Microsoft has reportedly scaled back on data center projects in the U.S. and Europe, according to TD Cowen analyst Michael Elias. This decision is not surprising given the growing competition from Chinese DeepSeek, a more cost-effective and efficient alternative. The emergence of DeepSeek has raised concerns about the timeline for achieving AI data capacity.
In a recent report by Bloomberg, Microsoft is said to have halted construction of data centers in Indonesia, the UK, Australia, Illinois, North Dakota, and Wisconsin. The company has withdrawn from negotiations for leasing space in various locations, including London and Chicago, and has delayed construction in Jakarta and Mount Pleasant, Wisconsin.
Elias first raised concerns about Microsoft’s AI computing capacity in February, noting the termination of AI data center leases. He later reported that Microsoft had walked away from data center projects in the U.S. and Europe, totaling approximately 2 gigawatts of electricity capacity.
The news of Chinese DeepSeek’s efficiency compared to other large language models has prompted discussions about the necessity of current data center capacity. This development could lead to AI peak demand capacity being reached sooner than anticipated.
Goldman’s AI and power baskets have already seen changes in response to these developments. The China AI basket is outperforming the US AI baskets, indicating a shift in the market dynamics.
Overall, the evolving landscape of AI technology and data center projects suggests a need for reevaluating capex spending and preparing for potential revisions in the near future.