In January, we saw the inauguration of a new administration, which has wasted no time in imprinting its agenda onto the government. We view some of the announced actions as positive for corporate profit growth (and therefore markets), while others are negative. At the same time, the $27 trillion US economy and the $63 trillion US equity market are to a fair degree independent of the federal government.
January was a healthy environment for both stocks and bonds, as economic data continued to come in stronger than expected, auguring healthy markets going forward. Below we review upcoming milestones in February, datapoints from January, and our view of initial Trump announcements and the late January DeepSeek surprise.
Key Data Points to Watch in February
- Corporate earnings continue. Corporate earnings for the fourth quarter of 2024 started being reported mid-January. Thus far, Q4 earnings growth has averaged a strong +12% year-over-year. The sectors with the strongest growth so far include Technology (+41%) and Financials (+28%). Looking forward, analysts expect 11.5% earnings growth in 2025 – a step up from 9% in 2024 – with YOY growth accelerating each quarter and ending 2025 at a 15% growth rate. If achieved, this should continue to support strong equity markets
- Jobs February 7. This will be the first jobs report under the Trump administration. Current expectations are for a gain of 148,000 nonfarm jobs created in January. This would reflect a modest deceleration from recent months; in the prior 6 months new jobs averaged about 160,000 each month. Consensus expects the unemployment rate to remain flat with the December level of 4.2%
- Inflation February 12 (CPI) and February 28 (PCE). The Federal Reserve (Fed) will be watching inflation data as one input to their decision on interest rates; a CPI or PCE report closer to 2% could give the Fed cover to resume lowering the Federal Funds Rate
January Takeaways:
- Fourth-quarter GDP of 2.7% was solid. The last several quarters, GDP has ranged between 1.6% and 4.4%; the Atlanta Fed GDPNow indicator is forecasting 3.9% for the first quarter of 2025
- Fed held rates steady. When the Fed met in late January, it held the short-term rate steady at a target range of 4.25% to 4.50%. Given that inflation has not yet declined to the Fed’s 2% target, we expect the Fed to continue to hold the rate steady for at least a couple more meetings
- Inflation is in line with expectations, but higher than Fed’s 2% target. Both the Consumer Price Index at +2.9% and the Core CPI (excluding food and energy) at +3.2% remained above the Fed’s target in December, which we believe will cause interest rates to stay higher for longer
- Bond yields modestly declined, driving bond indices higher. The Bloomberg Intermediate Government/Credit index appreciated by 0.6% during the month, while the Bloomberg High Yield index appreciated by 1.4%
- Stocks appreciated and the equity market broadened. The S&P 500 Index appreciated by 2.8%, and was outpaced by small-caps at +2.9% and international at 5.3%
Trump Actions Have Conflicting Economic Impacts
After President Trump’s inauguration on January 20, the new administration has issued a slew of executive orders and announcements, ranging from cutting the size of the federal government to pausing refugee programs. The actions that are most likely to impact the economy are tariffs, immigration, taxes, and regulations. Some that are likely to promote growth – and particularly corporate profit growth, which ultimately drives asset values – include lower taxes and less regulation. Others, including tariffs and limits on immigration, are likely to both increase inflation and lower corporate profits. Because the final shape of the orders, and the outcome of presumed litigation against them, is very uncertain, we cannot today draw firm conclusions about which of these impacts (higher corporate profit growth, or lower corporate profit growth and higher inflation) will have a larger impact. We will continue to carefully assess developments as they unfold.
How Deep Is the Impact of DeepSeek? Time Will Tell… |
By now, most investors have heard of DeepSeek, the artificial intelligence large language model (LLM) developed by a Chinese startup, which claims to offer models trained at a fraction of the cost of better-known models, like Meta’s LLaMA, Open AI’s o1-mini, and Google’s Gemini. The model’s performance in a variety of AI benchmarks show that its performance is comparable to better-known models. The DeepSeek white paper claims that the model was trained for $5 million, much less than other models. Both Meta and Google parent Alphabet intend to spend $60+ billion on capital expenditures in 2025, primarily on AI models; OpenAI has been spending over $5 billion per year; and the recently announced Stargate Project calls for spending $500 billion over four years.
