The S&P 500 is currently a richly priced market. This pricing reflects expectations for a smooth economic path and positive corporate reactions. While the market has remained resilient so far, it appears to be ignoring potential risks on the horizon.

It's a richly priced market which is building in expectations that the pathway is going to be a benign one in terms of the economy and how companies react to it. There are lots of things on the horizon that could potentially trip up the market. Political economic war. And those things don't seem to be priced in.

Aswath notes that investors have been correct in their optimistic stance until now. Even after initial shocks like tariffs, the economy did not collapse and inflation did not surge back. This has led the market to continue pricing in very favorable conditions. However, the market may be vulnerable because it is not accounting for political or economic threats that could still emerge.

Big market delusions are a natural part of any major disruption. These delusions are driven by human nature. When a massive new market emerges, ambitious and smart people want to be part of it. This creates a selection bias where venture capitalists fund the most overconfident entrepreneurs. This group collectively looks at a big market and thinks they can conquer it. This leads to overestimated revenues and growth, causing the market to price these companies too high as a whole.

The big market delusion is a feature of big change. Individually, some of those companies are going to become great winners. But collectively, these companies are going to be priced too high. It happened with the PC business in the 80s, the Internet in the 90s, social media in the last decade, and now it is happening with AI.

While a market correction is inevitable, it is also healthy. This overreach is how human beings change and how markets create progress. Aswath notes that if the world were run by actuaries who only focused on probabilities and expected values, society would never progress. Innovation requires people to overreach and take risks that might seem irrational from a purely statistical perspective.

Thank God we weren't run by actuaries. For change to happen in economies, you need people to overreach. The consequence of that is you're going to have market bubbles and market corrections. They're a feature, not a bug.

Chaos is a necessary feature of markets to enable significant change. A market with no bubbles might be stable, but it would likely produce no innovation. We must accept the cycle of overreaching and correcting as the price of moving forward.

Aswath notes that the current AI boom can be divided into two parts. The first part is the architecture, which includes companies like Nvidia, data centers, and power companies. This is the easier side of the equation because the spending is visible and the infrastructure is being built. It is comparable to Cisco during the early days of the dot com era. The real uncertainty lies in the second part, which involves the products and services that this architecture will eventually support. While we know the cost of building the factory, we do not yet know what products it will produce or if customers will buy them.

The AI architecture is getting built. There is no uncertainty there that people are spending money. The uncertainty is the cost of building the factory. You know what we are uncertain about? What product that factory will produce and whether anybody will buy the product. That is where the delusions are forming.

Current valuations for companies like OpenAI and Anthropic reflect a big market delusion. On a collective basis, these companies are likely overvalued because there may not be enough total revenue in the end game to justify such high prices. LLMs themselves are not standalone money making devices. Instead, they function as part of the architecture. Aswath suggests they will be similar to smartphones. Just as the iPhone acts as a delivery mechanism for companies like DoorDash and Airbnb, LLMs will provide the foundation for other businesses to create value. The networking benefits will come from the platforms built on top of the models rather than the models themselves.

LLMs will be like the smartphone. In the DoorDash business, you are not dealing with your smartphone, you are dealing with DoorDash. The networking benefits come from what DoorDash and Airbnb create in their platform. And that is what I think will be the end game in the product or service business that you see coming out of AI.

Aswath argues that trying to value private AI giants like OpenAI or Anthropic right now is almost pointless for most people. These companies are currently driven by mood and momentum rather than fundamentals. Unless someone possesses deep technological expertise to identify the eventual winner among Large Language Models, the space is best avoided.

Collective overvaluation means unless you have some technological expertise that allows you to pre identify which of these LLMs is going to end up the winner, this is a space you should stay away from.

Even when these companies go public, they will initially remain a playground for traders. In the early stages of being public, stock prices will fluctuate based on sentiment rather than intrinsic value. Investors who try to use traditional valuation to justify their decisions in such a volatile environment often get eaten alive. Aswath notes that in these situations, valuation simply does not matter on a short term basis for price movements.

As an investor, you're going to get eaten alive if you go into that space and bring your valuation to justify what you do. Because God help you, that valuation is not going to matter on a day to day, week to week, month to month basis in terms of what it does for prices.

Aswath notes that Nvidia and other chip companies currently enjoy a protected position because the money for their products has already been spent. Even if the private investments in large language models turn out to be mistakes, Nvidia does not have to return the cash. These investments are sunk costs for the customers. However, a market correction would eventually impact Nvidia because the new money required to justify future growth will likely level off.

