The first quarter of 2025 has been marked by significant geopolitical events and the extraordinary rise of Chinese AI, particularly with the emergence of DeepSeek. This update explores the implications of AI developments on personal asset allocation.
DeepSeek Assessment
The rapid advancement of AI has led to an increase in speculation and misinformation in the media. According to Bridgewater Research, the current leading publicly available reasoning model is OpenAI’s o1, with OpenAI having announced but not yet released its o3 model. DeepSeek-R1, a model developed in China, has emerged as a formidable competitor, equaling OpenAI’s o1 on several mathematical and reasoning benchmarks while operating at a significantly lower cost.
Bridgewater analysts highlight that DeepSeek’s accomplishment is particularly impressive because it demonstrates how near-state-of-the-art models can be produced at a fraction of the cost of Western counterparts. However, there are concerns that DeepSeek’s success may be somewhat overstated due to potential overoptimization. Some experts believe the model may be overly tailored to specific benchmark tests rather than demonstrating superior real-world reasoning capabilities.
In terms of financial investment, DeepSeek’s final training run reportedly cost only $6 million. However, this figure does not include expenses related to data acquisition, preliminary research, model architecture experimentation, algorithm development, or personnel salaries. Industry experts estimate the total cost to be at least $100 million, which aligns with expectations given the model’s performance and timing. A notable point of speculation is whether DeepSeek had access to Nvidia’s H100 AI chip, which is currently restricted under US export controls. While DeepSeek claims it did not use the chip, many believe otherwise. Nonetheless, the efficiency with which DeepSeek utilized available computing resources is a significant technological breakthrough.
AI Investment Market Impact
The launch of DeepSeek-R1 has already had a substantial impact on the financial markets, triggering significant turbulence in the tech sector. Nvidia, a dominant player in AI chip production, saw its share price plummet by 17% in a single day following the announcement of DeepSeek’s capabilities. This decline reflects growing investor concerns over the security of Nvidia’s market dominance. DeepSeek’s success has demonstrated that AI development may shift focus toward efficiency gains in software optimization, rather than purely relying on expensive proprietary hardware.
The increasing viability of open-source AI models suggests that AI advancements may become more widely accessible, accelerating global AI development outside of leading research labs. If AI models become commoditized, key questions arise: Which companies will benefit from this shift, and which will suffer losses? Furthermore, will leading AI firms continue making substantial investments to push AI progress forward, or will the potential for rapid replication discourage innovation? If investment in model improvement slows down, it could place significant downward pressure on broader economic growth.
Ray Dalio and Bridgewater analysts have noted that AI remains a high-stakes race toward artificial general intelligence (AGI), where the outcome could be a “winner-takes-all” scenario. Meanwhile, billionaire investor Mark Cuban has pointed out that many major AI companies, such as Amazon and Google, would likely avoid going public if they were being founded today, making it difficult for retail investors to gain exposure to the AI sector.
Market Valuation and Bubble Concerns
Renowned investor Howard Marks argues that financial bubbles are often more psychological than quantitative. He describes a bubble as a state of mind where investors develop an unwavering conviction that “there’s no price too high” for certain stocks. Classic indicators of a market bubble include extreme investor enthusiasm, widespread adoration of certain companies, a fear of missing out (FOMO), and the belief that prices can only continue rising.
The US stock market may be approaching the final stage of a bull market, in which asset prices have appreciated dramatically, and investors assume that conditions will only improve indefinitely. Historical parallels can be drawn with previous market cycles, such as the Nifty Fifty stocks of the 1960s and 1970s, which traded at unsustainably high valuations before collapsing.
Currently, the S&P 500 trades at a price-to-earnings (P/E) ratio in the low 30s, well above the post-World War II average of 16. Although Nvidia has been a key driver of AI innovation, its valuation now appears to be reminiscent of past speculative bubbles. Additionally, the S&P 500 has experienced back-to-back years of 20%+ returns, an occurrence that has only happened four times in history. This suggests that current US stock valuations are higher than global averages, driven in part by enthusiasm for AI and passive index investing.
Historically, when the S&P 500 has traded at today’s P/E multiples of around 22, ten-year returns have ranged between +2% and -2%. Given these valuations, cautious investors should temper expectations for future real returns while recognizing that betting entirely against the US stock market is unwise, as Warren Buffett has long advised.
Investment Environment
The US economy has demonstrated unexpected resilience despite restrictive monetary policies. Strong real GDP growth, easing labor market conditions, and falling inflation suggest that a “soft landing” may have been achieved, but Vanguard’s economic outlook attributes much of this robustness to supply-side factors rather than Federal Reserve policy alone.
Productivity growth and an expanding labor force have helped drive economic strength, though emerging policy risks under the new US administration could introduce uncertainty. Ray Dalio has expressed concerns about the long-term financial stability of the United States, warning that many countries eventually experience debt crises. Even fiscally disciplined nations, such as Switzerland, have shown signs of unsustainable policies, such as expanding pension systems without adequate funding.
Meanwhile, the concentration of market power in a handful of large companies continues to raise concerns. The “Magnificent Seven” stocks—Apple, Microsoft, Alphabet, Amazon, Nvidia, Meta, and Tesla—now account for over 32% of the S&P 500’s total market capitalization. This level of concentration surpasses the previous record of 22% during the height of the dot-com bubble in 2000.
Outside the US, inflation has slowed significantly, though its path has been uneven across different regions. Europe has struggled with stagnation due to weak external demand, low productivity, and lingering effects of the energy crisis. In China, policymakers have implemented fiscal and monetary stimulus measures, though Vanguard remains skeptical of China’s long-term growth potential. Overall, Vanguard expects global monetary policy to remain tighter than in the 2010s, reinforcing a new era of positive real interest rates.
Consequently, we look for higher potential returns outside of the US, without betting against it. We access specific jurisdiction through ETF, and our broker of choice Interactive Brokers.
Sources: Berkshire Hathaway Shareholder Letters, February 22, 2025 Bridgewater Associates, "What China’s DeepSeek Means for AI," Greg Jensen, January 31, 2025 Howard Marks, "On Bubble Watch," Memos, January 7, 2025 Vanguard Economic and Market Outlook for 2025
Guide Finances 