When I See A Bubble Forming, I Rush In To Buy, Adding Fuel To The Fire

Trade Thesis

There has been a sharp and growing divide between two breeds of investors. On one side there are investors who are talking about a world where artificial intelligence, or AI, will transform the way in which people live and how that creates a bullish narrative for companies that are in this space. On the other side, there are deep-value investors who look at what’s going on with AI-related stocks and say that it’s a bubble because there is far too much froth, thus valuations are too high.

The former group buy and the latter group short. What makes me unique is that I agree more with the latter group, but I would look to buy like the former group.

The famous quote in the title of this article was said by George Soros. When he saw a bubble, rather than avoiding it like traditional Graham and Dodd value investors, he would rush in to take advantage and further the reflexivity. This is exactly how I view Nvidia Corporation (NVDA) and Advanced Micro Devices, Inc. (AMD).

I learned this lesson the hard way with Tesla, Inc. (TSLA). Whilst I was never short it as I knew the dangers of shorting something that had the kind of momentum it had, I had never bought either, even though I was aware of the bullish thesis around it since 2018. I would justify it initially saying that it had too much debt and the sales weren’t high. Then when the price went higher I would say that it wasn’t profitable. Then when the price went even higher I would say the valuation was absurd. What I didn’t understand at the time was that the fundamentals and valuation weren’t driving the price, in fact, to this day Tesla is going for 73x earnings, and even with potential growth priced in the valuation is insanely high.

Instead, what drove the stock, and many like it was flows created by positive sentiment around EVs and eventually FOMO, and of course a macro environment where the cost of capital was low which acted as a tailwind for companies like Tesla.

My view of AI-related stocks is similar to the dot-com bubble. During the dot-com bubble, there were bullish arguments from investors that the Internet would become bigger in the future. They were absolutely right about that, but dot-com stocks still went down. This was due to a few reasons, the first of which is that they underestimated the amount of time it would take for the Internet to be used for everyday uses; the second reason was that business models at the time had no profitability to them, such as Pets.com, which would’ve been unprofitable regardless of the amount of revenue it got due to a business model that was essentially selling dollars for 50 cents; the third reason being competition, such as Yahoo being out-competed later on by other players in the space; and the fourth reason being valuation. For all of these reasons, investors were right on the narrative but wrong on the trade. In fact, this reminds me of another George Soros quote:

Every bubble consists of a trend that can be observed in the real world and a misconception relating to that trend. The two elements interact with each other in a reflexive manner.

With the above thesis presented, I want to go over what I believe a fair valuation for Nvidia and AMD would be and how to trade this bubble that is likely just starting to form.

Fundamentals and Valuation for Nvidia

When it comes to the fundamental for Nvidia there are two parts: the near future and the far-out future. The near future is more predictable and we have good estimates of what the next two years will look like, but the far-out future is where the estimates are like throwing darts in the dark, as it’s difficult to see where Nvidia will be a decade from now.

Here’s what management expects for Q1 2024:

NVIDIA’s outlook for the second quarter of fiscal 2024 is as follows:

-Revenue is expected to be $11.00 billion, plus or minus 2%.

-GAAP and non-GAAP gross margins are expected to be 68.6% and 70.0%, respectively, plus or minus 50 basis points.

-GAAP and non-GAAP operating expenses are expected to be approximately $2.71 billion and $1.90 billion, respectively.

-GAAP and non-GAAP other income and expense are expected to be an income of approximately $90 million, excluding gains and losses from non-affiliated investments.

-GAAP and non-GAAP tax rates are expected to be 14.0%, plus or minus 1%, excluding any discrete items.

With $11 billion of revenue expected and 70% margins, we’re looking at $7.7 billion in gross profits and around $6 billion EBIT (if we’re being generous). Annualize this and that $24 billion EBIT. We can presume that between now and say 2025 there is likely to be some growth, so EBIT might be closer to $30-$40 billion which makes for an EBIT multiple of 25-30x presuming that there is no stock dilution between now and then.

