The emergence of DeepSeek triggered violent turbulence in European and American stock markets, with Nvidia losing nearly $600 billion in market value in one day. Although the market has partially recovered, such huge fluctuations still cause investors to worry about AI, which is widely regarded as the next generation of transformative technology.
Vishwanath Tirupattur, global head of quantitative research at Morgan Stanley, believes that although DeepSeek’s breakthrough is significant, it will not lead to the collapse of AI-related capital expenditures.
Tirupattur said that DeepSeek's breakthrough, while impressive, was not entirely unexpected. The history of computer development is filled with cases of dramatic improvements in efficiency. DeepSeek's progress is essentially a significant efficiency improvement that will drive incremental demand.
The dramatic decline in computing costs in the 1990s provides a useful point of reference. Analysts at Morgan Stanley said the investment boom in the 1990s was driven by two factors: the speed at which companies replaced depreciating capital, and the continued sharp decline in the price of calculated capital relative to the price of output. If the efficiency gains brought about by DeepSeek reflect a similar phenomenon, it could mean that AI capital costs are falling, supporting corporate spending prospects.
Tirupattur also invokes the famous "Jevons Paradox". According to this theory, as technological progress reduces the cost of resource use, overall demand will increase, leading to an increase in total resource consumption. In other words, cheaper and more ubiquitous technology will increase its “consumption”, allowing AI to transition from the innovator stage to wider applications, opening the door to faster LLM product innovation and wider consumer and enterprise adoption. In the long term, this should lead to greater productivity gains and accelerate the realization of AI’s transformative promise.
Regarding the impact of DeepSeek on AI-related capital expenditures such as semiconductors, power and data center infrastructure, Joseph Moore, head of Morgan Stanley’s semiconductor team, believes that this development is unlikely to change semiconductor expenditures.
While the efficiency gains brought by DeepSeek are technically impressive, preliminary discussions with industry experts familiar with the technology indicate that it does not appear to have impacted the investment planning process.
Understanding the capital spending intentions of major technology players is critical to assessing whether these companies will significantly reduce AI infrastructure capital spending. Based on the information disclosed so far, the vast majority of known capital expenditures are for AI inference and non-AI use cases, not AI training.
Tirupattur concluded that from a macroeconomic perspective, there are good reasons to expect higher business spending and productivity growth. While there may be winners and losers at the stock level, the overall economy will be the winner. DeepSeek demonstrates the potential for efficiency gains, which in turn could foster greater competition and drive broader adoption of AI.