Reuters reports that Microsoft is on a shopping spree for AI startups, stockpiling top talent as it prepares for life after OpenAI. The company has set an internal goal of releasing its own frontier-grade large language model by next year.
The shopping list is ambitious—Microsoft even considered buying Cursor, a code-generation startup that competes with its own GitHub Copilot. This spring, Microsoft seriously evaluated acquiring Cursor but backed off due to antitrust concerns. Soon after, Elon Musk's SpaceX (which just acquired xAI in February) swooped in and sealed the deal.
Now the bidding war is centered on Inception, a startup founded by a Stanford team in mid-2024. Microsoft's venture arm M12 participated in Inception's $50 million seed round at the end of 2025. Inception is taking a novel approach: using diffusion models—typically used for image and video generation—to train large language models, generating and optimizing multiple tokens simultaneously for a major speed boost. Sources say Inception is seeking an acquisition price north of $1 billion.
These moves signal a real loosening of the Microsoft-OpenAI alliance. Their once-exclusive contracts have been progressively untangled: a revised agreement in late 2025 allowed Microsoft to develop AGI independently, and in late April, OpenAI was permitted to build products for Amazon and other direct Microsoft competitors.