Microsoft Accelerates Self-developed AI Models, Targets Top-tier Text, Image and Audio Capabilities by 2027

Microsoft is accelerating the development of its self-developed AI models, aiming to achieve top-tier text, image and audio processing capabilities by 2027 to compete with rivals like OpenAI and Anthropic. Mustafa Suleiman, head of Microsoft AI, emphasized this is a core strategic direction.

Recently, Microsoft released a voice transcription model that outperformed competitors in tests for 11 out of 25 major languages. However, this model is a specialized tool and differs from general-purpose large models. Meanwhile, Microsoft is deploying NVIDIA GB200 chip clusters and plans to boost computing power to cutting-edge scale within 12-18 months, with internal team adjustments.