AI computing power demand is insane, Microsoft's "New Cloud" transaction volume exceeds 60 billion USD

Microsoft has committed over $60 billion to "neocloud" data center companies, indicating that the company is accelerating its search for sufficient computing resources to meet its artificial intelligence computing power needs. Since early October, self-media statistics show that Microsoft's investment commitments to such companies have nearly doubled.

According to media reports citing informed sources, the largest single expenditure is approximately $23 billion, directed towards the UK startup Nscale. Under the arrangement with Nscale, Microsoft will gain access to approximately 200,000 of NVIDIA's latest GB300 chips deployed in data centers in the UK, Norway, Portugal, and Texas.

Just this Monday, Microsoft announced two new "neocloud" collaborations totaling over $10 billion. The Australian company Iren Ltd. will lease chip computing power to Microsoft for $9.7 billion, while Lambda Inc. has signed a "multi-billion dollar" agreement with Microsoft. These collaborations follow Microsoft's commitment to Nscale last month. According to informed sources, Microsoft's contracts with these neocloud companies are mostly five-year terms.

Like its peers, Microsoft is working to get data centers online as quickly as possible to meet the growing demands of customers and internal AI tool development teams. To this end, it has turned to neocloud suppliers—smaller-scale service providers that can quickly provide infrastructure, including CoreWeave and Nebius Group NV, to enhance its computing power supply capabilities.

Renting servers from neocloud companies can accelerate progress, as these suppliers have already resolved many logistical challenges, including power supply and chip acquisition.

Microsoft CEO Satya Nadella stated in a podcast last week that the current biggest bottleneck is not chips, but power.

A Microsoft spokesperson stated that the company's "global infrastructure strategy focuses on flexibility and optionality, based on short-term and long-term signals observed from customer demand. Microsoft employs a mixed approach of building its own data centers, leasing space, and collaborating with third-party suppliers, enabling it to rapidly expand computing power globally."

In a recent conference call with investors, Microsoft indicated it would increase investments to meet the demand for cloud computing and AI services. In the most recent fiscal quarter, Microsoft spent approximately $35 billion, primarily on data center leasing and server equipment.

Bloomberg analysts stated this week that the series of deals Microsoft has reached with neocloud companies "further reinforces our view that the entire industry is facing a severe capacity shortage, driven by the surge in demand for AI workloads."

Microsoft's competitors in the cloud services sector, such as Amazon, have not publicly disclosed as many actions in collaboration with neocloud as Microsoft has. Reports indicate that Google is renting some computing power from CoreWeave to collaborate with OpenAI. Meta has also previously rented computing resources from CoreWeave