the linchpin to unlock this industry.

We need to closely monitor Starship V3 launches. We just saw a Starship launch last week, and it went very well. If Starship isn't truly operational, there's no space energy and no racks to orbit. It's a necessary condition because the payloads required for launch are enormous. Therefore, SpaceX is definitely a company we must keep an eye on, although many other second-tier companies will also be affected.

Why isn't this just another dot-com bubble?

Josh: Next, everyone will definitely ask, why isn't this just another dot-com bubble? Gavin has been asked this question many times, and he gave a very strong answer, and I basically believe him; his argument is very convincing.

His logic was roughly this: the dot-com bubble of 2000 was debt-fueled. Many people borrowed large sums of money to invest in unproven theories and products that nobody actually used or cared about.

If we compare this to the current AI supercycle that Gavin mentioned, OpenAI and Anthropic alone are expected to reach $200 billion in ARR (Annual Recurring Revenue) this year. And this isn't just fabricated money; it's money already secured through contracts, a large portion of which—the program stated 40% to 60%—has already been prepaid by enterprise and retail clients. In other words, there is real money in circulation.

Looking at GPU computing power, ignoring model labs, let's look at who's buying products from Nvidia. Google, Microsoft, Amazon, and Meta are all paying with their own cash reserves, not borrowing. Amazon is just reaching the tail end of its free cash flow; if they started borrowing, we could be worried. But the key point right now is that they haven't leveraged.

Moreover, these are among the world's top five companies, and in a sense, among the smartest companies, given their market capitalization, scale, and status. In contrast, during the dot-com bubble, numerous unknown companies raised vast sums of money and then burned it out in irrational ways. This cycle, however, saw some of the world's smartest companies spending with unleveraged funds.

The quarterly reports we've discussed on the show in recent weeks also show that profits are optimizing around these actions, and the model is improving and becoming smarter. So Gavin's core argument is: this isn't an internet bubble because it wasn't driven by leveraged funds; and the bottlenecks we're talking about are constrained by physical atoms.

Buying a bunch of memory chips and GPUs is one thing, but Nvidia can't oversell GPUs, and Micron can't oversell AI memory chips because they don't have enough chip manufacturing facilities. So his simple argument is: if you can't oversupply the entire market, then it's not a bubble. We're limited by not having enough picks and shovels to do this, and that's exactly what he's investing in.

Another good point: Gavin believes that if TSMC had been able to supply the chips, Nvidia could have sold $2 to $3 trillion worth of GPUs this year and next. In other words, TSMC was a key link in the bubble.

The reason is that if TSMC could meet the needs of these companies and supply them with that many chips, it would consume a huge amount of capital. Currently, the charts show that there isn't a significant disconnect between CapEx (capital expenditure) and operating cash flow; the cash generated by the company is still sufficient to support construction.

But if TSMC told Nvidia tomorrow that they could triple their production capacity overnight, Nvidia wouldn't refuse; it would start spending huge sums of money to buy chips. Other companies would also be forced to borrow money to buy these chips, at which point the CapEx bubble would start to grow and widen the gap between it and corporate operating cash flow.

However, due to supply constraints at each stage—constraints in storage, chip manufacturing, and energy, especially TSMC's constraints in advanced chips—we couldn't actually accelerate the construction process that much. Therefore, TSMC prevented the bubble from accelerating.

As long as TSMC's chip production capacity remains limited, and as long as Samsung and other chipmakers don't overtake its market share, the growth rate is relatively sustainable. It looks fast, but there's still a lot of unmet demand because we're simply not building fast enough. As long as this dynamic exists, I don't think there will be any major problems for the time being.

EJ: One more thing, you can't assume demand remains static, because it doesn't. AI-related demand is growing exponentially, and it's growing faster than the production supply of these chips.

I can only think of two ways to disprove this theory. First, someone miraculously replicates ASML (the world's core supplier of extreme ultraviolet lithography machines), and suddenly a bunch of ASML competitors appear. For those unfamiliar with ASML, it can be understood as follows: it produces machines worth approximately $400 million, which TSMC and all major chip fabs (wafer fabs) need. The program stated that ASML only has one team manufacturing these machines in Norway, and the lead time is extremely long, with an order backlog already stretching to about 5 years.