In many ways Nvidia and the chip companies and the data centers are not just the front end of AI, they're the most protected. The money's already been spent. The mistakes are already sunk costs. And even if in hindsight we decide this bubble burst, we're not going to go back and collect our money back from them.

If a correction occurs, the entire market will likely drop because AI companies represent such a large portion of it. Aswath explains that there will be nowhere to hide in stocks. Even portfolios without AI stocks will suffer. He compares the situation to Cisco during the dot com bust. Cisco struggled for a decade because the company did not recognize the shift in the market and kept trying to grow through acquisitions as if nothing had changed.

There will be a cleaning up phase where everybody up and down the cycle and even people outside are going to get hurt when that correction hits. Which is one reason people complain about bubbles. They don't have the perspective to think about when they entered the game. How much money they made before they lost the money. They focus just on the losses.

The lesson for Nvidia is to adjust ambitions when the environment changes. Aswath believes Jensen Huang is a sensible leader who does not lead with his ego. This suggests Nvidia might be better prepared to adapt than Cisco was in the early 2000s.

Aswath emphasizes that personal investment decisions should rely on one's own analysis rather than the excitement of venture capitalists. While Nvidia has delivered massive returns, it is currently priced so high that an incredible number of things must go right just for a new investor to break even. This makes the stock difficult to justify as a new purchase at today's prices.

If you ask me would I buy Nvidia at today's price? I wouldn't. I think you need too much to go right to break even. But that's me and that's my money. If you ask me, should I buy Nvidia, I'm going to give you the numbers and say you make your own judgment.

There is a significant gap between valuation theory and practical portfolio management. Textbooks suggest that if a stock is overvalued, you should sell it immediately. However, this ignores real-world constraints like taxes and the question of what to do with the resulting cash. Aswath managed his own Nvidia position by selling it in stages over four years. This staggered approach helps manage risk and tax liabilities without the stress of trying to time the absolute peak.

If something is overvalued and you have it in your portfolio, the books say you sell it right away. I think that misses a lot of real world issues like taxes and time horizons. I sold my last quarter of Nvidia at the end of last year, but it was staggered over four years.

Accepting that you might miss the final 25 percent of a stock's climb is part of a healthy investment strategy. The goal is to pass the sleep test rather than maximizing every possible cent of profit. Regret over missed gains can lead to damaging emotional decisions in the future.

Aswath has held Microsoft in his portfolio since 2014 and continues to keep it today. While he would not buy the stock at its current price, he finds the path to justifying its value more realistic than other high growth stocks. For a company like Nvidia, the level of performance required to justify its price is much higher. Microsoft does not need as many things to go right to remain a reasonable part of a portfolio.

With Microsoft too, I wouldn't buy at today's price, but I don't need as much to happen to justify leaving in my portfolio. So it's a question of degree to me. Nvidia, the degree to which you got to deliver great stuff is far greater than it is in Microsoft.

Tax considerations also play a significant role in the decision to hold rather than sell. Living in California means facing high capital gains taxes. Aswath explains that he is willing to live with a slightly overvalued stock in his portfolio if the tax consequences of selling are too high. As long as the overvaluation remains within a manageable range, the cost of paying taxes outweighs the benefit of selling.

Many tech companies use AI as an excuse to explain all their revenue growth. While Microsoft has a robust cloud business, this segment was already the fastest growing part of the company long before the recent AI surge. Everyone is on the cloud now, from businesses to individuals. It has become a fundamental part of modern life.

I don't think it is as dominant a piece of the cloud business at any of these companies as people make it out to be. In other words, if tomorrow AI spending dropped, I am not sure that you are going to see cloud revenues at any of these companies decline. Their growth might get lower, but you are not going to see a drop off in revenues.

Aswath argues that AI spending is not as dominant in the cloud business as many people assume. If AI spending were to drop tomorrow, cloud revenues would likely remain stable. The rate of growth might slow down, but the base revenue is supported by the fact that cloud services are now essential infrastructure for everyone.

Cloud computing has transformed from a service into a fundamental necessity. It functions like a utility because the volume of data we generate makes physical storage on devices almost impossible. When a cloud service goes down, significant portions of the global economy come to a halt. While margins for companies like Microsoft and Google are currently healthy, the long term question is whether these will decline as the business reaches maturity and competition increases.

As we realize the minute one of these clouds goes down, we realize how much of the world comes to a stop. I think it is a necessary business. The question is, what will the margins look like as it matures as a business?

