Name: Microsoft Plans Data Center Push to Triple Its Computing Power
Uploaded: 2026-09-11T20:39:14+00:00
Duration: 308 s

Speakers

Anurag Rana — Senior Analyst, Bloomberg Intelligence

Summary

Anurag Rana discusses Microsoft's plan to more than triple its data center capacity from 12GW to 38GW, spending close to $200 billion annually. He views Microsoft as a major AI beneficiary and highlights continued industry data center capex, with Nvidia capturing roughly half of the spend. He also warns that public opposition to data centers could delay permits and power access, potentially softening the AI trade over the next 6-12 months.

Microsoft plans to expand data center capacity from 12GW to 38GW, spending near $200B a year.
Anurag Rana says Microsoft is a major AI beneficiary and remains free cash flow positive.
Data center capex is still rising across hyperscalers, with backlog exceeding capacity.
One gigawatt of data center capacity costs $40-60B, and half goes to Nvidia.
Public opposition to data centers could delay permits and power access.
This could soften the growth rate of the broader AI trade over 6-12 months.
Microsoft is outsourcing some capacity to neoclouds.

Ideas

Anurag Rana Senior Analyst, Bloomberg Intelligence 2:11

Microsoft benefits from AI data center expansion.

Microsoft plans to more than triple its data center capacity from roughly 12GW to 38GW, spending close to $200 billion a year to redesign infrastructure for GPU computing, with new data centers in Texas and Wisconsin. Anurag Rana views Microsoft as one of the biggest beneficiaries of AI, expects it to remain free cash flow positive this year, and sees the data center expansion as a positive despite public opposition that could cause delays.

Anurag Rana Senior Analyst, Bloomberg Intelligence 2:53

Microsoft outsourcing boosts neocloud demand.

Microsoft cannot fulfill all its data center needs in-house and is outsourcing capacity to neoclouds, creating demand for those providers as hyperscaler AI infrastructure expands.

Anurag Rana Senior Analyst, Bloomberg Intelligence 3:23

Data center opposition may soften AI trade.

Public opposition to data centers is the single biggest obstacle to the AI buildout, spanning both Democratic and Republican communities. If permits and power access become harder to secure, it could delay projects and soften the growth rate of the entire AI trade over the next 6 to 12 months.

Anurag Rana Senior Analyst, Bloomberg Intelligence 4:25

AI data center capex continues, not tapering.

Almost every major cloud provider is increasing capex next year, and Bloomberg Intelligence's data center model shows capex continuing over the next few years rather than tapering, because backlog revenue on balance sheets exceeds the capacity to fulfill it. One gigawatt costs $40-60 billion, supporting continued investment in AI data center infrastructure.

Anurag Rana Senior Analyst, Bloomberg Intelligence 5:01

Half data center capex goes to Nvidia.

Data center buildout costs about $40-60 billion per gigawatt, and roughly half of that spending goes to Nvidia, making it a key beneficiary of Microsoft's and the industry's AI data center capex.

Microsoft Plans Data Center Push to Triple Its Computing Power

Summary

Ideas

Up Next