Since Microsoft dropped a cool $1 billion on OpenAI in 2019, their exclusive partnership has been the tech equivalent of a power couple - until now. Today, the two firms announced an amended agreement allowing OpenAI to take its models on a tour of other cloud providers, not just Microsoft's Azure. Microsoft keeps its license for OpenAI's IP and models through 2032, and Azure remains the "primary cloud partner" (assuming Microsoft can keep up), but that license is now non-exclusive. The 20 percent revenue share OpenAI pays Microsoft continues, but with an unspecified cap and only guaranteed through 2030. Crucially, that revenue is now "independent of OpenAI's technology progress," sidestepping the infamous "AGI clause" that would have nixed exclusivity upon achieving artificial general intelligence - a milestone about as clear as a foggy London morning.
This amendment arrives two months after Amazon and OpenAI inked a $50 billion deal to run some models on Amazon Web Services. The Financial Times reported Microsoft threatened legal action over that dalliance, but today's amendment should make that a moot point. In a staff memo obtained by CNBC, OpenAI Chief Revenue Officer Denise Dresser noted the Microsoft partnership had "limited our ability to meet enterprises where they are - for many that's [Amazon] Bedrock," adding that interest in running OpenAI's models through Amazon's cloud has been "frankly staggering." Amazon CEO Andy Jassy chimed in on social media, saying he was "excited to make OpenAI's models available directly to customers on Bedrock in the coming weeks," alongside a new Stateful Runtime Environment, because apparently builders need even more choices to pick the right model for the right job.