The War of the AI Roses
In the 1989 film The War of the Roses,1 a seemingly picture-perfect marriage between Oliver Rose (Michael Douglas) and Barbara Rose (Kathleen Turner) deteriorates over time to the point where disagreements turn into sabotage as neither side will agree to move out of their shared mansion. They keep living together, but it's increasingly an absurd situation. That's where my mind went when reading the latest about the drama between Microsoft and OpenAI.
Publicly, both sides continue to put out statements that everything is fine, so as to try to deflect such reporting. But they read as half-hearted attempts at best. There is just too much smoke coming out of the kitchen for there to be no source of heat.
I started writing about this friction back in March, around the Inflection acquisition hackquisition, which even back then clearly felt like a hedge of sorts – albeit an expensive one – against what had happened with the brief ouster of Sam Altman at OpenAI. By May, a tension felt palpable even to us outsiders. And a series of events in the intervening months has seemingly made things even more awkward. From boardroom drama to personnel challenges to sales rivalries to more Microsoft hedging – while that initial hedge started to produce fruits of its own...
And now consider today's stories. The New York Times has the headline 'Microsoft and OpenAI’s Close Partnership Shows Signs of Fraying'. The story, reported by Cade Metz, Mike Isaac, and Erin Griffith (with additional reporting by Karen Weise and Tripp Mickle) – so, five reporters working sources – at one point cites "19 people familiar with the relationship between the companies" in reporting how strained the five-year partnership has become. Nineteen.
Meanwhile, not to be left out of the reporting fun, The Wall Street Journal today has a story entitled 'The $14 Billion Question Dividing OpenAI and Microsoft'. That story, reported by Berber Jin and Corrie Driebusch (with additional reporting by Tom Dotan) – so, three reporters working sources – details how the complicated process to morph OpenAI from a non-profit to a for-profit company is further driving a wedge between themselves and Microsoft, and this is likely to get even messier once the situation is sorted.
Again, these are just the latest in a series of stories from many different publications over an extended period of time that have painted a picture not unlike the plot of The War of the Roses. We're perhaps not at the point of throwing vases at one another yet. But we're seemingly not far from that either as the two sides not only remain bedfellows, but have furthered their commitment to one another in this latest OpenAI funding round, which saw Microsoft invest yet again. I had been wondering just how much (and in what form) the investment would be, and if my skills of deduction are correct, it seems like roughly $750M in new commitments came from Microsoft recently – taking the previous $13B invested up to $13.75B, as the WSJ reports. It's still not clear if that's cash or cloud credits (or some split), but they're roughly the same thing here since most of that money flows back to Microsoft anyway to run OpenAI's models on Azure servers.
But actually, as the NYT report details, Microsoft initially didn't want to invest more money into OpenAI. Spooked by the ouster of Altman – an event which we're coming up upon the year anniversary of, even though it feels like five years ago – Microsoft started to think about hedging, reports... well, basically everyone at this point. Of course, once this round was coming together, Microsoft basically had to invest. They're already so wedded to the company that not only would it optically look bad, it would potentially harm their own large stake.
Well, not exactly a stake in the company – yet – but the stake in the profits of the company, which also don't exist yet, of course. I loved this short paragraph in the WSJ piece:
Microsoft, other private investors and OpenAI employees currently own rights to future profits generated by a for-profit subsidiary controlled by OpenAI’s nonprofit board. Their returns are capped based in part on when they were issued these profit participation units.
Write a short story in six words: Future profits. For-profit. Nonprofit. Profit participation.
The end result, all sides hope, is a for-profit company with a non-profit arm. Basically the inverse of the structure today. But it's insanely complicated, more so because of the stakes involved, quite literally:
In a sign of how significant the outcome will be for Microsoft and OpenAI, both have hired investment banks to advise them on the process. Microsoft is working with Morgan Stanley and OpenAI has tapped Goldman Sachs, according to people familiar with the matter.
In addition to figuring out how big a piece of the restructured artificial intelligence company Microsoft will own, the two sides must figure out what governance rights it will have.
It is unusual for nonprofits to convert to for-profit companies and particularly so for a company of OpenAI’s size and value.
