|
 | American Resilience Model Portfolio |
| Monthly Portfolio Review: January 2025Publication date: February 3, 2025 |
| | | Current portfolio holdings |
|  | | FOR SUBSCRIBER USE ONLY. DO NOT FORWARD OR SHARE. |
| | | The American Resilience portfolio produced a total return of 6.0% in January, compared with a total return of 2.8% for the S&P 500 Index. Long-term interest rates started to come down after the mid-month Consumer Price Index (CPI) report, which led to broad-based upside in stocks. President Trump’s announcement of Stargate, a $500 billion AI infrastructure project, briefly pushed up AI stocks, but this was abruptly undermined by news flow related to China’s DeepSeek. Performance this month was led by Air Products & Chemicals (APD), up 16%, after activists succeeded in their bid to change the composition of the board and replace the CEO. Shares of Thermo Fisher Scientific (TMO), up 15%, and Roper Technologies (ROP), up 11%, also performed well after reporting earnings. We offer additional perspective on the DeepSeek controversy. We continue to have an optimistic outlook for AI-related stocks, while emphasizing diversification across investment themes and sector exposures.
|
| | | The American Resilience portfolio generated a total return of 6.0% in January, which compares favorably to the S&P 500 Index return of 2.8%. For the six months ending January 31, 2025, the portfolio generated a total return of 10.5%, versus 10.1% for the S&P 500.
In January, the portfolio’s top performing stocks were Air Products & Chemicals (APD), which returned 16%; Thermo Fisher Scientific (TMO), which returned 15%; and Roper Technologies (ROP), which returned 11%.
The worst performing stocks in the portfolio were Eaton (ETN), which returned -2%, and Texas Instruments (TXN), which returned -1%. All other names in the portfolio generated positive returns in January. |
| Rate-driven recovery
As we noted in last month’s review, optimism in the wake of Trump’s victory gave way to concerns over rising interest rates. December was a difficult month for most stocks as long-term interest rates crept up.
Concerns around interest rates persisted through the first two weeks of the new year. The yield on the 10-year Treasury, which was approximately 4.2% at the start of December, reached a high of approximately 4.8% in mid-January.
The peak in bond yields more or less coincided with the S&P 500 Index, as well as the S&P 500 Equal Weight Index, reaching their lowest levels since Election Day. |
|  | Total Return (last 3 months) -S&P 500 and S&P 500 Equal Weighted |
| Relief came in the form of the Consumer Price Index (CPI) report, which was released by the Bureau of Labor Statistics on January 15, 2025. Yields fell and stocks rallied as core inflation readings came in lighter than expected.
By the end of January, yields on the 10-year Treasury were approaching 4.5%.
While investors fretted over rising interest rates in December and the first half of January, once long-term yields started to decline after the mid-month CPI report, market sentiment shifted back towards the themes of growth, innovation and optimism which fueled the November rally.
A busy month for AI
Shortly after Trump’s inauguration, he announced a blockbuster deal in the tech space. The $500 billion Stargate announcement, while light on specifics, signaled a strong commitment by the administration to the Artificial Intelligence (AI) build-out.
Most tech stocks, along with stocks in other sectors that are linked to the AI theme, rallied in the wake of this news. The renewed enthusiasm around AI would only last a few days, however.
Stargate was announced on Tuesday, January 21. By Sunday evening, January 26, a small Chinese AI start-up named DeepSeek became the main topic on every investor’s mind.
DeepSeek offers an AI model similar to OpenAI’s ChatGPT. Like ChatGPT, it can be downloaded as an app onto your phone (given data privacy concerns, we would not necessarily recommend this).
DeepSeek quickly shot to the top of the charts on the Apple app store. Meanwhile, media attention was directed towards a December 2024 technical paper which claimed that DeepSeek was created for a fraction of the cost of other AI models.
As we previously noted, the timing around the DeepSeek hysteria was curious, almost as if the press attention was intended to pour cold water on Trump’s bold initiative to cement U.S. technological dominance in AI.
On Monday, January 27, AI-related stocks tanked, including NVIDIA (NVDA). Just after the Stargate announcement, NVDA was trading very close to its all-time highest levels. NVDA ultimately traded down approximately 16% last week.
Thanks to the DeepSeek news flow, the market pivoted from conviction that an enormous investment boom was underway to create a vast infrastructure of AI data centers to something of a panic that all of this capital spending is pointless.