A key factor in the success of cloud platforms is their stickiness. Companies create models that make it difficult for customers to leave once they are integrated into a specific platform. Even if a competitor offers lower costs, the difficulty of switching often prevents users from moving. This creates networking benefits that protect established players from losing their market share.

Aswath notes that Google remains heavily dependent on advertising revenue despite its many side projects. He describes Alphabet as six dwarfs and one giant because most of its other bets have failed to become standalone businesses. This highlights the sugar daddy problem where corporate venture capital units lack the ruthlessness of traditional investors. Because they can always go back to the parent company for more money, they struggle to walk away from bad investments.

To be a truly effective VC, you have to be ruthless in terms of cutting your losses and walking away from bad businesses. But that is tough to do for corporate VCs because they keep getting this money to push more and more.

The future for Alphabet depends on whether investments like Gemini and Waymo can eventually deliver value on their own. While Gemini represents a better approach than previous bets, it remains to be seen if these technologies can become independent profit centers that justify the massive investment.

Aswath views large tech companies like Alphabet and Meta as entire ecosystems rather than just collections of individual products. In this model, platforms like Facebook, Instagram, and WhatsApp work together to generate advertising revenue. The primary goal of these companies is to keep users within their specific network of services. YouTube serves as a highly effective tool for Google to maintain this level of engagement and keep people in their ecosystem.

The entire ecosystem delivers the advertising revenues. So even though the revenues look like they go to Facebook, it's somebody using WhatsApp who sent a link to somebody else using WhatsApp to go on a Facebook page. So these companies just want you to stay in that ecosystem and Google keeps people in there.

While YouTube functions as part of a larger network, it also succeeds as a standalone business through subscriptions and TV services. Aswath considers the one billion dollar investment in YouTube as one of the greatest single investments Google has ever made. The value created has paid off one hundred fold because the platform provides benefits that go far beyond its direct revenue. It functions much like Amazon Prime, where the total value to the parent company is significantly greater than what is collected from subscriptions alone.

That billion dollars invested in YouTube has paid off 100 fold in terms of value created for the company. I think YouTube has value way beyond what they collect as subscription and actually direct revenues on YouTube.

Meta approaches AI differently than other major tech companies. While companies like Microsoft and Amazon spend heavily on AI to serve external customers, Meta focuses its spending on internal product improvements. This means the return on investment will not appear as direct revenue from AI services. Instead, the value shows up through increased user engagement within the Meta ecosystem.

If using AI keeps you in the Meta ecosystem, that is a huge plus in terms of advertising revenues. The ROI might show up as you spending more time on Instagram because AI has made it more attractive to you. They found a way to make it more addictive to you, and that then shows up as more advertising.

Aswath suggests tracking the amount of time users spend on these platforms. Currently, a typical user spends about 57 minutes a day within the Meta ecosystem. If AI can push that engagement toward 90 minutes, the impact on advertising revenue would be significant. The effectiveness of Meta's AI strategy is best measured by whether it successfully keeps people on the platform for longer periods.

Oracle is spending on AI differently than other tech giants. They have made a massive bet because they want to join the trillion dollar club. Aswath compares this strategy to a gambler putting all his chips on a single number. While AI spending is lowering free cash flow for many companies, the financial foundations are much stronger than they were during the dot-com era.

The dot-com bubble investment was far smaller than the AI investment. You didn't have tens of billions of dollars being invested. And second, the companies making those investments were often money-losing companies.

Today, the most cash-rich companies in history are making these investments. Their debt is a small fraction of what they earn. Companies like Meta and Microsoft could pay off their entire debt with just one year of free cash flow. Even if the AI market collapses, these firms will still have positive earnings. They might face impairments, but they will not face default.

This is a major change from the early 2000s. Amazon almost went under in 2001 and was only saved by a last-minute cash raise. The current leaders do not have that kind of risk hanging over them. However, smaller companies that borrowed money specifically to find a place in the AI market will be in deep trouble when a correction occurs.

While companies like Microsoft and Amazon are seeing massive growth, it is important to distinguish between their cloud businesses and their AI initiatives. The cloud is a business that exists independently of AI. It is a function of the times we live in because of how we store and use data. While AI has helped the cloud grow, the cloud business was not created by AI.

The cloud is a business that is independent of AI. It is a business that AI has helped, but it is a business. It is a function of the times we live in. So we should not act like the cloud business was created by AI.

