Microsoft is pushing forward with its self-developed AI model research with unprecedented intensity, aiming to build a cutting-edge system that can rival OpenAI and Anthropic in the coming years. Microsoft's AI head, Sundar Pichai, clearly stated that the company's core goal is to create the most advanced model in the industry.
According to the latest disclosed strategic plan, Microsoft aims to make its self-developed models among the best in the world in text, image, and audio processing capabilities by 2027. This means Microsoft is trying to break its long-term reliance on external partners in the field of general large models.
Powering with Computing Capacity and Early Success of Self-Developed Models
To support this ambitious vision, Microsoft is deploying large-scale clusters of NVIDIA's latest GB200 chips. The company plans to increase its underlying computing power to the forefront of the global scale within the next 12 to 18 months, providing a solid hardware foundation for model iterations.
As a阶段性 achievement, Microsoft released a new speech transcription model on April 2nd. Test data shows that in 25 major languages, the model has outperformed current market competitors in 11 of them.
Breaking Free from Agreements to Achieve Technological Autonomy
Previously, Microsoft faced many restrictions when developing general large models due to its agreement with OpenAI. However, after the agreement was adjusted and "unbundled" last year, Microsoft has cleared internal obstacles and fully shifted towards the path of technological autonomy.
CEO Nadella emphasized in an internal meeting that achieving AI capability autonomy within the next three to five years is the group's core goal. This strategic shift indicates that Microsoft will transition from an AI technology integrator to a top-tier developer with core autonomous capabilities.
