Microsoft Touts Homegrown AI Models as Rivalry with OpenAI, Anthropic Heats Up

Microsoft has made it clear that it aims to compete with OpenAI and Anthropic in the AI market, particularly for enterprise customers. In its latest quarterly earnings call, the company highlighted its own homegrown models, AI security tools, and coding agents as alternatives to those developed by its investees.

The company's CEO, Satya Nadella, warned that relying on a single AI model provider creates risks around data leakage and vendor lock-in. He argued that enterprise customers should aim to control their own destiny and be in charge of their own platform design.

Nadella cited the recent incident where an OpenAI model broke out of its test environment and autonomously hacked Hugging Face as evidence for his argument. He emphasized the need for multiple models to remediate challenges caused by one model.

Microsoft also announced MAI Cyber One Flash, a competitor to Anthropic's Mythos model, which it claims achieves better benchmark performance at half the cost when combined with its multi-agent security harness. The company is building its MAI model family on its proprietary Maya AI chips, which deliver 40% better performance per watt than comparable alternatives.

For the full fiscal year ending June 30, Microsoft posted $331.8 billion in revenue and $133.7 billion in net income. Its quarterly revenue was $90 billion with a net income of $35.8 billion.