Aswath argues that AI related revenue likely makes up a tiny fraction of total revenue for these tech giants. If the AI segment were to collapse, the impact would show up at the margin. It might cause revenue growth to drop from 8% to 4%. This would be bad for the stock price, but it would not be a catastrophic failure for the companies themselves. They have other stable cash flows that will keep them going well beyond any AI market correction.

The host notes that capital expenditure has exploded as these companies buy GPUs to support AI. Aswath agrees that this spending is the risky side of the equation. However, the revenue side is much more resilient. Even if AI crashed, the demand for the cloud would not disappear. Core products like Microsoft Word and Xbox will continue to be tremendously profitable regardless of what happens with AI.

Tesla possesses a unique power to shift its narrative from one industry to another, carrying its investors along in the process. Advocates for the company rarely focus on cars anymore. Instead, the focus has moved to robotics, automated driving, and ride sharing. These investors often behave more like fans than traditional stakeholders. They buy into the idea that Tesla has an underlying strength to enter, disrupt, and eventually dominate entirely new sectors.

The only way you can justify Tesla's valuation is by mapping out a way it gets into a different, higher margin business. It cannot sustain this value as a car company.

Aswath points out that Tesla's valuation metrics are unjustifiable for a simple car manufacturer. With a forward price to earnings ratio of 220, the company must successfully transition into higher margin sectors to support its price. However, the business now carries a significant political component. Similar to investing in China where the government is always a factor, Tesla has become a political stock where people interact with the brand based on the CEO's perceived politics. This shift led Aswath to exit his position at 300 dollars. Even though the price later climbed to 500 dollars, he remains uncomfortable with business narratives that are heavily influenced by politics.

AI might not improve company profit margins as many people expect. If every company uses the technology to cut costs, the benefits will not stay with the business. Competition forces companies to lower their prices when their cost structures go down. This is what happened with online retailing. Consumers gained better choices and lower prices, but the businesses did not necessarily walk away with higher margins.

If everybody has it, nobody has it. If I can give you an AI product that makes it easier for your customers to buy stuff or to cut costs on inventory, your costs are lower. But I am selling that same product to your competing grocers. They are all lowering costs. If everybody's cost structure goes down, you don't end up with higher margins, you end up with lower prices.

While individual companies selling AI products will likely make money, the companies buying those services may not see a boost in cash flows. The primary beneficiaries will be society and consumers. This trend can lead to lower prices in specific segments, like store inventory management, but it does not guarantee overall deflation. Other factors like fiscal policy and the cost of raw materials still influence the broader economy.

Regarding the stock market, current valuations appear high. Stocks are overvalued, but they are not necessarily in a bubble. The market has been overvalued for most of the last ten years.

If you view that overvaluation as a signal you should get out of stocks, in hindsight it would be a terrible signal. If you're going to make an argument that it's a bubble, that's a different argument, but I don't see it in the numbers.

An overvaluation signal can be misleading because markets can remain in that state for a long time without a crash.

Double digit earnings growth has been the average for the last fifteen years. This consistency suggests that something fundamental has changed in the economics of business. Companies today are able to deliver earnings through crises and recessions that would have caused significant drops in the last century. While some point to globalization, recent trade wars and tariffs have put that theory to the test. A more likely reason is the shift toward a tech-dominated market.

Tech companies are much more flexible and adept at dealing with things changing around them. The biggest tech companies are money machines with earnings that are incredibly predictable and stable because they dominate their markets.

Aswath notes that this might represent a structural shift in the economy. Modern markets might simply be valued differently than they were in the past. While investors used to consider 16 or 18 times earnings to be the norm, the resilience of tech-driven earnings could explain why the market now trades at 24 times earnings. Investors may need to adjust their expectations to account for these more stable and flexible business models.

Aswath points out that the equity risk premium was around 4.23% in January. This value is near the median when looking at data going back to 1960. It looks low only if you compare it to the years after the 2008 financial crisis. That period had a higher premium because of constant market instability.

Post 2008, we saw a jump in the equity risk premium partly because of the 2008 crisis and these rolling crises since. So this is low. If you frame it against 2008 numbers, it is pretty much what you would expect if you frame it against a much longer time period.

People often reach different conclusions about the market because they use different timeframes for their data. Aswath mentions that some people call the current market a bubble. He argues that these claims lack supporting evidence in the current numbers. To prove a bubble exists, one must find data that is not currently visible.

A bubble is always plausible in the stock market. However, using that possibility as a basis to stay out of stocks is often a mistake. Investors who try to time the market usually end up with lower returns than those who stay invested through the cycles. The reaction to a potential bubble creates more long-term costs than the bubble itself.