I'd venture to say that it's not just "unusual", it's "unprecedented" at this scale and scope. Even if the companies nail such a transition, there are undoubtedly going to be lawsuits galore just given the money involved. And underpinning that governance question is this:
Further complicating matters is the likelihood that the larger Microsoft’s stake is, the more scrutiny it could invite from antitrust regulators already taking numerous actions to try to tame the power of giant tech companies.
I've tried to guesstimate where all the various entities – including, notably, the left-over non-profit arm – will land with regard to equity. 10 days ago, I landed at this potential structure coming out of the transition:2
- Microsoft: 33%
- OpenAI Non-Profit: 25%
- Sam Altman: 5%
- Khosla: 5%
- Thrive: 5%
- Other Co-Founders: 5%
- Others: 12%
If Microsoft really ends up owning one-third of the company (and that would be quite a bit lower than their reported 49% profit-share – which itself is complicated as it starts as a 75% profit-share before recouping capital, etc), the governance question becomes even more complicated. Microsoft walked away from their board observer role a few months ago (perhaps to ensure Apple didn't also get a seat at the table in the room where it happens – and/or perhaps to ensure the government didn't start getting too interested in their deal with OpenAI). That role was only put in place after Altman's ouster. Microsoft said they were fine giving up that role after only a few months with a better board structure now put in place. But can they really control that much of OpenAI without having any formal oversight? I highly doubt it, for fiduciary reasons if nothing else.
Switching over to the NYT story:
OpenAI expected to spend at least $5.4 billion in computing costs through the end of 2024, according to documents reviewed by The New York Times. That amount was expected to skyrocket over the next five years as OpenAI expanded, soaring to an estimated $37.5 billion in annual computing costs by 2029, the documents showed.
A week ago, I guesstimated that just based off of their own (leaked) models, OpenAI might need to raise another $25B in capital before they are able to turn a real profit. As the NYT piece notes, Microsoft has restructured at least some of their contracts with OpenAI, presumably to make the costs less onerous on the startup, but we're undoubtedly still talking tens of billions of dollars required that the company doesn't yet have. Microsoft will need to and want to watch this closely – especially as Wall Street inevitably starts to get antsy about their AI spend. The OpenAI relationship has immensely helped Microsoft's stock – perhaps to the tune of a trillion dollars or more in our current AI-crazed environment (see also: NVIDIA, once again worth more than Microsoft). But it can turn and hurt that stock too if things aren't kept in check.
Perhaps that's in part why they're okay with OpenAI turning to Oracle for help with NVIDIA chips and servers (even though Microsoft, optically, still wants credit for brokering such deals). And maybe even why they were okay with Apple investing, before the iPhone maker thought better of it – which is now another awkward point of tension because Apple will undoubtedly spike OpenAI's bill due to Microsoft thanks to ChatGPT usage when it rolls out as a part of Apple Intelligence later this year... OpenAI undoubtedly also saw an Apple investment as a way to help counterbalance the Microsoft investment. Alas...
At the same time, Microsoft sitting in the OpenAI board room would ramp up in awkwardness fast as the two increasingly compete directly. Especially since doing so means Microsoft shifting more focus to the aforementioned Inflection team, now fully integrated into an internal Microsoft AI org being run by Inflection co-founder Mustafa Suleyman. Does OpenAI have some thoughts on that now?
But in March, Microsoft paid at least $650 million to hire most of the staff from Inflection, an OpenAI competitor. Inflection’s former chief executive and co-founder, Mustafa Suleyman, oversees a new Microsoft group that is working to build A.I. technologies for consumers based on OpenAI software. He is also the point person for Microsoft’s long-term effort to build technologies that could replace what the company is getting from OpenAI, according to two people familiar with Microsoft’s plans.
“Microsoft could be left behind if it is only using OpenAI technologies,” said Gil Luria, an analyst at the investment bank D.A. Davidson. “It is a real race — and OpenAI may not win it.”
Some OpenAI executives and employees, including Mr. Altman, are angered that Mr. Suleyman is at Microsoft, according to five people familiar with the relationship between the two companies. Mr. Suleyman’s team is part of a group of Microsoft engineers who work directly with employees at OpenAI. Dozens of Microsoft engineers work on-site at OpenAI’s offices in San Francisco and use laptops provided by OpenAI that are set up to maintain the startup’s security protocols.
