76report

January 27, 2025

China’s DeepSeek Rattles Markets: Deep Trouble or Opportunity?

Many AI-related stocks came under intense pressure today. Investors reacted negatively to recent reports that a Chinese start-up called DeepSeek has been able to develop an AI model very inexpensively.

Among the biggest casualties in today’s trading was AI bellwether NVIDIA (NVDA), which declined 17%. The tech heavy Nasdaq Composite Index was down approximately 3%.

More than a dozen other stocks within the S&P 500 Index saw double digit declines, which led to an approximate 1.5% decline in the index. It is interesting to note, however, that a majority of stocks within the S&P 500 in fact traded higher today.

The market action today may be unsettling, but we encourage investors to consider the following key points:

(1) We do not have a perfect understanding of what resources have gone into DeepSeek. Take reports out of China with a grain of salt.

(2) It is not time to give up on AI. The possibility of a productivity breakthrough in AI has both positive and negative implications for stocks, even those directly tied to the AI build-out.

(3) Always be diversified. Today’s action provides a healthy reminder of the need to stay diversified across industry sectors and investment themes.  

What we know about DeepSeek

DeepSeek is a Hangzhou, China based AI start-up that was formed in 2023. In a recent technical paper, engineers at the company claimed that, for a mere $6 million, they were able to create an open-source AI model that is competitive with products like OpenAI’s ChatGPT.

The backstory here is that China has faced legal restrictions on importing NVIDIA Graphic Processing Units (GPUs). As a result, firms like DeepSeek have had to figure out process changes that allow them to do more with less.

The accuracy of the $6 million figure—and whether or not it properly accounts for all resources that went into the project—is impossible to confirm.

It is widely acknowledged, however, that DeepSeek was able to achieve a level of performance that competitors like OpenAI needed considerably greater investment to achieve.

The timing of the news from DeepSeek is interesting in that it follows Trump’s announcement, shortly after getting inaugurated, of the $500 billion Stargate AI infrastructure project, which we discussed last week.

To put this in context, just days after the President announced that the U.S. will be pursuing an aggressive investment campaign to achieve technological dominance in AI, we have a response from a Chinese firm indicating that they can create AI capabilities for a fraction of the cost.

Given that China is the only meaningful challenger to the U.S. in this competition over AI supremacy, there is reason to interpret the information coming out of DeepSeek through a political lens.

Many observers, including Elon Musk (who is also investing heavily in AI through his Grok initiative), have expressed doubts as to whether the amount of money and hardware allocated to DeepSeek’s model was quite as described.

Why did stocks sell off?

The operating assumption of markets has been that vast amounts of capital will be deployed by the large tech platforms, often referred to as hyperscalers, towards the development of AI capabilities.

By far, the largest impact on the major stock indices today came from NVDA. NVDA dominates the market for AI GPUs, which are the key technological commodity that hyperscalers must purchase to create their AI data centers.

NVDA is of course one of the most successful stock market stories of all time. Just a few weeks ago, NVDA shares closed at an all-time high of $149 per share.

NVDA has benefited from the perception that it will continue to see extremely strong demand, especially as Trump policies focus on accelerating AI development.

The DeepSeek news has now prompted investors to re-evaluate just how necessary it will be for the tech sector to pour hundreds of billions of dollars into AI chips and related expenditures. The sharp decline in NVDA suggests the market is now pricing in a diminished long-term outlook for GPU sales.

Other stocks that stand to benefit from a massive AI build-out traded down as well. Like NVDA, many of these have been extremely high performers and have recently benefited from some giddiness around the Trump AI push.

The worst performing stock in the S&P 500 today was actually Vistra (VST), which traded down some 28%. VST is an electrical power generation company that is expected to be a major beneficiary of AI data center demand.

It is worth noting that, even with today’s sell-off, VST is only slightly negative on a year-to-date basis. Before today, the stock had rallied some 39%—just in the month of January alone.

Not all stocks were weak today. In fact, about two-thirds of all S&P 500 stocks were up on the day. Many benefited from the rotation of capital away from high-flying AI names to unrelated industries.

Is the AI party over?

