OpenAI is set to save roughly $97 billion on its Microsoft commitments through 2030 under the terms of their newly restructured partnership, according to The Information. The Information reports that the savings stem from the latest renegotiation between the two companies, which reshapes how OpenAI pays for compute and what Microsoft gets in return.
This is significant because OpenAI’s single biggest cost line is compute, and Microsoft has been both its largest backer and its largest bill. Trimming nearly a hundred billion dollars off that tab over the next five years changes the math on OpenAI’s path to profitability and on how aggressively it can chase Google, Anthropic, and Meta.
What changed
The original 2023 arrangement locked OpenAI into Microsoft’s Azure as its exclusive cloud provider, with Microsoft taking a sizable cut of revenue in exchange for cash and compute credits. That deal has been quietly unwinding for months. OpenAI now runs workloads on Oracle, CoreWeave, and its own Stargate buildout with SoftBank. Microsoft, in turn, has been diversifying its model lineup beyond GPT.
The $97 billion figure, as detailed in The Information, reflects what OpenAI would have owed Microsoft for compute and services under the old terms versus what it now owes under the new ones.
Why it matters for the industry
- OpenAI’s burn rate gets a serious haircut. The company has been on track to lose tens of billions annually as it scales training and inference. Cutting nearly $20 billion a year in projected Microsoft costs makes the IPO math look very different.
- Microsoft is no longer the gatekeeper. For two years, the assumption was that OpenAI couldn’t move a workload without Satya Nadella’s blessing. That era is ending. OpenAI can now route training and inference where it’s cheapest or fastest.
- The cloud wars just got hotter. Oracle, CoreWeave, and Google Cloud all stand to pick up OpenAI workloads. Expect aggressive pricing from every hyperscaler chasing frontier-model customers.
- Stargate gets validated. OpenAI’s own data center buildout with SoftBank looked like a moonshot when announced. With Microsoft loosening its grip, OpenAI now has the runway to actually own its compute stack.
What to watch next
The restructured deal almost certainly clears the way for OpenAI’s long-rumored conversion to a fully for-profit entity and an eventual public offering. Microsoft keeps a stake and presumably a revenue share, but loses the exclusivity that made it the de facto operator of frontier AI.
For practitioners and enterprise buyers, the practical takeaway is straightforward: OpenAI is no longer Azure-only in any meaningful sense. Expect GPT models to show up in more places, on more clouds, at more competitive prices.
Full breakdown of the deal terms is available at the original source.