DeepSeek & AI: Evolution, Not Disruption
Key Takeaways
- DeepSeek’s model is unlikely to replace industry leaders but represents an expected evolution in AI development
- AI chip demand will remain strong due to the Jevons paradox—efficiency gains often drive higher usage
- US AI stocks are expensive, and better AI-related opportunities may exist in emerging market
Your OCIO office has naturally been following this story with great interest; it touches not only US stocks in this space, but also China and the rest of emerging markets, which are full of stocks exposed to the AI theme—and not just chipmakers, but many others in the broader supply chain including those making components for data centers. We want to share our initial thoughts with you here. As an advisor that runs strategies in both developed and emerging markets, we offer what I hope you’ll consider a balanced perspective on this hot topic.
First, based on the process by which DeepSeek built and trained its model—while they undoubtedly did a number of clever and very cool things to achieve the efficiencies they’re reporting—their approach has been hyped a little too much, in our view. For one thing, there are legitimate questions as to how low training costs truly were and how much they leaned on incumbents like OpenAI to develop their own model. It seems very unlikely to us that their model will supplant those developed by OpenAI and others in the space. Still, expecting models like DeepSeek’s to proliferate over time seems very reasonable. That’s actually how technological revolutions like this usually play out, with first-movers overcoming the big, expensive obstacles and smaller, scrappier players eventually riding their coattails to their own breakthroughs.
Along those lines, however, there’s an important corollary. Suppose you’ve been reading about DeepSeek in the media. In that case, you’ve probably seen references to the ‘Jevons paradox’: the notion that as a resource becomes more efficient to use, demand for that resource may actually increase. This is the right way to think about how AI usage will play out in the face of DeepSeek and similar advances that inevitably hit the market. In short, it’s early innings in terms of how AI will be utilized, and the demand for chips will not majorly weaken. Fundamentally, we don’t believe DeepSeek is a threat to the AI theme.
But another angle to this story is also worth considering: Is DeepSeek a threat to AI theme stocks? Our Chief Research Officer, Dr. Phil Wool, spoke with Reuters recently about that, and they kindly published some of his comments here. As mentioned, DeepSeek will not fundamentally hurt the AI supply chain in the long run. Still, we believe this saga that unfolded over the last week has further exposed how expensive US tech has become—i.e., how much of that fundamental growth is already priced in. But this has also been obvious to anyone watching Mag 7 earnings over the last few quarters, where companies often ‘beat’ and the market is still disappointed because expectations have gotten so high. Investors are becoming increasingly skeptical about the crazy AI capex these companies are doing, which is continuing indefinitely.
So, it’s clear there was a massive amount of growth being baked into the price of companies like Nvidia and ASML, which means there’s little margin for error when any potential threat to those assumptions emerges (and DeepSeek isn’t the only threat—US containment of China’s chip demand is another big one). I think the upshot for investors is that it pays to be active, do research, and be more selective in this theme. As a firm that runs both US and EM strategies, we truly believe there are much better opportunities within the emerging markets to buy stocks at much better valuations that will continue growing as the data center buildout continues unabated. We expect EM to outperform DM over the next decade, and this is one of the channels through which that happens: relatively cheap EM tech catches up with its relatively expensive DM counterpart.
Advisory services offered through Sowell Management, a Registered Investment Advisor. The views expressed represent the opinion of Sowell Management. The views are subject to change and are not intended as a forecast or guarantee of future results. This material is for informational purposes only. It does not constitute investment advice and is not intended as an endorsement of any specific investment. Stated information is derived from proprietary and non-proprietary sources that have not been independently verified for accuracy or completeness. While Sowell Management believes the information to be accurate and reliable, we do not claim or have responsibility for its completeness, accuracy, or reliability. Statements of future expectations, estimates, projections, and other forward-looking statements are based on available information and Sowell Management’s view as of the time of these statements. Accordingly, such statements are inherently speculative as they are based on assumptions that may involve known and unknown risks and uncertainties. Actual results, performance or events may differ materially from those expressed or implied in such statements. Investing in securities involves risks, including the potential loss of principal. While equities may offer the potential for greater long-term growth than most debt securities, they generally have higher volatility. International investments may involve risk of capital loss from unfavorable fluctuation in currency values, from differences in generally accepted accounting principles, or from economic or political instability in other nations. Past performance is not indicative of future results.
