Microsoft’s CEO of AI, Mustafa Suleyman, stated that building the largest artificial intelligence models in the world is expensive and duplicative. The company partners with OpenAI to access its Copilot model, which has recently gained a “memory” feature to retain key facts about users.
Suleyman emphasized that Microsoft aims to develop AI self-sufficiently in the long term but acknowledged that building state-of-the-art models is capital-intensive and unnecessary for duplication. Instead, the company is focusing on specific use cases and waiting three or six months behind the frontier to access more affordable computing power.
Microsoft has an existing partnership with OpenAI, which it relies on for AI models, and also partners with CoreWeave for supplemental computing power. The company’s Copilot model will gain this “memory” feature, previously introduced in OpenAI’s ChatGPT chatbot, allowing users to access top-flight large language models.
However, Microsoft has recently added OpenAI to its list of competitors after investing $13.75 billion in the startup. Despite this, Suleyman believes that the partnership will continue for at least until 2030, citing the benefits of a long-term relationship with OpenAI.
Source: https://www.cnbc.com/2025/04/04/microsoft-ai-chief-sees-benefits-to-ai-models-that-are-months-behind.html