The $300 billion (€278 billion) AI outfit has been bogged down in tortuous talks with Vole over a contract set to run until 2030. The deal rewrite is crucial if OpenAI wants to let investors own equity and eye up a possible IPO.
But insiders say key sticking points still separate the two firms and talks could now rumble past the 31 December deadline.
If they don’t shake hands by then, SoftBank has the right to yank its $10 billion (€9.3 billion) investment according to its terms. That would scupper OpenAI’s pitch to drum up even more capital.
At the heart of the deadlock is Microsoft’s grip on OpenAI’s API, which it hosts exclusively on its Azure cloud. That makes Redmond a critical gatekeeper to the models powering ChatGPT.
OpenAI wants to loosen those reins and sign deals with Google and Amazon Web Services, just like its rival Anthropic. That could bump up API sales, which currently make up about a quarter of its €11.1 billion annual recurring revenue.
Microsoft isn’t exactly thrilled at the idea of sharing its prize asset. The firms are discussing a compromise that might let OpenAI serve non-Azure government punters only.
Then there’s the row over intellectual property. Microsoft wants full access to how future models are trained, while OpenAI might prefer the tech giant just gets to use them, not see how the sausage is made.
One particularly juicy point is the so-called AGI clause. This bit of legalese says Microsoft could lose IP access if OpenAI ever creates “artificial general intelligence” in other words, something that does most jobs better than humans.
Microsoft supremo Satya Nadella wants that clause gone. OpenAI, not so much. They know it gives them heavy leverage.
One source familiar with the talks said: “OpenAI having the AGI clause is negotiating chit. It’s a threat, but it’s more like mutually assured destruction because if it doesn’t go by year-end, they won’t be able to raise any money again and Sam [Altman] knows that.”
How these squabbles pan out will affect how much equity Microsoft bags in the restructure. The Redmond giant is expected to end up with between 30 and 35 per cent of OpenAI, having already lobbed in more than $13 billion (€12 billion).
Despite the back-and-forth, a deal is still the most likely outcome, insiders reckon, though it might not arrive before the fundraising deadline.
Both firms put on a brave face in a canned statement: “We have a long-term, productive partnership that has delivered amazing AI tools for everyone. Talks are ongoing and we are optimistic we will continue to build together for years to come.”
Even if Microsoft gets sorted, OpenAI still has to wrangle with other shareholders and attorneys-general in California and Delaware before it can change its structure.
The bean-counting trench warfare is being led by OpenAI’s chief financial officer Sarah Friar and Microsoft’s counterpart Amy Hood, according to those familiar with the talks. The idea is to replace the current profit-sharing setup with proper equity.
OpenAI’s last two funding rounds baked in clawback clauses. One in October valued it at $157 billion (€145 billion). The March round SoftBank led valued it at $300 billion (€278 billion) and gave backers the right to walk if OpenAI doesn’t restructure by the end of 2025.
Executives at OpenAI aren’t too worried. They think SoftBank won’t bolt and believe the firm can still trouser the cash even with its current setup, thanks to rapid growth since SoftBank signed on for a $40 billion (€37 billion) round in March.
The company is now chatting to investors about a secondary sale that would give it a $500 billion (€463 billion) valuation, a tidy bump for SoftBank. There’s also talk of a “primary” raise at an even loftier number, according to sources in the loop.
And while the whole thing remains a giant question mark, OpenAI’s €7.7 billion March funding tranche was apparently massively oversubscribed.