DeepSeek & AI: Evolution, Not Disruption
DeepSeek & AI: Evolution, Not Disruption
Key Takeaways
Your OCIO office has naturally been following this story with great interest; it touches not only US stocks in this space, but also China and the rest of emerging markets, which are full of stocks exposed to the AI theme—and not just chipmakers, but many others in the broader supply chain including those making components for data centers. We want to share our initial thoughts with you here. As an advisor that runs strategies in both developed and emerging markets, we offer what I hope you’ll consider a balanced perspective on this hot topic.
First, based on the process by which DeepSeek built and trained its model—while they undoubtedly did a number of clever and very cool things to achieve the efficiencies they’re reporting—their approach has been hyped a little too much, in our view. For one thing, there are legitimate questions as to how low training costs truly were and how much they leaned on incumbents like OpenAI to develop their own model. It seems very unlikely to us that their model will supplant those developed by OpenAI and others in the space. Still, expecting models like DeepSeek’s to proliferate over time seems very reasonable. That’s actually how technological revolutions like this usually play out, with first-movers overcoming the big, expensive obstacles and smaller, scrappier players eventually riding their coattails to their own breakthroughs.
But another angle to this story is also worth considering: Is DeepSeek a threat to AI theme stocks? Our Chief Research Officer, Dr. Phil Wool, spoke with Reuters recently about that, and they kindly published some of his comments here. As mentioned, DeepSeek will not fundamentally hurt the AI supply chain in the long run. Still, we believe this saga that unfolded over the last week has further exposed how expensive US tech has become—i.e., how much of that fundamental growth is already priced in. But this has also been obvious to anyone watching Mag 7 earnings over the last few quarters, where companies often ‘beat’ and the market is still disappointed because expectations have gotten so high. Investors are becoming increasingly skeptical about the crazy AI capex these companies are doing, which is continuing indefinitely.
So, it’s clear there was a massive amount of growth being baked into the price of companies like Nvidia and ASML, which means there’s little margin for error when any potential threat to those assumptions emerges (and DeepSeek isn’t the only threat—US containment of China’s chip demand is another big one). I think the upshot for investors is that it pays to be active, do research, and be more selective in this theme. As a firm that runs both US and EM strategies, we truly believe there are much better opportunities within the emerging markets to buy stocks at much better valuations that will continue growing as the data center buildout continues unabated. We expect EM to outperform DM over the next decade, and this is one of the channels through which that happens: relatively cheap EM tech catches up with its relatively expensive DM counterpart.
Advisory services offered through Sowell Management, a Registered Investment Advisor. The views expressed represent the opinion of Sowell Management. The views are subject to change and are not intended as a forecast or guarantee of future results. This material is for informational purposes only. It does not constitute investment advice and is not intended as an endorsement of any specific investment. Stated information is derived from proprietary and non-proprietary sources that have not been independently verified for accuracy or completeness. While Sowell Management believes the information to be accurate and reliable, we do not claim or have responsibility for its completeness, accuracy, or reliability. Statements of future expectations, estimates, projections, and other forward-looking statements are based on available information and Sowell Management’s view as of the time of these statements. Accordingly, such statements are inherently speculative as they are based on assumptions that may involve known and unknown risks and uncertainties. Actual results, performance or events may differ materially from those expressed or implied in such statements. Investing in securities involves risks, including the potential loss of principal. While equities may offer the potential for greater long-term growth than most debt securities, they generally have higher volatility. International investments may involve risk of capital loss from unfavorable fluctuation in currency values, from differences in generally accepted accounting principles, or from economic or political instability in other nations. Past performance is not indicative of future results.