Being for the Benefit of OpenAI

The worst-kept secret of the past year in the world of tech has now been spelled out more explicitly by the company: OpenAI aims to restructure itself as a for-profit entity. Not only has basically every major news outlet reported on this shift over the past many months, the company itself hasn't really shied away from talking about it in public, albeit more indirectly. But as we enter 2025 – the 10 year anniversary of OpenAI – the company is now making it quite clear: it's for-profit or bust, perhaps quite literally.

The reason for those earlier indirect confirmations seemed pretty clear: given the high-profile nature of OpenAI, arguably the most-important company working in the most important field at the moment, everyone was always going to be up-in-arms with such a shift. On paper, the move from a non-profit to a for-profit has some real bait-and-switch vibes. And that's why the company has kept repeating over and over and over again that no matter what ultimate form the company takes, the non-profit will always be a part of the equation. And beyond the potential for bad press, there are very real legal matters involved in moving from a non-profit to a for-profit, as the latter entails certain promises made to the public – promises which would need to be altered, if not broken, to make such a shift.

That is going to play out over the next many months. And this new post from OpenAI essentially lays out their high-level case for not only why they should be allowed to do this, but why they need to do this. And, in fact, the two elements, they argue, are related. Essentially: a lot has changed in the past decade. And whereas OpenAI thought that not only could they operate within the non-profit confines back then, they thought it would perhaps be a better structure to attract the best talent working on world-changing technology for the benefit of humanity.

At the very least, with the benefit of hindsight, that was naive. Of course, with the benefit of emails shared via legal discovery, we also now know that there were many more shades of gray in all of this. OpenAI wasn't without competition in 2015, in fact, they were specifically set up to compete with Google which had just acquired DeepMind, the true pioneering startup in the field. One of the original investors in that team was Elon Musk, who knew what the company was building and seemingly knew what Google scale and resources could do to such a mission. Again, this isn't speculation, this is all explicitly spelled out in emails.

And so Sam Altman's notion of an AI subsidiary of Y Combinator was turned into a new entity. And again, the non-profit element seemed just as much about being able to compete with Google's DeepMind from the name on down, OpenAI.

To be fair, many of those emails revealed that there was um, alignment, around creating an entity that wouldn't be controlled by Big Tech™ . But those emails also reveal that people like Ilya Sutskever, who had come from working on AI at Google (via an earlier acquisition), and Greg Brockman, who had been the CTO of Stripe, were more dogmatic about the mission, at least in those early days. Musk seemed to quickly realize that the only way to truly compete in the coming AI wars was to become a "real" company, albeit one he controlled. Altman seemed open-minded on either approach, but quickly realized that sticking with the non-profit structure gave him the path to lead the entity. And once Musk cut ties, Altman was clearly the right person to come up with some unconventional ideas to keep the non-profit going without Musk's money.

Enter, Big Tech™ . Specifically, Microsoft.

This was pretty clearly the moment when OpenAI actually changed from a non-profit. The convoluted "capped profit" structure was a clever "hack" around the obvious: OpenAI was now a non-profit in name only. And even their name was now misleading, as there wouldn't be much that was particularly "open" about OpenAI going forward. That was five years ago.

With that context, it's no wonder that nearly all of the founding team – beyond Altman and Brockman (and researcher Wojciech Zaremba) – left in the subsequent years. Oh yes, and that whole unfortunate coup business a year ago. OpenAI was simply not the same entity that was founded back in 2015. And after that break-up with Musk – and really, even before then – it couldn't be. People don't want to hear this, most notably many of the idealists involved in the early days of OpenAI, but it's simply hard to see a world in which OpenAI could compete right now as a non-profit. Hence, the formal shift towards becoming a for-profit!

To that end, unlike many other commentators on the subject, I find OpenAI's case for their metamorphosis compelling. Yes, it's a huge shift. Yes, it goes against many of their founding ideals. Yes, it's complicated and messy. But it's the only way they're going to be able to ultimately compete with the largest companies in the world who have now shifted essentially all of their resources into AI.

Musk is upset about this for a myriad of very obvious and very conflicted reasons – beyond the history, now there's xAI, of course. Meta also doesn't want the shift to happen for comically conflicted reasons – and their posturing that their worried about the precedent and public here is truly laughable. Both clearly know the obvious truth: if OpenAI doesn't make this shift, they're dead in the water. The end state would only be a full bail-out by Microsoft in the form of an acquisition, assuming we're allowed to do such deals again in post-Lina Khan America (still not entirely clear when it comes to Big Tech™, to be fair).

I also don't think OpenAI's stance that the remaining non-profit arm of the company would benefit from this shift is total bullshit. Depending on the structure, they will certainly be one of the most well-capitalized non-profits ever. Assuming they can sell off their eventual equity position in the broader OpenAI in tender offers over time, they'll have billions upon billions of dollars to work with on their mission. Maybe they can even bring back some of the more idealist original team members to steer said mission.

Clearly, this is all complicated, to say the least. But it doesn't mean OpenAI is aiming for the wrong thing here. There's some slight disingenuousness via the sin of omission simply given what we know about the history of the company thanks to those emails, but they're also not wrong in pointing out that for the company to compete right now – in arguably the most competitive environment ever – they need to make this shift. And yes, there's a world in which the new non-profit benefits immensely from such a shift too.

That still does not guarantee success, of course. Again, they're going up against the biggest companies in the world, all of whom are specifically focused on their prize. And that includes their own major partner, Microsoft. This new structure simply gives OpenAI a shot.

We can all joke about the billions burned and the lack of a path to profitability, but we've seen this story play out hundreds of times before. Sometimes it works out – Uber, Airbnb, etc – more often, it does not. OpenAI, right now, is still in the former camp, in my book. But the scale is above and beyond even those two others mentioned. OpenAI is likely going to need something more than $20B still to get to those elusive profits. It's just the reality of the situation right now. The others can afford to play with their profits, OpenAI cannot without help. And that help will only keep coming if there's real upside potential for the helpers. Love it or hate it, this is the game on the field right now. They have to play the game on the field.

And that game is going to change a lot in 2025, just as it did in 2024, just as it did in 2023, in 2022... We can all guesstimate how much money OpenAI will need to achieve profitability, let alone "AGI" (whatever that means), but it's going to be wrong. That includes estimates from the company itself. There are just too many variables and unknowns even now. And to combat that, OpenAI needs flexibility. Otherwise, they're just a failed non-profit. And who does that benefit?

One more thing: the Microsoft part of this equation remains the most interesting outstanding element to watch. Not only because of that relationship which is, um, complicated. But because of the position where it ultimately places Microsoft in the AI arms race. With an undoubtedly large and actual equity stake, do they kiss and make up with OpenAI? Rip up that AGI prenup? Does this morph more directly into a Microsoft/OpenAI vs. Google/DeepMind battle? Does the relationship go the other way as a result, with OpenAI getting even more independence from Microsoft (and vice versa)?