Second, we've created a completely different type of LLM that doesn't require as many GPUs or as much storage. But we're not seeing any signs of this yet.

I saw a news item today about SK Hynix (a major global supplier of high-bandwidth memory). It's the leading memory manufacturer and supplier for Nvidia GPUs, and practically a top dog in the AI memory field. It's currently receiving offers of approximately $50 billion to $100 billion from Google and Microsoft, who want to secure supplies for the next three years to fund its expansion.

This illustrates how hungry these large companies are for storage, and this is just one sub-sector within AI components. SK Hynix, on the other hand, says: "I don't want to guarantee your supply; I'll just raise prices." Its operating margin is approximately 70%, which is almost unbelievable in the semiconductor industry.

So Gavin's all-in move makes sense. It doesn't look like a bubble, though the market might react that way in the short term. When we opened our stock portfolio before recording today, almost all of them were down, but that was more of a reactionary reaction. The directional goal here is: we will only need more GPUs, more semiconductor chips, and there isn't enough supply, nor enough manufacturers.

Gavin's portfolio

Josh: The conclusion is: electricity and wafers. That's it. They're two brick walls, two limiting factors preventing us from accelerating too fast. As long as electricity and wafers remain valuable, demand is strong, and supply is limited, there are still good days ahead.

If you want Gavin's portfolio of TLDRs (TLD, TL;DR version), I can read about his largest holdings. Again, this is not investment advice. This is what Gavin holds, not what we hold. I don't know if these stocks will go up, down, or just sit still.

His largest position is somewhat counterintuitive, a QQQ put position (Nasdaq 100 ETF put option). Overall, he is bearish on the market, which is noteworthy. His second largest position is Astera Labs, at approximately 7.4%, ticked by ALAB. His third largest position is Unity, the 3D software company.

There are many more: Ciena (optical networking equipment company), Micron, Nvidia, Amazon, Lumentum (optical communications and laser devices company), Alphabet (Google's parent company), Coherent (optoelectronics and materials company), Roblox (gaming platform), EchoStar (satellite communications company), Twilio (cloud communications platform), and Wayfair (furniture e-commerce company). This person invests in everything.

If you're interested, you can check out his 13F; we'll include a link in the description. But that's Gavin's point: the bottlenecks are power and wafers. As long as these constraints remain, it's basically a one-way upward trend. EJ, how do you absorb this information? How would you process it?

EJ: The market has been volatile ever since Leopold's 13F came out. While recording this episode, I increasingly realized that Gavin is a bit like an older, smarter Leopold. He's been in this industry for a long time. Maybe he doesn't have $13 billion in AUM, but I feel like he'll still be around in 10 years.

If you're thinking, "I don't want to chase AI developments every minute, every hour, every day; I just want to put my money there and see how it grows over the next few months or years," then Gavin's portfolio might be a good reference. Of course, this is not investment advice.

He's taking a more cautious, long-term, and future-oriented approach. If his trend predictions ultimately come true, like his early bets on Nvidia and Cerebras, there could be exponential returns in the coming years. But all of this is built on one core belief: we're not in a bubble.

I'm curious if the audience agrees. Obviously, most people aren't as technically savvy or as deeply involved in the grassroots as Gavin. But after listening to this episode, do you think we're in a bubble? Or not? What are the reasons for and against? Is there anything we've missed? Josh, before we end, do you think we're in a bubble right now?

Josh: I think we're definitely in a bubble. The question is, what stage of the bubble are we in? That's debatable. Right now it looks more like an early stage, so hopefully it stays that way. According to Gavin, as long as TSMC continues to limit chip production, we're fine.

This is the overall outlook. We've already talked about Leopold, whose success is currently measured in quarters; now we're talking about Gavin, whose success is measured in decades. Many people's own answers might fall somewhere in between.

If you enjoyed this episode, don't forget to share it with your friends. Also, tell us which asset class you're most optimistic about. Maybe it's not a particular theory, but a stock code worth paying attention to. I find this exciting because everything is moving quickly, with lots of fluctuations, both up and down, and it's very engaging. See you tomorrow, good morning.