The IPOpenAI
Over the weekend, I got an email asking a pretty straightforward question: why on Earth does a startup need to raise $40B? And actually, it's the same basic question that has come up a few times in a range of conversations over the past several days, ever since the reports of OpenAI's would-be mega fundraise first surfaced last week. It's a good question! I think it's worth taking a step back here, because it truly is an incredible amount of money.
To answer the question at the highest level, OpenAI needs to raise that amount of money in order to pay for the compute needed to train and execute their AI models. In that, I would bucket research and development on said models and other technology.1 And related to that, they obviously need money to hire and retain talent. But that's going to be in the millions of dollars-range, not billions – the AI ecosystem isn't that competitive. Yet.2 And yes, there are other general expenses for operating a business – undoubtedly more for legal, regulatory, and compliance costs than your average startup. And they also now have data licensing agreements with various players that are expensive as well. But again, the vast majority of the billions upon billions are going towards compute, and the related infrastructure.3
On the AI front, OpenAI is hardly unique here. But as the company currently believed to be the leader in the space, their needs – and thus, expenses – are seemingly higher than any other startup right now. Anthropic and xAI have been growing in terms of spend up until now, but this round will kick things up an order of magnitude. And we can probably expect those other players to at least try to raise such amounts to keep up with the Joneses, as it were. Elon Musk will undoubtedly be able to. Anthropic would likely have to go back to Amazon and/or Google, which will be especially awkward since Microsoft just seemingly tapped out leading up to this OpenAI round, offloading the 'Stargate Project' to Oracle and others – including, yes, SoftBank.
The capital canon is back, baby.
But actually, part of the reason why OpenAI needs to raise this amount of money is also because of those Big Tech players. Because each of them is insanely profitable with their own cloud infrastructure, they can spend in ways that any startup cannot. Well, until SoftBank and/or the sovereign wealth funds step in, it seems. But even then, Microsoft is talking about spending $80B in CapEx this year. Meta, $65B. Amazon, Google, etc. They'll all be in the same general boat.4 And that boat is still bigger than OpenAI's. That's the benefit of having your own cloud, and tens of thousands of existing employees to deploy at a problem, and being a trillion-dollar-plus public company.
There are downsides to the latter too, of course. Which is why they have to be careful in how they talk about such spend. Wall Street is watching. Wall Street is not yet watching OpenAI. They are watching SoftBank (in Japan, at least). But this is what Masa Son does. If you buy his stock, you're sort of signing up for a wild ride at this point. Also, SoftBank has their incredible ARM bet to back them up – which they're ready and able to use on the debt front. Masa Son always seems to have one bet up his sleeve that gives him a way to justify any spend. Yes, even with WeWork. But again, we're pushing new envelopes with this deal...
Speaking of going public and this deal. Assuming OpenAI raises the $40B target, that would be more than any company has ever raised in an IPO. The current record-holder is Saudi Aramco, which raised just under $30B in 2019.5
After that was another company Masa Son knows well: Alibaba, which went public raising $25B in 2014. Third is SoftBank itself! $23.5B in 2018.6 The largest fully American company IPO was Visa at just under $20B in 2008. The largest tech company was Facebook at $16B in 2012.
Yes, OpenAI here is raising far more than double what Meta (né Facebook) did in their IPO. More than double any American IPO. More than any IPO, ever.
It brings up an interesting question: if they were to go public, could OpenAI even raise as much money as they're raising here? I mean, it depends on the deals they cut leading up to it, and the terms, of course. But just from "traditional" IPO backers, they probably could not! So in that regard, this round is quite literally bigger and far more important to them than an IPO.
In a way, it's the ultimate extension of what has been happening with the IPO market itself. The biggest and "best" companies opt to stay private with a seemingly endless supply of capital at their disposal, while the merely "good" companies are the ones that test the public markets now. Some will say this is unfair, and some companies do have other reasons to go public, but directionally, this is obviously true. And OpenAI is about to put an exclamation point on it!
So, back to the question: why does OpenAI need to raise this much money? It's a combination of competition, actual needs, and yes, because they can. And that's mixed with investor appetite to fund the next big thing, and now in particular, one investor's long-standing goal of doing just that in the form of Masa Son.
Perhaps it helps to think of the capital flow this way: SoftBank -> OpenAI -> Microsoft -> NVIDIA. It's over-simplified, of course.7 But again, at a high level, you can sort of frame it this way. It's the equation that made NVIDIA rise from a unique computer graphics card player to the most valuable company in the world, in record time.
And now Wall Street is in the process of knocking them down a few pegs over concerns that the market is in over its skis, as perhaps showcased by DeepSeek. But if and when OpenAI gets that $40B, a lot of it will eventually funnel down to NVIDIA after filtering through Microsoft (and now Oracle). If you want to add the other complex layer to this, a lot of that capital will undoubtedly be debt leveraged off of SoftBank's ARM holdings. And why is ARM currently valued so highly? Because of the AI boom, of course.
M.G. Siegler
1 Though all of these companies will break this out in separate categories of spend. But unless they're public, or you're an investor in said company, you don't get to see said spend. And even when you do for the tech giants, a lot of it is obfuscated so as to try to maintain competitive advantages and spook Wall Street as little as possible!
2 Well, unless you includes stock options. Then we might be getting there...
3 Don't forget that OpenAI is apparently committing some $18B - $19B to the 'Stargate Project' itself (well, SoftBank's money, of course), and a huge amount of that will go towards not only servers, but construction and eventually power! When you go one level higher and hear $100B and $500B, that's why those numbers, while perhaps too pie-in-the-sky and more marketing, are even more incredible.
4 Well, except Apple. Though I would argue that sooner or later, they're going to need a bigger boat...
5 Even with inflation, that's around $35B -- less than the OpenAI round.
6 Technically their second time going public.
7 There are many other players getting paid along the way. And a lot of the spend is staying within Microsoft too. Though because of their OpenAI deal, it's been complicated from an accounting perspective. Now SoftBank can pay them via OpenAI! (And Oracle too.) One takeaway here might be that OpenAI needs their own cloud, powered by their own chips...