Some OpenAI staff recently complained that Mr. Suleyman yelled at an OpenAI employee during a recent video call because he thought the start-up was not delivering new technology to Microsoft as quickly as it should, according to two people familiar with the call. Others took umbrage after Microsoft’s engineers downloaded important OpenAI software without following the protocols the two companies had agreed on, the people said.
Seems like a totally normal, healthy relationship, for your main backer to be openly working to subvert you. On that topic...
Still, OpenAI employees complain that Microsoft is not providing enough computing power, according to three people familiar with the relationship. And some have complained that if another company beat it to the creation of A.I. that matches the human brain, Microsoft will be to blame because it hasn’t given OpenAI the computing power it needs, according to two people familiar with the complaints.
Oddly, that could be the key to getting out from under its contract with Microsoft. The contract contains a clause that says that if OpenAI builds artificial general intelligence, or A.G.I. — roughly speaking, a machine that matches the power of the human brain — Microsoft loses access to OpenAI’s technologies.
The clause was meant to ensure that a company like Microsoft did not misuse this machine of the future, but today, OpenAI executives see it as a path to a better contract, according to a person familiar with the company’s negotiations. Under the terms of the contract, the OpenAI board could decide when A.G.I. has arrived.
It's not quite that straightforward, since, if memory serves, Microsoft would get to keep everything OpenAI had built up until that point of AGI, which would be cleaved off and thus, effectively end the go-forward relationship between the two companies. But that also feels like it's a long way away – if we ever get there. As I wrote back in May:
From the get-go of the partnership, the talk had been that OpenAI might be content to simply let Microsoft take ChatGPT and the other products that were made along the path to AGI. That AGI was the only real goal and Altman and team had some sense of how much money, data, and compute it would take to get there, and so the non-profit went for broke and the for-profit was born. The soul, essentially, was sold.
But what you increasingly hear now is that AGI might be nowhere near reality and in fact, may never be. This isn’t helped by the fact that no one has an actual definition of the mythical technology, but instead it gets sort of the porn-y vague “you’ll know it when you see it” hand waves more and more.
While AGI is still getting lip service from OpenAI, I believe you’re going to increasingly see the focus more on regular old non-AGI product development. Like GPT-4o, which soundslike Scarlett Johansson, quite literally, brilliant. Altman’s brief comments upon the announcement — no, not “her” — almost seems to indicate as much. And whatever they’re about to do with Apple may be the logical next step down that path.
Said another way, OpenAI can keep saying they’re going down the path to AGI without actually doing it. That’s perhaps unfair, as they, like most AI companies, still would love to stumble upon such a creation. But I’m just not sure how realistic of a goal it is anymore as the importance of AI minus the “G” comes more and more into focus. To say that another way, I think OpenAI can still be one of the most important companies in the world without creating AGI. And I think they may have come to this realization as well.
But that leaves the relationship with Microsoft in an even more precarious place. In a normal world, Microsoft would just acquire the company outright. But we don’t live in a normal world and they can’t do that. And so they have to continue to partner with a partner with which they’re increasingly at odds. While that partner also partners with other partners which will put the two sides even more at odds. It’s all quite odd.
And here we are, all these months later, and it's even more odd with the two more at odds. If Microsoft's consumer Copilot efforts fail to take off in the way they're hoping is that better or worse for the relationship with OpenAI? I'm guessing we're going to find that out soon enough, perhaps even before we know what Microsoft's ownership stake in OpenAI ends up being.
At one point in The War of the Roses, Oliver and Barbara have the following back-and-forth:
Oliver: With my money! It's a lot easier to spend than it is to make it, Honeybun!
Barbara: You might not have made it if not for me, Sweetcakes!
So how does the film end? I'll let ChatGPT summarize:
After a series of escalating battles over their house, Oliver and Barbara Rose’s final confrontation leads to both of them falling from a chandelier, crashing to the floor. They lie gravely injured, and in their final moments, Oliver reaches out for Barbara, but she flicks his hand away. The movie closes with their demise, underscoring the destructive consequences of their bitter divorce, leaving their lawyer Gavin D'Amato to reflect on the futility of their conflict.
Awkward.
M.G. Siegler
M.G. Siegler
M.G. Siegler
M.G. Siegler
M.G. Siegler
1 Directed by Danny DeVito!
2 My math leaves 10% open for wiggle-room and various other needs, like an option pool, etc.