Over the weekend, new open-sourced AI models were made public by Chinese company DeepSeek that appear to be cheaper to produce AI results equally as effective as OpenAi and Google’s Gemini. There are some apparently false headlines, such as just how cheaply DeepSeek’s R1 model was created, but it was created not using the latest (most expensive) Nvidia chips. We won’t go into the technical details, but they apparently used an innovative architecture and technical approach to gain efficiency. In short, this announcement is creating a knee-jerk backlash among AI stocks as it raises concerns about lower future demand for Nvidia’s chips and AI data centers.
We are still researching DeepSeek and assessing its actual implications in a clear minded way as is our approach. Our initial take is that the market reaction is overblown and likely reflective of already stretched valuations and investor crowdedness. We’d also note that their privacy policy contains some concerning language. DeepSeek collects information you share with Apple or Google if you sign in with that login. Given how much pushback TikTok has received in this country, we think this language could impede adoption.
We’d also point out Javon’s paradox, which says that innovations that unlock efficiencies ultimately creates increased demand. There are few examples in the world of technology where faster and cheaper computing didn’t mean better. Does this mean that Nvidia will need to lower the price of their Blackwell chips? Maybe, but that is normal in the semiconductor chip industry. Chip prices always decline over time.
AI is still in the very early stages of its development and usage. There is no question that the Total Addressable Market is enormous and there is lots of room for many participants. There will be competitors that emerge to challenge Nvidia chips. There will be multiple winners across the chip architecture and data center infrastructure. This is just the beginning. Here are just a few recent announcements from leading AI companies:
In short, AI is just at its beginning. DeepSeek’s successful optimization of AI reasoning models is additive to the industry’s development but is in the short run creating enormous uncertainty for AI stocks.
What Does This Mean for Portfolios?
To simplify this discussion let’s create four buckets of stocks: AI Chips, Hyperscalers, Date Center infrastructure suppliers and everything else.
AI Chips
Top on the list is Nvidia. The key question is does this DeepSeek reduce the demand for high-end GPUs (e.g., Blackwell and Ruben). It seems unlikely given the above capex announcements that anyone is cancelling their Blackwell orders. Remember, Blackwell chips are currently in a supply deficit. Nvidia can’t make enough fast enough to satiate demand from what we know. Could this impact the next generation? Maybe. However, as alluded to above, there has never been too much computing power. So, we will have to watch this situation evolve. We are significantly underweight NVDA.
Collateral damage is being experienced in other chip stocks (ASICs, custom chips, memory). Among these is Micron, which produces memory chips used in Nvidia GPUs. These powerful AI chips require more powerful memory chips. So, this stock faces some of the same concerns Nvidia is experiencing. We have minimal exposure to these names. This could be a great opportunity to add to these positions.
Hyperscalers
In this group we would include Google, Microsoft, Meta, Amazon and Oracle. These are the companies building huge Large Language inferencing models and, therefore, spending billions on chips and data center infrastructure. If this DeepSeek development pushes down costs, then this is a positive development. Importantly, no one is talking about there being less future demand for AI computations. So, the demand underpinnings for the hyperscalers’ services are not in question. In short, it may become cheaper to run AI models, which would be positive for these names. This is likely why these stocks are faring relatively better today.
Data Center Infrastructure
In this group are companies that make the products needed to build and operate AI data centers. Companies like Eaton and Vertiv produce the power and cooling systems that AI data centers need. The market is concerned that fewer AI data centers will be built or perhaps GPU innovation could lead to cooler chips that require less power. Even prior to the DeepSeek announcement, we thought this would happen. Again, if the compute power become cheaper and demand for more AI Reasoning grows, we ‘re not sure why that would translate to fewer data centers.
One important theme in our portfolio is Electrification. The power needed to run these AI data centers is enormous. Many of these AI data centers will be powered by natural gas. The big downward move in natural gas today and related stocks (like LNG and WMB) makes little sense if it specifically relates to the DeepSeek announcement.
Everything else
In short, the DeepSeek announcement will likely have very little impact. However, AI innovation is a net positive. The promise of AI is that it may deliver material productivity enhancements. That could lead to margin expansion and higher profitability. Nearly every company this earnings period is talking about how they are using and implementing AI. Some are developing new products and services. Some are enhancing existing ones. Others are simply able to improve customer service and worker productivity. AI innovation should benefit everyone else either through higher effectiveness or lower cost or both. In other words, these companies benefit from industry developments like this one.
The Bottom Line
The dramatic moves today in AI stocks are emblematic of crowded trades. The AI trade is very popular among professional and retail investors. When news flow emerges that raises potential concerns for popular stocks, air pockets can be created. Many AI stocks were already trading at robust multiples, which in most cases seemed justified given the power of AI. On days like today, panic wins. On days like tomorrow, calmer heads who understand business fundamentals pick through these technically battered stocks.
Disclosures
LIM is a Registered Investment Advisor based in Dallas, Texas and registered with the Securities and Exchange Commission. Registration does not imply a certain level of skills or training. LIM is a company with purpose, dedicated to creative and unique thinking. We focus on portfolio valuation and research, along with superior client experience. We seek to identify investment opportunities by looking at economic factors, security valuation and human behavior. We start with the fundamentals of portfolio management and valuation. Then we build on these fundamentals with unique thinking and creative intelligence gathering to form a viable investment thesis. We believe this approach leads to dynamic global portfolios with increased return and managed risk. LIM utilizes Charles Schwab & Co. Inc. (“Schwab”), a FINRA-registered broker-dealer, member SIPC, as its custodian of assets. LIM is independently owned and operated and not affiliated with Schwab.
Additional disclosures:
This document may contain forward-looking statements based on LIM’s expectations and projections about the methods by which it expects to invest. Those statements are sometimes indicated by words such as “expects,” “believes,” “will” and similar expressions. In addition, any statements that refer to expectations, projections or characterizations of future events or circumstances, including any underlying assumptions, are forward-looking statements. Such statements are not guaranties of future performance and are subject to certain risks, uncertainties and assumptions that are difficult to predict. Therefore, actual returns could differ materially and adversely from those expressed or implied in any forward-looking statements as a result of various factors.
This material represents an assessment of the market and economic environment at a specific point in time and is not intended to be a forecast of future events or a guarantee of future results.
Information is based on data gathered from what we believe are reliable sources. It is not guaranteed as to accuracy, does not purport to be complete and is not intended to be used as a primary basis for investment decisions.
This document is a general communication being provided to you for information purposes only. This communication is educational in nature and not designed to be a recommendation for any specific investment product, strategy, plan design feature or any other purpose. By receiving this communication, you agree with the intended purpose described above. Any examples used in this material are completely hypothetical and for illustration only. The document is for the sole use of the person to whom it is addressed and is privileged and confidential. Use by anyone other than the addressee is strictly prohibited.