Microsoft has halted data center projects in the United States and Europe that were set to consume two gigawatts of electricity, according to analysts from TD Cowen. This decision stems from an oversupply of capacity compared to Microsoft’s current demand forecast. The tech giant’s retreat from new leasing agreements is primarily influenced by its choice not to support additional workloads from OpenAI, the creator of ChatGPT. Investor concerns regarding the substantial investments in artificial intelligence by U.S. tech companies have grown due to slow returns and competition from the Chinese startup DeepSeek, which offers AI technology at significantly lower costs. Following Microsoft’s withdrawal, Google is reportedly stepping in to fill the void in international markets, while Meta Platforms is doing the same in the U.S. Despite this pullback, Microsoft reassured stakeholders that it plans to invest eighty billion dollars in AI infrastructure this fiscal year, maintaining that growth will continue across all regions.
Apple chips are set to be produced in the United States at a faster rate, according to TSMC. The company, which has faced delays in establishing its first chip plant, claims that future facilities will be built more rapidly, with the next plant expected to come online in two years. TSMC’s initial plant took five years to complete and primarily produces chips for older Apple devices. Despite this progress, former Intel CEO Pat Gelsinger remains skeptical, arguing that without research and development in the U.S., the country cannot achieve semiconductor leadership. TSMC’s U.S. expansion aims to address these concerns, with plans to produce three-nanometer chips by 2028 and two-nanometer chips before 2030.
Why do we care?
Microsoft’s halt on data center projects, primarily due to the decision not to support additional OpenAI workloads, hints at a more cautious approach toward AI infrastructure investment. This could signal a slowing momentum in cloud AI services growth, affecting clients who anticipated increased capacity for AI-driven workloads. The lack of rapid returns on AI investment and competition from cost-efficient alternatives like China’s DeepSeek reflect growing skepticism about the short-term profitability of large-scale AI infrastructure.
Despite halting specific projects, Microsoft is still committing substantial funds to AI infrastructure. In an interview with CNBC, Mustafa Suleyman, the CEO of Microsoft AI, explained the advantages of focusing on artificial intelligence models that are three to six months behind the latest developments. He emphasized that this approach allows for lower costs and a sharper focus on specific use cases. There’s wise words there.