DeepSeek won’t cause the Fed to tighten monetary policy. It won’t cause a financial crisis or a credit crunch. Of course, it might have a negative wealth effect on Nvidia’s (NASDAQ:NVDA) shareholders.
Chinese firm DeepSeek has taken the evolution of AI to a new level with its cheaper Language Learning Model (LLM). This is not a “black swan” event for the stock market, but it is a “gray swan,” with potential positives and negatives.
Although DeepSeek disrupts the AI status quo, it should speed up the proliferation of AI and the realization of associated productivity gains. DeepSeek also could affect the U.S. Federal Reserve’s considerations.
Beyond this, the U.S. stock market’s historically high valuation doesn’t worry us. Even if the Mag-7’s P/Es take a hit owing to DeepSeek, we expect that the P/Es of the other 493 stocks in the S&P 500 could go higher as earnings growth should support valuations.
The digital revolution is evolving: DeepSeek is a Chinese AI lab that has rocked the world of artificial intelligence (AI) by developing a competitive LLM that reportedly outperforms ChatGPT but was developed at a fraction of the cost, with much less time required to “teach” the program. It also functions with cheaper and less powerful Nvidia (NVDA +8.93%) GPU chips. And it is available on an open-source basis.
Will DeepSeek cause a bear market?
Since the late 1920s, there have been 22 bear markets in the S&P 500. Over the same period, there have been 17 recessions. In other words, more often than not, bear markets are caused by recessions. And more often than not, bear markets and recessions are caused by the tightening of monetary policy.
DeepSeek won’t cause the Fed to tighten monetary policy. It won’t cause a financial crisis or a credit crunch. It won’t cause a recession even if it causes American AI companies to reduce their capital spending on AI infrastructure. Of course, it might have a negative wealth effect on Nvidia’s shareholders.
We view DeepSeek as a “gray swan” event, a spinoff from the term “black swan” event. Black swan events are unexpected developments with mostly negative consequences. They’ve been known to cause recessions and bear markets, but not all black swans have had negative consequences. Likewise, gray swans are unexpected events, but they have both negative and positive consequences.
The negative consequence of DeepSeek is that it challenges the business models of American companies that expected to use their exclusive access to Nvidia’s most expensive and powerful chips to dominate and profit from the AI revolution. The good news is that they should be able to follow DeepSeek’s lead in lowering the cost of AI infrastructure spending.
That should offset some of the potential revenues lost by having to compete with DeepSeek and other AI startups. They will still profit from AI by converting more of their pre-AI products into AI-driven ones. Also good: More competition in the AI economy will give business and individual consumers more bang for their AI bucks.
The digital revolution is all about data processing, i.e., processing more data faster and faster at lower and lower costs. From this perspective, AI is an evolutionary development in the digital revolution. So much data can be processed that we need LLMs to make some sense of it all and use it to increase productivity.
Making Sense of the Market
Will all the AI hype that has sent the Magnificent Seven’s valuations soaring prove to be a dot-com bubble 2.0?
Could investors’ negative reactions to DeepSeek be linked to the S&P 500’s valuation? We’ve been fielding concerns that the stock market is overvalued for at least the past two years. Indeed, valuation multiples tend to ride the escalator up during economic expansions and the elevator down during recessions. Rising bond yields have raised the question of whether the S&P 500’s forward P/E ratio has peaked at its current level of 22, suggesting a turning point for U.S. stocks.
We haven’t been very concerned by the stock market’s valuation, and we still are not. We believe the relationship between the S&P 500 forward earnings yield and the 10-year Treasury yield (TMUBMUSD10Y 4.526%) is returning to its historical norm, as opposed to much of the 21st century, when the Federal Reserve depressed rates and bought bonds. Strong expected earnings growth is propelling valuations to justifiable levels. Furthermore, any decrease in the “Magnificent-Seven’s” P/Es may be made up by rising valuations in the rest of the stock market, i.e., the S&P 493.
Consider the following:
- (1) Fed’s stock valuation model
From 1985 through 2000, the S&P 500 forward earnings yield and the 10-year Treasury note yield tracked one another relatively closely. They diverged through 2020, mostly because the bond yield fell relative to the forward earnings yield. Bond prices were boosted by the Fed’s low interest rates and quantitative easing policies during most of that period. The two yields are now equal for the first time since 2022, meaning it is more likely that stocks are fairly valued relative to bonds than overvalued.
- (2) Valuation versus earnings
Valuations typically fall when something in the financial markets breaks because of the Fed’s tightening monetary policy, leading to a credit crunch in the real economy. The U.S. economy skirted that scenario during the recent round of tightening. But the advent of DeepSeek raises the question of whether the large sums being dished out by the Magnificent Seven on data centers, semiconductor chips, and AI models can generate an adequate return on investment. Will all the AI hype that has sent Magnificent Seven valuations soaring prove to be a dot-com bubble 2.0?
The profit margins of the Magnificent Seven will likely benefit from lower costs as they use AI more efficiently and cost effectively internally.
We believe that earnings growth will be the primary driver of stock-market returns through the end of the decade. Analysts currently expect S&P 500 companies to grow their earnings at an annual rate of 17.9% over the next five years. While not dissimilar from what was expected in 2000, it is also at a cheaper multiple and in line with prepandemic expectations. Investors tend to pay higher prices for stocks with sustained high earnings growth.
The profit margins of the Magnificent Seven will likely benefit from lower costs as they use AI more efficiently and effectively internally, which may net out losses on large capital expenditures. Even if the Magnificent Seven’s collective forward profit margin falls from its current 25.5%, the rest of the S&P 500 will likely benefit from cheaper AI tools and more productive employees, boosting their collective forward profit margin from the current 12.0%. That could help maintain the overall market’s valuation, as the forward P/E of the S&P 493 would likely rise even if the valuation of the Magnificent Seven falls.
If workers can become more productive faster and companies can cut costs quicker, then the Fed has a huge disinflationary tailwind to consider.
- (3) The Fed effect
Will the Fed officials care about the DeepSeek-triggered stock market hullabaloo? Perhaps. On one hand, if stock prices fall enough, they could be concerned about a declining wealth effect weighing on consumer spending. We don’t believe this is likely, but it is possible.
The Fed may be more interested in how the vastly cheaper AI model will help permeate AI throughout the economy. If workers can become more productive faster and companies can cut costs quicker, then the Fed has a huge disinflationary tailwind to consider. Fed officials may even say that it means the neutral rate of interest is lower, emboldening them to cut the federal-funds rate further.
We, on the other hand, believe that AI-boosted productivity growth would boost real GDP growth and keep inflation subdued. This would imply that interest rates don’t need to be lowered. If the Fed lowers interest rates in this scenario, it risks causing a speculative bubble in risk assets.
So the outlook is either the Roaring 2020s scenario (to which we ascribe 55% subjective odds) or a market meltup a la the 1990s (25% odds). We don’t see the DeepSeek development as a reason to increase our odds of a Stagflationary 1970s scenario (20% odds). At worst, it is a gray swan, not a black one.