Microsoft gains a large, recurring buyer as Meta directs hundreds of millions of dollars a year to Azure for access to AI models and computing power, a Bloomberg report said on Thursday. The spend is striking because Meta has committed enormous resources to build its own AI stack, yet still needs outside capacity to run increasingly costly models.
Meta’s appetite for third-party compute sits alongside an aggressive internal build. The company has invested in data centres, custom AI silicon through a partnership with Broadcom, and expanded its physical footprint. In July Meta announced a venture with BlackRock and infrastructure investors to develop a data-centre campus in El Paso, Texas, and Reuters reported the company’s free cash flow fell sharply in the second quarter as it increased spending on chips, servers, energy and data centres.
That combination of build and buy is not new. Meta began using Azure virtual machines for AI research in 2021, and in 2022 selected Microsoft Azure as a strategic cloud provider. Under the deal the company expanded use of Azure’s supercomputing resources, including a dedicated cluster containing 5,400 NVIDIA A100 GPUs for large-scale research workloads. More recently Meta broadened its supplier mix, announcing in April a partnership with Amazon Web Services to add tens of millions of AWS Graviton cores for agentic AI workloads.
For Microsoft the commercial upside is clear. Azure can monetize customers that prefer not to internalize every layer of AI infrastructure, even when those customers operate their own extensive compute fleets. Meta’s reported payments validate Azure’s positioning as an infrastructure layer for both traditional enterprises and the world’s biggest AI developers.
The relationship also sharpens an emerging competitive angle. Meta has explored selling AI computing capacity and models to outside customers, a move that would place it in potential competition with Microsoft Azure, Amazon Web Services and Google Cloud. That would transform some of Meta’s internal investments into a commercial product while keeping it a buyer of rival cloud services for the foreseeable future.
What happens next will matter for cloud economics and market share. If Meta continues to split workloads across its own farms and external clouds, Azure and other providers stand to collect substantial recurring revenue. If Meta pivots to commercialising its excess capacity, the company could become both a supplier and a competitor in a market where scale and capital intensity determine advantage.