The big losers today were companies that were perceived beneficiaries of AI capital spending.

Investors have rapidly transitioned from great optimism towards this theme towards a sense that all of this spending is pointless. The market is now implicitly pricing in reduced AI investment—but the actual impact of the DeepSeek revelation is unclear.

The main premise of bears on AI stocks seems to be that there will be less need to invest heavily in AI chips and related infrastructure because it has just been proven that AI can be manufactured more efficiently.

Yet this runs counter to historical patterns in demand for technology—where if a product offers more value for less money, demand actually rises.

Microsoft (MSFT) is one of the most important AI hyperscalers. MSFT CEO Satya Nadella basically made this point in a social media post today on X.

Jevons paradox strikes again! As AI gets more efficient and accessible, we will see its use skyrocket, turning it into a commodity we just can't get enough of. - Satya Nadella (1/27/2025)

Nadella included a Wikipedia link to Jevons paradox, “which occurs when technological advancements make a resource more efficient to use, thereby reducing the amount needed for a single application.”

Energy is one of the most prominent examples of Jevons paradox (named after William Stanley Jevons, a 19th century British economist). The cheaper energy gets, the more energy tends to get consumed.

Nadella is implying that as AI gets less expensive to produce, more of it will be consumed. A similar observation can be applied to internet bandwidth—the more high-speed internet access that consumers are offered at low cost, the more they want.

We are not surprised by the extremely negative reaction to the DeepSeek news, especially in the context of very strong recent performance and high embedded expectations for stocks in the AI theme.

We are not convinced, however, that capital spending plans will be substantially scaled back.

Major AI players will likely try to learn from DeepSeek how they can reconfigure their investment plans to get more bang for their buck. But the total number of bucks they intend to allocate may not get reduced.

There has always been an expectation of productivity gains along the AI journey—technology is constantly becoming better, cheaper and more effective.

If it now takes less money to create the same quantity of AI, it can be argued that there is even more reason to allocate money to AI.

Stay diversified

The DeepSeek sell-off serves to remind investors of the need to maintain diversification. The AI theme has been a powerful force in markets in recent years and has to some extent distorted markets.

The top 10 holdings of the market-cap weighted S&P 500 now represent 35% to 40% of the index. With the exception of Warren Buffett’s Berkshire Hathaway (BRK), these are essentially all technology platform companies that have benefited, in some cases substantially, from AI-related enthusiasm.

Within our Model Portfolios, we have focused on differentiated investment ideas and avoided mega-cap stocks like NVDA, as many investors already have substantial exposure to these names.

We do have some exposure to AI-related names that we are optimistic about. At the same time, we have exposure to a variety of other investment themes and industries.

Even within technology, not every stock is an AI build-out play. In some cases, tech stocks can be beneficiaries of deflationary forces within AI.

TXN: An AI deflation beneficiary

Texas Instruments (TXN), for example, is a semiconductor stock that traded up today. We hold TXN within both the American Resilience and Income Builder portfolios (3% dividend yield).

As we described in an earlier note, TXN is a supplier of analog chips, which consist of very basic and inexpensive technology (used in sensors, for example).

TXN’s products, which may only cost a few dollars per unit, are at the opposite end of the spectrum of NVDA’s GPUs, which can be tens of thousands of dollars each.

Analog chip makers like TXN stand to benefit from declining AI costs because it is the proliferation of AI—in vehicles, robotics, industrial equipment—that drives demand for analog chips.

The cheaper and easier it is to deploy AI, the more demand there will be for TXN’s products.

The long-term opportunity

Technology moves in fits and starts. Tech stocks can be quite volatile. Yet technology stocks have also been an extraordinary driver of stock market returns for decades.

DeepSeek has challenged widely held assumptions about AI and has emerged as an unexpected source of uncertainty, which the market never likes.

Investors should not have all their eggs in the AI basket, but technological innovation and strong policy support from the new administration are a powerful combination.

We continue to see strong potential for AI-related development to drive growth across the economy.

Click HERE to learn more about our Model Portfolio subscription plans.

Disclaimer

FOR SUBSCRIBER USE ONLY. DO NOT FORWARD OR SHARE.