Fra Merril Lynch
What Should Investors Make of the DeepSeek Developments?
Investor assumptions about the dominance of U.S. technology leaders were challenged last week following the release of a new research paper by China’s AI startup DeepSeek. The company’s latest R1 large language model chatbot was able to deliver comparable performance with top U.S. firms in coding, reasoning and problem solving, despite reportedly being developed at a fraction of the cost, consuming much less power and using fewer processing resources. The news sparked a major Equity selloff in the U.S. Technology sector, particularly among upstream segments of the AI supply chain, including a record one-day market capitalization loss of close to $600 billion for a leading provider of high-end semiconductors. But does the DeepSeek moment spell an end for the AIdriven market advance of recent years? And how might its ripple effects change the broader outlook for global equity markets in 2025 and beyond?
It is notable that the biggest declines in last week’s volatility were concentrated among the providers of AI-enabling hardware, particularly those with the most extended valuations. This in our view reflects the immediate-term concern for investors that more efficient AI approaches could reduce demand for advanced chips and the specialized equipment used for their manufacturing. However, most individual constituents within the S&P 500 (including leaders in software and services) actually moved higher during the initial volatility of January 27. And this to us reflects what should be the larger consequence—an increase in global AI adoption, which is likely to be pulled forward by a reduced cost of implementation and a higher return on investment as developers around the world build on R1’s open-source model design. Ultimately this could make for greater expenditure on these enabling resources to keep up with demand. And while DeepSeek may represent a step gain in the efficiency of today’s AI models, further innovation will require ongoing infrastructure spending to deliver future breakthroughs in capability.
In this regard, it is instructive that a number of the U.S. hyperscalers reiterated their capital spending commitments on Q4 earnings calls last week. A deeper U.S.-China geopolitical divide could however come as a headwind for AI-driven gains in both markets. For China, tighter controls on access to U.S. and other Western hardware might eventually do more to hamper associated advances in local AI development. Despite the restrictions on cutting edge hardware, DeepSeek’s R1 was not developed totally independently of inputs from U.S. suppliers. Until China’s domestic chipmakers and semiconductor capital equipment makers can narrow the performance gap with Western suppliers, more effective controls could potentially pose a greater challenge for next-generation AI models.
And for the U.S., stronger export curbs could undermine revenues for firms with significant China exposure. We nonetheless expect the benefits of more widespread AI deployment to be felt across the broader global economy, beyond the leading technology enablers in any individual market. The result is likely to be higher productivity and labor saving in a range of sectors including manufacturing, agriculture, retail, healthcare and transportation, especially in the face of longer-term constraints from demographics and migration. Structural effects in terms of higher growth and lower inflation should come as fundamental supports for equity returns. And we would expect wider participation in the equity market advance as input costs decline and corporate profitability improves. These developments also in our view reinforce the case for investor diversification by industry, sector and region.