Investors responded by focusing on a couple of questions. First, was this really possible without access to the export-controlled NVIDIA H100 chips, which are prohibited from being sold to China? Second, if LLMs can indeed be developed this cheaply, will demand for the massive AI datacenters dramatically decline, eviscerating demand not only for NVIDIA’s advanced chips, but also for other vendors to data center builds, and even the utility companies expected to provide those datacenters’ massive power requirements? Yes, NVIDIA stock fell by 17% on Monday Jan 27, wiping out $600 billion in market capitalization – the largest single-day loss for a US company in history – but industrial companies like Eaton declined by 16%, Quanta declined by 18%, and utility Vistra Energy declined by 28%. These companies had all been viewed as material beneficiaries of rapid datacenter buildouts, as evidenced by their stocks’ rapid increase over the past year: Eaton +22%, Quanta +51%, and Vistra +281%.
Why was the reaction so large?
Many of these stocks were trading at high price-to-earnings multiples. Prior to that day, NVIDIA was trading at 40x 2025 earnings (albeit only 26x 2026 earnings), Eaton was at 30x (5-year median 22), Quanta was at 35x (5-year median 21), and Vistra was at 27x (5-year median 12). When stocks are expensive, changing perceptions of market position can have a greater impact.
What does the market’s reaction say about the market?
Interestingly, despite several AI infrastructure-related stocks seeing dramatic declines, the overall market action was healthy. Both the Dow (+0.7%) and the equal-weighted S&P 500 (+0.1%) were up on the day. Six of the eleven S&P economic sectors appreciated, with even Consumer Discretionary and Financials up. Small- and mid-cap stocks both outperformed large-cap stocks, and a few software stocks like Meta and ServiceNow – both of which stand to benefit from cheaper AI services – also increased.
While the S&P 500 fell by 1.5%, a breadth indicator closely watched by Ned Davis Research (the NDR MultiCap Advance/Decline Line) rose – a rare combination. According to NDR, a positive Advance/Decline Line on the same day as a 1% S&P 500 decline has occurred only eleven times since 1980. Further, after such events, the S&P 500 is on average higher both one month later and one year later.
What does it mean for tech stocks, and for the economy?
Caveats abound about both the veracity and implications of the initial DeepSeek publication. Did it in fact have access to NVIDIA chips beyond what was claimed? Did the model also have access to OpenAI – evidence suggests that it did. Was the dollar investment claimed truly apples-to-apples?
While it’s premature to evaluate whether the low cost to train can be taken at face value, let’s assume for argument’s sake that it is order-of-magnitude correct. In that case, it is not a given that demand for NVIDIA’s chips would decline by the decline in cost to train. Indeed, basic economics tells us that the lower the cost of something, the higher demand is. Would demand increase enough to fully offset the material cost decline? Perhaps not immediately, but probably in less time than one might think. Consumers and businesses have a multi-decade history of using more memory, more storage, and more CPU power whenever those things get cheaper – just think about your computer or phone from 20, 10, or even five years ago compared to what you have today.
The advances that DeepSeek claims to have used, including using a Mixture-of-Experts (MoE) process and using Reinforcement Learning (RL) in the process, are likely to continue to move the field forward and will no doubt be adopted by existing players. Will the fact that DeepSeek is open-source reduce the competitive moats that have been assumed to be held by the large incumbent players? Perhaps.
Diversification and the value chain.
Going back to the market response, why were stocks like Meta, ServiceNow, and Salesforce actually up on the day? Our view is that as AI gets cheaper and more widely deployed, value capture will spread beyond the hardware makers like NVIDIA, Broadcom, and ASML, to the broader tech ecosystem, which includes not only the hyperscaler providers like Amazon Web Services and Microsoft Azure but software providers like ServiceNow and Salesforce. The pace at which this migration will occur may have been sped up (or not), but the direction is likely determined. Value that has been accruing to the chipmakers will broaden to other areas of technology. Capitalism has always been characterized by creative destruction, the precise nature of which is not always predictable. This is why diversification remains a core principle when investing.