Abstract:
Microsoft plans to more than triple its data center capacity, a move that is expected to help the company alleviate its computing power shortage. Previously, Microsoft had to refuse some artificial intelligence (AI) and cloud service businesses due to insufficient computing power. People familiar with the plans said Microsoft's data center network capacity around the world will exceed 38 gigawatts by 2032, up from the current level of about 12 gigawatts. By then, its capacity will exceed New York State's peak electricity demand.
These data give the outside world a glimpse of how much computing power Microsoft hopes to obtain after spending huge sums of money to build data centers. The company's capital expenditures reached $145 billion in its most recent fiscal year, and analysts expect that spending to continue to grow in the coming years.
The roadmap includes Microsoft-owned and leased data center facilities but does not include data centers acquired from CoreWeave Inc. Wait for the computing power rented by so-called new cloud service providers. They caution that plans are subject to change because data center construction takes years and changing customer preferences and the emergence of new technologies could overturn current assumptions.
Microsoft and its peers also have to contend with a backlash against data centers. Polls show that a majority of Americans oppose the construction of data centers in their areas.
Microsoft declined to comment on the specifics of its data center plans.
Microsoft's data center in Virginia
Informed sources revealed that of Microsoft’s current approximately 12 gigawatts of data center capacity, only approximately 2 gigawatts are dedicated to AI-specific chips. That share is expected to rise to about one-third of the 38 gigawatts of capacity Microsoft plans to bring online by 2032, they said.
Microsoft has limited new cloud service subscriptions in several key data center hubs such as the United States and Europe due to capacity shortages.
The key issue facing Microsoft is whether it can put enough data centers into operation in time to meet the accelerating demand for cloud computing and AI services.
Comments