25-30x EBIT is definitely an expensive multiple as a good terminal multiple would likely be around 12-15x if there were no future growth prospects. But there are future growth prospects, so the 25-30x EBIT multiple doesn’t sound too absurd.

I went ahead and used a DCF calculator to determine what kind of growth rate we’d need to see for Nvidia to be at fair value:

First off when it came to the earnings I used the estimate of next year’s annualized EBIT of $24 billion. This is far easier than plugging in EBIT on a per-share basis. Also, I’m choosing to use EBIT as it’s hard to determine what both interest expense(or leverage for that matter) and taxes will be many years from now.

Growth in demand for GPU chips will likely take place over the next decade, so I used 10 years as the major growth period, then afterwards a perpetual growth rate of 5%, which is conservative. As for the discount rate, 12% seems about right even though it is higher than what a major benchmark would return, as there is of course added risk in the growth of the company. The unknown is the growth rate.

To get to a $1.075 trillion market capitalization, a 20% annual growth rate is needed for 10 years. If you are a Nvidia shareholder you need to ask yourself if that’s possible based on your own assumptions.

In my opinion, having a 20% growth rate is definitely possible. In fact, a Goldman Sachs report had the following to say

Breakthroughs in generative artificial intelligence have the potential to bring about sweeping changes to the global economy, according to Goldman Sachs Research. As tools using advances in natural language processing work their way into businesses and society, they could drive a 7% (or almost $7 trillion) increase in global GDP and lift productivity growth by 1.5 percentage points over a 10-year period.

While I don’t have the caliber of knowledge to know how exactly they got to these figures, from the research I’ve done it seems to be common consensus that these large multi-trillion dollar growth figures are possible.

The issue here is twofold. First is competition. There is no guarantee that Nvidia will be the leader in GPU chips a decade from now, in fact, if you look at the largest companies in the tech world today and compare that to the dot-com bubble times, only one of them, Microsoft has continued to stay a mega-cap; all other tech companies are down big time from their peaks. The second and more important question is whether technology will change so much a decade from now that the chips that Nvidia is delivering currently and plans to deliver over the next few years will be considered obsolete for what will be needed a decade from now. An example of a past disrupter is the cloud; prior to the cloud, hardware would be used to store data, but when the cloud came along, it was cheaper and more convenient to store data on the cloud, so external hard drive disks became obsolete and companies producing them had to quickly pivot. My worry would be that Nvidia or other companies in the space would have similar issues in keeping up with innovation.

Fundamentals and Valuation For AMD

Let’s run a similar DCF model for AMD to see what kind of growth is needed.

AMD looks more richly valued compared with Nvidia, but it’s due to its terrible starting point. Below is the DCF model:

I used similar figures above. There are some major differences with AMD though. The first is that it has an EBIT that is at breakeven, which of course wouldn’t work with this DCF model since anything times zero is zero. So, I choose to use 2 billion of EBIT as the base or starting off point. I think $2 billion of EBIT is a reasonable starting off point as the last time quarterly operating income was positive the company did around $500 million in operating income, so annualizing that to $2 billion would work as a good base.

It needs 27% annual growth for ten years to justify its valuation with the same inputs as Nvidia. While what I presented so far might make it sound like Nvidia is the sure bet since its innovation is further along and needs less growth, the slim to non-existent margins of AMD make it tougher to do this analysis.

When you look at it from a P/S ratio perspective AMD looks a lot better. The P/S for AMD is 9 while that of Nvidia is 42. The main issue with AMD is its lack of margins on current revenue; if it were to get margins like Nvidia its P/E ratio be 20 or so.

Because of all of the above reasons, I find AMD to be the “cheaper” play, but you get what you pay for in that margins are terrible hence the risk is higher, and they are just starting out with making new chips whereas Nvidia is far ahead.

