AI cloud business: Microsoft reins in OpenAI dependence

Microsoft’s AI cloud business is increasingly tied to OpenAI, with OpenAI reportedly driving about 70% of Azure’s AI revenue. In April 2026, Microsoft revised its OpenAI partnership terms to preserve primary cloud access while capping the revenue share through 2030—shifting Microsoft from dependence toward partial diversification. At Microsoft Build 2026, the company launched seven proprietary in-house AI (MAI) models and highlighted chip co-design efforts to gain more control over the stack from silicon to software. Microsoft also broadened third-party model access via Azure Foundry, listing providers such as Anthropic, Mistral and Meta. The Mistral partnership is a key piece of the AI cloud business diversification plan. Microsoft expanded collaboration with Mistral in July 2026, including investment in European GPU infrastructure. AMD’s Helios rack-scale systems are expected to ship to Azure in the second half of 2026 for frontier inference workloads, which support large-scale inference demand. Traders should watch the “70% OpenAI revenue” figure over upcoming fiscal periods. If it falls while total AI revenue rises, Microsoft’s diversification thesis is working. If it remains high, it suggests customers still pick Azure primarily for OpenAI access, limiting Microsoft’s leverage despite more alternative models.