By Rajwa Quasim <\/p>\n
Microsoft is increasingly positioning itself as a broad AI platform provider as it develops its own models while partnering with technology giants like Anthropic and OpenAI.<\/p>\n
This is evident from Microsoft\u2019s quarterly financial reports which show gains from its investment in Anthropic despite the declining value of its stake in OpenAI during the quarter. The company reported a $3.2 billion gain from its investment in Anthropic this quarter, boosting diluted earnings per share by 33 cents. In 2025, Microsoft invested $5 billion in Anthropic under a circular agreement<\/a> that committed Anthropic to purchase $30 billion worth of Microsoft\u2019s Azure cloud services.<\/p>\n READ: Microsoft CEO Satya Nadella warns companies relying on one AI model may not survive<\/a>\u00a0(July 28, 2026)<\/p>\n The company\u2019s investment in OpenAI did not fare as well as the company\u2019s stake down by $600 million trimming earnings per share by 7 cents. But on a full year basis, it still delivered a $5 billion gain. Microsoft owns around 27% of OpenAI.<\/p>\n For years, OpenAI was Microsoft\u2019s primary AI partner where it relied on Azure for cloud infrastructure while powering products such as Copilot. Now, Microsoft is strengthening<\/a> its portfolio as it broadens its AI strategy. CEO Satya Nadella said the company\u2019s model catalogue now includes more than 11,000 AI models from developers such as OpenAI, Anthropic, Mistral, and Microsoft\u2019s in-house MAI family. The company also announced new AI models covering image generation, voice transcription, coding, cybersecurity, along with its first reasoning model \u2014MAI Thinking One. Microsoft said the models are optimized to run on its in-house Maya 200 AI chip that can deliver 40% better performance per Watt.<\/p>\n Among the new launches is MAI- Cyber-1- Flash, a cybersecurity model aimed at competing with Anthropic\u2019s Mythos models. Nadella said it delivers comparable performance at about half the cost when combined with Microsoft\u2019s multi-agent security harness, although independent verification is not yet available. The company also plans to launch a unified Copilot \u201csuper app\u201d that combines chat, coding, coworkers and agentic autopilots into a single platform for consumers and businesses.<\/p>\n Nadella recommended use of AI models from different companies, including OpenAI and Anthropic, instead of relying on just one provider. Microsoft benefits when customers use these models because they run on their Azure cloud platform. At the same time, they encourage customers to avoid becoming dependent on any single provider. \u201cIf you look even at the Hugging Face incident, the biggest thing that we should take away from that is you can\u2019t sort of depend on any one model,\u201d Nadella said. \u201cYou will maybe need multiple models to even remediate some challenges that get caused by one model. Like that\u2019s the way to think about it, right? Which is you can\u2019t be subject to a refusal of one model.\u201d<\/p>\n
It also has stakes in both companies.<\/p>\n