THE first few weeks of 2025 have begun with some turbulence in global stock markets due to a number of disruptive events.
For example, Donald Trump’s new US administration is looking to roll out trade tariffs, raising concerns about inflation and global supply chains.
However, topping the list of disruptive events was the emergence of DeepSeek, a formerly little-known Chinese artificial intelligence (AI) startup.
The company unnerved the markets in late January with the release of a large language model that delivered performance comparable to well-known mainstream models but, crucially, at a significantly lower cost.
Market reaction and misconceptions
The market’s immediate reaction was one of concern, with several major players in the AI supply chain experiencing sharp declines.
This reaction seemed justified. If leading AI models can now be developed at a significantly lower cost, less investment may be needed than previously thought.
This reasoning appeared to drive market moves, as concerns arose that companies heavily investing in AI infrastructure might scale back their spending plans.
However, this argument overlooks a key dynamic, “elasticity” in AI spending. Although AI delivery costs have been falling significantly for years, companies have continued to ramp up investment in AI infrastructure.
Why? There are two primary reasons. Firstly, efficiency gains are reinvested. As costs decrease, companies reinvest those savings to build even more advanced AI systems.
And secondly, lower costs drive faster customer adoption. As AI becomes more affordable, accessibility increases, and companies are more willing to experiment, innovate and use AI, leading to greater overall demand for AI infrastructure.
AI investors not scaling back spending plans
While DeepSeek’s innovation is undoubtedly important, we believe it is unlikely to reduce overall AI infrastructure spending.
In the weeks following DeepSeek’s emergence, key investors in AI infrastructure, including Magnificent Seven companies such as Amazon, Microsoft, Google, and Meta, have reiterated, and in some cases even increased, their capital spending plans.
At this stage, it would seem that they see no reason to scale back investment.
This leads us to believe that DeepSeek has not and will not change the trajectory of AI investment. To be clear, AI remains a thematic investment with inherent risks.
However, these risks did not increase with DeepSeek’s emergence. The primary concern, whether AI demand will materialise at the scale expected by infrastructure investors, remains unchanged.
A key worry is the absence of definitive “killer” AI use cases. While this creates some uncertainty, we have already seen AI monetisation in areas such as advertising (Meta) and anticipate the emergence of new, unpredicted applications.
After all, few could have foreseen Uber arising from mobile internet adoption.
AI continues to be a powerful investment theme, but, given the fast-moving nature of the technology, careful monitoring of developments and a disciplined investment philosophy are essential. —Feb 24, 2025
Main image: The New York Times
