SoftBank, Indeed

I'm old enough to remember when $6.6B was an unheard of amount of money for a startup to be raising. That is to say, I'm at least a few months old. But faster than you can say DeepSeek, OpenAI is apparently going back to the well. Actually, it's more like SoftBank is bringing the well to them. And well, it's filled with a lot of money. $25B, to be exact. Okay, technically "up to" $25B, The Financial Times reports. It could be as "low" as $15B, Bloomberg confirmed.¹ Regardless, if a deal gets done, it will be more than double that record-setting round.

For the sake of the argument, let's call it $25B – a number which also happens to exactly line up with the back-of-the-envelope math I did to determine how much more capital OpenAI would need to raise in order to hit their projections on getting to profitability. Granted, that was back in October, and a lot has changed since then – including, of course, the unveiling of 'The Stargate Project'. That is to say, I'm assuming OpenAI will actually now need to raise even more than $25B to get to their profitability goal. And with this money, Masa Son is probably going to push them to move even faster and think even bigger, because that's what he does. Still, $25B sounds like a nice little start.

Ten more things I think are worth calling out/thinking through/questioning here:

1. If SoftBank does the full $25B, that would take their total capital invested in OpenAI shares to $27B (SoftBank invested $500M into the primary round of funding a few months ago, and spent another $1.5B buying up employee shares). It would also more than double the amount of money that OpenAI has raised to date (around $20B). And, well, they need it.
2. Both of SoftBank's stock purchases last year were done at the $157B (technically, $156.6B) valuation. No word on valuation here, but you have to believe it would be some sort of up round given Son's desire to plow money in here. Historically, he hasn't cared too much about valuation, it would seem (see: WeWork). At the same time, OpenAI is obviously in a slightly tricky position at the moment given the DeepSeek situation. If nothing else, DeepSeek and other competitors aim to undercut their API business. Though their ChatGPT subscription sales remain more important to the business. Even a flat round would push the post-money here past $180B so let's go ahead and assume this round gets done north of $200B. SpaceX now sits at $350B and as much as Sam Altman might like to stick it to Elon Musk here, hard to believe this round could more than double the valuation set just a few months ago – even for Masa. ByteDance is closer to $300B,² but you have to believe that is probably also a bridge too far here.
   1. Update: A new report from WSJ puts the total amount being raised at $40B (with the $15B - $25B from SoftBank) and the valuation at $340B. That's an interesting number in the context of SpaceX above. Is this a don't-poke-the-Elon equation? Regardless, it would be a stunning rise in value in just a few months. Thanks, Masa!
   2. Update 2: Meanwhile, The Information puts the valuation at "around a $300 billion valuation including the new capital". Sounds like one of these sources may have gotten their pre-money and post-money details mixed up. I'm not sure how much it really matters at these stratospheric levels – again, unless you're trying to specifically pass, or not pass, another company in that regard...
3. No word on structure either, but given that OpenAI is still not yet a for-profit company selling actual equity in the business, presumably it will be done in a similar fashion to the last round, where the money would convert to debt if the company can't convert their structure. Then again, adding $25B in debt on top of the $6B+ from the last round would probably break the company. No one really seems to believe they won't be able to do the conversion, but, I mean, Elon Musk is going to fight like hell to try to railroad it – and now Meta and others are joining the fight against it. I'd still bet on it happening as I think it makes sense given how much the game on the field has changed (and despite the bad precedent it would set). Still, they might need to be more mindful in how to structure this amount of capital.
4. Assuming this money does buy an eventual equity stake in the company, it's hard to know exactly how much of OpenAI SoftBank would own because it's going to be insanely complicated to convert all the prior rounds into equity – namely because the non-profit will also convert into a major shareholder, perhaps the major shareholder. I've tried to spitball equity stakes before, but this massive amount of capital would put a wrench in that hypothetical cap table. Will this massive slug of capital even be enough to make SoftBank a 10% owner of eventual OpenAI equity?
5. They'd undoubtedly still be well behind Microsoft in that regard simply because of the earlier entry points of Microsoft's investments. Still, this would push SoftBank's money invested well past Microsoft's – more than double, actually: $27B vs. $13B.³
6. Yes, that would also be more money than SoftBank put into WeWork: $27B vs. $16B. They'll be hoping for a better outcome here...
7. It would still be slightly less than they spent to acquire ARM, which was roughly $32B back in 2016. (And is now a public company worth around $150B – with SoftBank still owning around 90%. They'll be hoping for that kind of outcome here...)
8. This new $25B is on top of the the "more than" $15B they've committed to Stargate, which is being set up as a separate company, where OpenAI and SoftBank will both initially own about 40% of the company.
9. To get that 40%, OpenAI will also be committing about $19B in capital to Stargate – money, which, as I've noted before, they don't currently have. They were clearly going to have to raise new capital to do that deal, I just don't think anyone thought it would also be from SoftBank. Because this means SoftBank is investing $15B+ into Stargate and also potentially investing $25B into OpenAI so that OpenAI can invest $19B into Stargate. That would leave just $6B going back to OpenAI's actual balance sheet. Again, they're going to need to raise again, even after this insane round!
   1. Update: Per the above cited new WSJ report with the valuation number, if OpenAI succeeds in raising $40B, that would obviously give them quite a bit more breathing room after the Stargate investment – $21B or so. Would that be enough to get them to the profit promise land?
10. SoftBank also doesn't have enough cash to do all of this investing! But when has that stopped Masa before? He'll undoubtedly raise debt to do it,⁴ which will also put SoftBank, which already has quite a bit of debt on their books, into a tricky spot if something goes awry.⁵ Thank god for ARM – though they're already leveraging that position too...

So yeah, there's seemingly a lot of nuance and quite a few potential headaches with this new mega fundraise for OpenAI. But, well, they need the capital. And this always felt inevitable. As I wrote last summer:

It seems likely that OpenAI is going to need a new major cash partner at some point. Maybe they hoped that would be Apple. Or maybe it will be the Saudis. Or even the ultimate backstop: SoftBank.

Nailed it. Well, if this deal gets done.

Should Apple Switch Off Siri?

In our Age of AI, Siri continues to embarrass Apple…

SpyglassM.G. Siegler

Post-Explosion, Seeking the DeepSeek Fallout

A guess: DeepSeek itself will matter less than the reaction to it…

SpyglassM.G. Siegler

¹ And while there is some chatter about how advantageous a leak of such infromation would be for a number of these players: OpenAI, Microsoft, SoftBank, etc -- especially in the wake of DeepSeek -- CNBC is also confirming the talks, so they all have either the same source(s), perhaps trying to manipulate a situation, or these talks are very real. Or both, I suppose...

² Though perhaps falling soon, at least a bit, by way of being forced to realize some of their TikTok stake in the form of a massive equity sale to... Oracle, Microsoft, and Elon? And the US Government? Someone? Anyone?

³ And yes, as we're all well aware by now, a lot of Microsoft's "money" was in the form of compute credits. SoftBank's capital will not be...

⁴ Though he could raise capital from various soverign wealth partners for this deal too. Perhaps even MGX, another Stargate partner started with capital from Abu Dhabi, just to further intertwine all the parties here...

⁵ SoftBank sure likes to net out their debt with footnotes galore. But overall debt seems to be around $100B currently.