A broader market rally
As DeepSeek took over the news cycle, many AI-linked stocks were put under pressure. But with more optimism on the inflation outlook and interest rates coming down, we saw a rotation of capital into other areas of the market.
The Technology sector, dragged down by NVDA and other AI plays, ended up being the worst performing sector in the S&P 500 in January, although the damage was relatively mild (down 1%). |
|  | | Investors sold AI-related stocks and moved money into other growth areas that were not necessarily reliant on AI capital spending. Health care in particular performed well and delivered a 7% total return in January.
Cyclical sectors, like Industrials and Materials, which are especially sensitive to the interest rate trajectory, also performed well in January, reversing much of their dismal performance in December.
Positive momentum on the interest rate front was reinforced by the Federal Reserve meeting on January 29. The Fed’s statement, along with Fed Chair Jerome Powell’s press conference performance, painted a relatively benign picture of the economic outlook.
The Fed kept interest rates flat at the meeting. Powell described an economy that appeared healthy from a growth and employment perspective as inflation rates continue to moderate.
Our latest thinking on DeepSeek
The investment world spent much of the last week of January conducting a fire drill on DeepSeek and trying to understand the full implications. Over the course of the week, many important variables came to light.
We continue to have an optimistic long-term view of AI as an investment theme for stock market investors. There are four aspects to the DeepSeek story that should ease investor concerns over the potential negative impact.
(1) DeepSeek likely required much more money than initially assumed.
Initial reports on DeepSeek, based upon the December technical report, led investors to believe the Chinese had accomplished for a mere $6 million what their counterparts in the U.S. needed hundreds of millions if not billions to achieve.
As industry experts focused on the details of the DeepSeek business model, a consensus has emerged that a proper accounting of all the resources that went into DeepSeek would lead to a significantly higher figure than $6 million.
DeepSeek may have been created for meaningfully less than other AI models, suggesting the Chinese have perhaps discovered valuable process innovations or are leveraging lower human capital costs. But the gap between DeepSeek and U.S. peers is likely not as extreme as initially feared.
(2) DeepSeek copied ChatGPT.
It has also come to light that DeepSeek leaned heavily on the intellectual property of ChatGPT as it trained itself. As White House Crypto/AI Czar David Sacks explained in a recent interview, DeepSeek may have had millions of interactions with the ChatGPT model (which were possibly in violation of OpenAI’s intellectual property rights).
Sacks noted that DeepSeek was so thoroughly influenced by ChatGPT that users of the DeepSeek program observed that it even self-identified as ChatGPT at times.
(3) The bulk of the multi-billion dollar AI build-out is not for training but inference.
DeepSeek was allegedly able to develop an AI training model relatively inexpensively, but the next chapter of AI relates more to inference.
Training refers to the process of the computer network learning how to answer questions and create graphical content. Inference refers to the process of actually answering these questions and delivering this content. |
| | Major misunderstanding about AI infrastructure investments: Much of those billions are going into infrastructure for *inference*, not training. Running AI assistant services for billions of people requires a lot of compute. Once you put video understanding, reasoning, large-scale memory, and other capabilities in AI systems, inference costs are going to increase. - Yann LeCun, Chief AI Scientist at Meta (1/28/2025) |
| | The planned investment of hundreds of billions of dollars into AI data centers is all about mass market delivery of AI services, which remains very resource-intensive. |
| (4) Cheaper AI means more AI.
There is another possible misinterpretation of the DeepSeek story. Investors sold AI stocks based on the assumption that getting greater productivity out of hardware like NVIDIA Graphic Processing Units (GPUs) is a negative signal for AI capital spending.
We are in the early stages of the AI revolution, but if it follows the patterns of other technology roll-outs, like the Internet and mobile phones, the costs associated with delivering this technology should continuously decline. And declining costs should incentivize greater usage, creating even more demand.
Investors should therefore continue to expect technological advancements that generate more artificial intelligence at lower cost. With no apparent ceiling on the AI services that business and consumers will ultimately want to have, declining costs may actually spur more investment in AI, rather than less.
Diversification still makes sense
Artificial intelligence remains a powerful force within the economy, but the DeepSeek shock demonstrates that even growth-oriented investors should not be singularly focused on any particular trend. Investors who had exposure to other growth areas were rewarded in January.
Further underscoring the need for diversification, as we write, market focus is shifting towards Trump’s recently announced tariffs on imports from Mexico and Canada.
