Microsoft Swallows OpenAI’s Core Team – GPU Capacity, Incentive Structure, Intellectual Property, OpenAI Rump State
OpenAI is nothing without its people.
OpenAI has imploded. Satya Nadella has somehow masterfully navigated this weekend’s wild events and has ended up with the core OpenAI research team absorbed into Microsoft.
And we’re extremely excited to share the news that Sam Altman and Greg Brockman, together with colleagues, will be joining Microsoft to lead a new advanced AI research team.
To quickly recap for those who weren’t caught in the drama loop and enjoyed their weekends, Sam Altman (CEO) was fired by OpenAI’s board for vague reasons. This was done without the board discussing their decision with any of OpenAI’s investors, including Microsoft. Greg Brockman (Chairman and President) was also kicked off the board and demoted. He subsequently quit in revolt. The OpenAI board, which has no legal obligation to shareholders or really to anything besides AI safety, has had plenty of opportunity to explain their drastic actions, but they have not done so thus far, even internally. Multiple other key leaders on GPT-4 and GPT-5 such as Jakub Pachocki and Szymon Sidor have also quit.
Mira Murati, the CTO, was appointed as the interim CEO. On Saturday night, in what we are calling the night of a hundred hearts, most OpenAI employees, including Mira, publicly gave their support to Sam and Greg. The employees and Microsoft wanted Sam back.
On Sunday, there were negotiations to rejoin the firm, but in the end the board decided against it. Instead appointed as CEO Emmett Shear, the founder and long-time CEO of Twitch.tv. Emmett Shear has publicly stated he wants to significantly slow down AI progress.
Sam and Greg were considering creating a brand-new startup, but that would have likely caused a >1 year speed bump. Instead, now there is a new subsidiary within Microsoft.
I’m super excited to have you join as CEO of this new group, Sam, setting a new pace for innovation. We’ve learned a lot over the years about how to give founders and innovators space to build independent identities and cultures within Microsoft, including GitHub, Mojang Studios, and LinkedIn, and I’m looking forward to having you do the same.
There is a mass exodus of the core OpenAI team leaving and joining Microsoft. This new organization within Microsoft will get hundreds of technical staff from OpenAI.
Incentive Structures
The OpenAI for-profit subsidiary was about to conduct a secondary at a $80 billion+ valuation. These “Profit Participation Units” (PPUs) were going to be worth $10 million+ for key employees. Suffice it to say that this is not going to happen now, and the OpenAI board has foolishly destroyed the chance of generational wealth for many of the team. Despite this literal fumbling of the bag, key OpenAI employees who leave will be treated extremely well.
Part of Satya’s incredible deal with Sam and Greg is likely that these key OpenAI employees that join Microsoft will have their now worthless PPUs pseudo-refreshed for equity in Microsoft which vest over multiple years. There will be compensation packages that are $10 million plus ($1e7+) for those who were with OpenAI for multiple years.
There is likely also a huge incentive-based pay for all the huge bets and risks this new team will be making, which will align and incentivize the OpenAI team to do what they do best, accelerate.
The narrative that risk takers chasing generational wealth won’t want to join Microsoft and instead chase the start-up life is quite moot. There is a possibility that this subsidiary may also be allowed to grant its own equity to employees in a form that is not directly Microsoft stock.
Compute Is King
Microsoft had previously placed huge bets on OpenAI, with plans for >$50B annual datacenter spend to race to AGI and deploy their GPT-4 based copilot products. Our data shows one of OpenAI’s next training supercomputers in Arizona was going to have more than 75,000+ GPUs in a singular site by the middle of next year.
Our data also shows us that Microsoft is directly buying more than 400,000 GPUs next year for both training and copilot/API inference. Furthermore, Microsoft also has tens of thousands of GPUs coming in via cloud deals with CoreWeave, Lambda, and Oracle.
There are a few big question marks on what OpenAI has guaranteed. Most of the Microsoft investment in OpenAI is in the form of compute credits. While there are agreements on the sizes of supercomputers that must be delivered, we believe Microsoft was on track to blow way past those goals and deliver OpenAI more than legally required, meaning a rebalancing is possible.
Microsoft can likely claw back or not deliver quite a bit of what it had planned for OpenAI. These compute resources can be routed to the new internal team. Furthermore, given how killer Microsoft’s legal team is, it’s possible that an even large portion of what is already delivered or soon to be delivered can be clawed back.
If the new team were to spin out and make their own startup, they would have had tremendous difficulty acquiring enough compute to build a GPT-5 scale model before Anthropic or Google. Given there is a sort of runaway escape velocity here, this would put them at a huge disadvantage in the race to AGI. By joining Microsoft, the former OpenAI team will still have access to the necessary compute resources next year.
It is very likely that this development accelerates spending further and Microsoft’s orders for GPUs will have to go up yet again in order to fulfill the OpenAI contract and give the new company everything they need to build GPT-5 next year.
Intellectual Property
Within a few hours of Sam being fired, Satya minced no words in his tweet “supporting” OpenAI’s new leadership. It was effectively a thinly veiled threat that said “I don’t need you.” (which he doesn’t for copilot deployment)
We have a long-term agreement with OpenAI with full access to everything we need to deliver on our innovation agenda and an exciting product roadmap
Microsoft has full legal rights and access to the weights of the base GPT-4 model as well as the various fine-tuned versions and DALL-E 3.
If the team went down the startup path, they would have had to spend significant time rebuilding GPT-4. Instead, at Microsoft they will have access to much of the IP they require for future products.
What’s more important to understand is if Microsoft has legal direct access to all the data and code used for pre-training and RL. It is obviously all stored on Azure, but if the new Sam-led internal team can freely access that, they can basically start exactly where they left off without much of a hiccup. If they cannot get it, then we estimate that it could possibly lead to only a 4-6 month delay vs prior. While this small of a delay sounds insane to say... Talent is everything.
OpenAI Silicon
The silicon efforts at OpenAI are effectively dead now without Sam and certain key individuals. Those efforts were completely independent of the efforts at Microsoft regarding which we detailed specifications, performance, and more here.
These plans are far more immediate and realizable now inside the Microsoft subsidiary.