In a related development, Microsoft unveiled a new speech-to-text model that reportedly outperformed competing systems in benchmark tests across 11 out of the 25 most widely spoken languages worldwide. The model is designed to filter background noise in crowded environments, making it particularly effective in real-world use cases. It is expected to be gradually integrated into services such as Microsoft Teams in the coming months.
At the same time, Microsoft is expanding its computing infrastructure to support the development of more comprehensive AI models. Since October, the company has begun deploying an architecture powered by advanced Nvidia GB200 chips to reach the computing scale required for next-generation models within 12 to 18 months.
Previously, Microsoft’s efforts were limited by the terms of its partnership with OpenAI, which prevented the company from developing its own general AI models in exchange for access to technologies such as ChatGPT. However, this restriction was lifted after the two companies renegotiated their agreement last year, allowing Microsoft to pursue the development of its own models.
Internally, Microsoft has also undergone a recent restructuring. Responsibility for the AI assistant Microsoft Copilot has been transferred to Jacob Andreou, while Mustafa Suleyman is focusing on model development reflecting the challenges Microsoft has faced in building a mainstream competitor to tools like ChatGPT and Google Gemini.
These developments come after Microsoft recorded its worst quarterly stock market performance since the Global Financial Crisis of 2008, as investors questioned the company’s ability to convert its massive AI investments into sustained and measurable growth.