The implementation of these tariffs led to some pressure on markets at the end of the day on Friday. We plan to provide further color on this subject as trading resumes this week. |
| | | The top performers in the portfolio in January were Air Products & Chemicals (APD), which returned 16%; Thermo Fisher Scientific (TMO), which returned 15%; and Roper Technologies (ROP), which returned 11%.
The worst performing stocks in the portfolio were Eaton (ETN), which returned -2%, and Texas Instruments (TXN), which returned -1%. The remaining stocks in the portfolio produced a positive return. |
| APD shares advanced following the company’s January 22 annual shareholder meeting. Of the four candidates nominated by activist hedge fund Mantle Ridge to the board of directors, three were elected.
CEO Seifi Ghasemi and his two allies in the proxy contest failed to win election to the board. It is now widely expected that Mr. Ghasemi will step down and be replaced by Mantle Ridge’s preferred candidate, Eduardo Menezes.
Mr. Ghasemi had championed a clean energy strategy that was both capital intensive and high risk. His refusal to back away from this approach invited pushback from shareholders, depressed the company’s valuation, and created a window of opportunity for Mantle Ridge’s Paul Hilal.
Thanks to Mantle Ridge’s activism, the company is expected to implement many of the changes proposed by Mantle Ridge and other investors. These generally involve a return to the more predictable, and less capital intensive, industrial gas model. A large cost reduction program is also anticipated.
The change in the strategic direction at APD led to a number of upgrades and improved long-term earnings estimates by analysts who cover the stock, which helped drive upside in the shares.
Shares of TMO, one of the world’s leading providers of instrumentation and analytical equipment and services in the health care sector, advanced after a strong fourth quarter earnings report on January 29, which showed an acceleration of organic growth.
While companies involved in the AI build-out have captured the imagination of growth-oriented investors over the past year or two, health care is another area that is also in many ways linked to AI innovation as a downstream beneficiary of the technology.
As TMO’s CEO Marc Casper said on last week’s earnings call, “Our colleagues around the world are actively deploying Generative AI, finding new ways to improve the customer experience, streamline internal processes and enhance our products and services.”
ROP is a tech stock play, but is similar to TMO in that it leverages AI technology, rather than creates it.
Like TMO, ROP also delivered impressive fourth quarter earnings results in January. ROP posted 16% free cash flow growth for the year and indicated an encouraging outlook for 2025.
ROP offers a suite of mission critical software tools across a wide array of niche markets. The company is increasingly deploying generative AI throughout its business, which leads to incremental growth opportunities and stickier customer relationships.
Shares of ETN, a leading industrial supplier of electrical components, traded down modestly in January.
As a very direct beneficiary of AI spending, shares of ETN were recovering nicely from December weakness on the heels of Trump’s Stargate announcement. The uncertainty created by DeepSeek knocked the stock back down, however.
We continue to view ETN as a compelling long-term play on the AI-driven electrification theme. We understand the market’s concerns around DeepSeek, which may linger for a while, but view the current pressure on the share price as a buying opportunity.
In contrast with many semiconductor names, shares of analog chip maker TXN declined only modestly in January.
As we explained in our recent note, TXN does not make chips that are used to produce AI. Rather, demand for its chips increases as AI applications get embedded in vehicles, industrial equipment and even robots.
The cheaper and more abundant AI becomes, the more need there will for analog chips that are required by devices (like sensors and power management systems) that are controlled by AI systems. |
| | |  | | |  | | |  | | | Oracle Corporation (ORCL) |
|  |  | | |  |  | | |  |  | | |  |  | | |  |  | | Arch Capital Group (ACGL) |
|  |  | | Air Products & Chemicals (APD) |
|  |  | |  |  | |  |  | | |  |  | | Thermo Fisher Scientific (TMO) |
|  |  | | |  |  | | |  |  | | |  |  | | |  |  | | The 76research American Resilience Model Portfolio is designed to provide exposure to businesses that operate with competitive advantages in structurally attractive markets. The objective is to identify businesses that can survive and thrive across different macroeconomic environments and whatever geopolitical crises may unfold. The holdings are intended as long-term investments to drive portfolio compounding with minimal need to realize taxable gains. Emphasis is placed on critical markers of business quality such as barriers to entry, physical scarcity of assets, balance sheet strength, effective capital allocation and durable long-term growth drivers. These assessments are paired with careful consideration of valuation and risk. |
| | FOR SUBSCRIBER USE ONLY. DO NOT FORWARD OR SHARE. |
| | | |  |
|
|
|
| |
|