How To Trade These Stocks

Because I come from a background of an old-school deep value investing framework with global macro added into the filtering process, it’s difficult for me to buy into any of these companies with their current valuations. On top of that the CAGR I could make comparably with my current framework is higher than would I would get from both of these stocks over the next couple of years. Because of this, I would look to treat both of these stocks as a speculation opportunity by playing into the George Soros framework of speculating on bubbles that are likely to get bigger through reflexivity.

To be a good speculator one needs to understand how to manage their downside as that is the biggest risk in these large run-ups. In my opinion, setting up a trend-following system to trade these stocks would likely be the best way to follow the momentum.

Below is a chart of AMD with both a 50 and 200-day EMA:

There are a lot of different trend-following strategies that would work here. A simple trend-following strategy that is easy to get signals on is for a shorter-term EMA like the 50 day to cross a longer-term EMA like the 200 day. On the chart, it is represented with the plus sign. This would be a simple binary system where you are either long or short.

While that system would work, I’d likely want something more dynamic. My personal preference would be to use the slope of the 200-day EMA. When it is in an upslope it is a buy signal and vice versa for a downslope. If it is flat or within a margin of error then no position is taken. As for the stop loss here, we could estimate how low the stock price would need to go till the EMA would flip in the other direction. For example, if a 5% drawdown in the stock would cause the EMA to flip from upward trending to downward trending, then a 5% stop loss could be used. Also, one’s risk budget could be adjusted to this. So if the drawdown till it hits the stop loss is 5% and my risk budget allows for a max drawdown of 10%, then I can buy the stock with 2x leverage. Of course, as the price goes exponentially higher the drawdown till it hits the stop will go up, so the leverage would need to go down by the same amount as the price goes higher. On the flip side if the price goes lower you can increase leverage as the risk to stop is lower.

This creates a dynamic trend-following system where downside risk is controlled to one’s own risk tolerance through a risk budget, leverage can be used without the risk of ruin, and profits can be taken as the price goes higher and vice versa for increasing position sizing during drawdowns.

Using Options When Overbought

Although a trend-following system sounds great there are some downsides. The first is that when the underlying asset isn’t trending or when it’s the price very choppy it doesn’t work well. I’m not too worried about this risk as it usually doesn’t occur when there are quick run-ups in price like we have seen in both Nvidia and AMD. The second downside is that the risk to stop is very high when the stock is overbought. An example of that is Nvidia currently.

Below is a chart of price action on Nvidia:

Nvidia is so overbought at its current levels that even a drawdown of 25% would still leave it in an uptrend. Because of this, the risk to stop in a trend-following system would be far too large for most people’s risk budgets, so very small position sizing would need to be used. The issue with very small position sizing is that little of the upside is captured. Because of this using options would work better to capture the upside when overbought.

The issue with buying options is that the implied volatility is far too high. Selling downside optionality takes on the same downside risk as buying though. So instead I find the best options strategy when these stocks are overbought to be a 2 by 1 call front spread.

For those not familiar let me provide an example. If a stock is at $100, you can buy one of the $100 strike calls, and sell two of the $130 strike calls, both on the same expiration. If the premium for the $100 strike call is double the premium on the $130 strike calls then this trade can be put on for no debit paid. The breakeven would be $160 and the profit zone would be between $100 and $160. As long as the stock stays below $160 prior to expiration then this trade is profitable.

The benefit of call front spreads in a high volatility environment is being able to capture the upside with no downside risk and no premium paid. The only risk is if the stocks go up past their breakeven price. While this is possible, I don’t find it likely after the stock has already had a large run-up and with the implied volatility so high it provides net options sellers with a lot of room for error.

Takeaway

The takeaway here is that I think that both the fundamental and valuation picture for AMD and Nvidia is very blurry, in that so much of the growth priced in is uncertain. Yet, on the flip side, if the secular growth trend is going to play out as many bulls think, AMD and Nvidia could be severely undervalued.

Because of this, I’m looking to play it similar to how George Soros would think about trading bubbles. This means using risk-mitigating strategies like trend-following and options strategies to capture the upside while limiting the downside.