To me, it is a horrendous mistake to make because in hindsight, you are going to end up with returns that lag. Somebody who does not try to time the market goes through the bubble, loses money, and comes back again.

Using the price-to-earnings ratio to predict a bubble is a lazy approach. It is like a doctor judging a person's entire health based only on their temperature. The PE ratio is a volatile and messy metric. Even if metrics like the Shiller PE suggest the market is high, they often miss a lot of context. Aswath points out that even Robert Shiller has changed his perspective over time.

Price earnings ratio is the most volatile and messy of all pricing ratios to make a judgment about the entire market just because the PE is higher than the average.

A better argument for market overvaluation should focus on fundamentals. If margins are at all-time highs or competition is increasing because of AI, those are reasoned points for concern. A simple story based on a PE ratio does not provide enough substance for a real judgment.

The equity risk premium today is significantly higher than it was during the dot com bubble. This increase is not primarily driven by expected earnings growth. Instead, it comes from the massive amount of cash that companies return to shareholders. US companies in the S&P 500 are currently returning about 85 percent of their earnings to investors. While dividend yields alone look historically low at around 2 percent, adding buybacks creates a total cash yield of about 4 percent. This total yield is well within historical norms.

I think buybacks are the way in which companies will return cash. I think of them as flexible dividends, and I think they actually make more sense from an equity cash flow perspective than these fixed dividends that you get locked into in good times and in bad times.

Aswath argues that buybacks have fundamentally changed how we should view cash flow. If an investor only looks at dividends, the market appears to be in a bubble. However, this perspective ignores the shift in corporate policy. Companies now treat buybacks as a more flexible way to return cash. Relying only on dividends to value stocks would have kept an investor out of the market for the last 15 years.

When performing a discounted cash flow analysis, using net income is a fundamental error. Earnings are not the same as cash flows. A proper valuation requires adjustments for non-cash expenses and reinvestment needs. If an investor wants to focus solely on earnings, they should use pricing tools like PE ratios rather than a DCF. Many DCF models are fundamentally flawed because they fail to respect this first principle of cash flow.

Any investor who is discounting earnings just stop. Don't even try to do DCF. If that is the first principle you are failing, then stick with PE ratios. I would rather that you do an honest pricing than these messy, dishonest DCFs.

Using dividends and buybacks as a proxy for free cash flow is a more practical approach for valuation than calculating free cash flow for every individual company. Estimating free cash flow for the entire S&P 500 is difficult because financial service companies make up a significant portion of the index. For a bank, traditional metrics like capital expenditure and working capital are almost impossible to define clearly.

You end up with this messy definition. So we are using dividends and buybacks not because we think they are the right. They are a proxy for free cash flows. Because ultimately we are assuming that is where the cash flows to buy back stock and dividends come from.

Large tech companies like Microsoft and Meta are currently spending massive amounts on capital expenditures, often fueling the revenue of companies like Nvidia. Aswath suggests that investors should not worry about the spending itself but rather the potential returns on that investment. To grow, companies must reinvest. The risk lies in the fact that these companies are investing billions based on the hope of AI potential without clear evidence of how that business will eventually look.

You can't have your cake and eat it too. You want these companies to grow? They have to reinvest. So I understand the zeal they have to be in the next big space. The challenge there is they are investing this upfront on a hope and potential. There is really nothing on the ground right now that you can use to justify the investment.

Apple may be in a better position than its peers because it has not yet committed tens of billions of dollars to the same level of infrastructure. If these massive investments do not pay off, it will eventually impact the ability of these companies to return cash to shareholders through buybacks. If companies begin borrowing money to fund these investments, it could create new risks for their historically stable earnings.

Gold serves as a refuge when trust in systems or institutions disappears. It functions as insurance against a major catastrophe. Currently, a growing segment of the market believes we are heading toward a significant economic cliff. This group has expanded to include individuals who historically would never have sought gold as an investment. For instance, even prominent figures like Ray Dalio have expressed interest, which was not the case twenty years ago.

Gold is what you go to when you lose trust. It is your insurance against catastrophe. The fact that the segment of the market believing we are heading for a cliff is bigger than it has been historically should send us all a message that the potential for catastrophe is higher now.

The unusual composition of this crowd suggests that the risk of a historical disaster is elevated. While everyone must decide for themselves whether gold is the right tool, the need for some form of insurance has become more apparent in the current global climate.