If all goes according to OpenAI’s financial plans, Microsoft will close a $10 billion investment deal into the artificial intelligence startup before the end of this month, as Jeremy Kahn and I reported yesterday and according to documents seen by Fortune.
…
Microsoft’s bet on OpenAI appears to be even bigger than was previously known. The documents suggest that, prior to this deal, Microsoft had already poured $3 billion into the company—$2 billion more than has been publicly reported. If the current deal is completed at the figures being discussed, the cap table in the documents states that Microsoft will have contributed a total of $13 billion in capital to OpenAI, underscoring how important it believes the technology behind ChatGPT and DALL-E 2 is to its future.
…
Microsoft would be able to make as much as $92 billion from its collective investment, and venture capitalists that participate in the tender offer would be able to garner up to $150 billion. (An OpenAI spokeswoman declined to comment for this story, and a Microsoft spokesman didn’t respond to a request for comment.)
…
Microsoft will receive preferential treatment when it comes to OpenAI profits. The documents lay out how investors will be reimbursed once OpenAI starts posting a profit. “First close partners” will be reimbursed their principal first (it’s unclear whether “first close partners” refers to OpenAI’s early investors, Khosla Ventures and Reid Hoffman’s foundation, or other subsequent investors in the company). Once that has happened, 75% of OpenAI’s profits will flow directly to Microsoft until the sum that Microsoft invested in OpenAI is reached. Here is a graphical representation of how the economics are structured:
While the terms look like a win-win for Microsoft, it could end up being quite a while before Microsoft, or any of the other investors, see a meaningful return on that investment. Documents show that, as of the end of last year, OpenAI was projecting a loss of more than $508 million for 2022. The company has projected $1 billion in revenue in 2024, as was first reported by Reuters, but it’s unclear what it expects its costs to be in the years ahead. According to the documents, OpenAI expected that its costs in 2022 would total somewhere around $544.5 million.
If the deal completes, this investment is a big deal. OpenAI can afford to jump a few orders of magnitude of scale given a cash infusion that big.
This investment might represent potential evidence that OpenAI leadership believes that scale is indeed all you need.
[Note that whatever evidence this represents is pretty weak, until/unless they actually scale that far.]
If the cash infusion is completoon, it may be a reason to accelerate timelines (to the extent that one believes OpenAI intends to advance a few orders of magnitude of scale).
Alternatively, I don’t think people in the LW/EA sphere should have their beliefs on AI development and timelines basically unchanged upon learning that OpenAI received a $10B cash infusion.
Edit: the considerations mentioned here are significantly weakened if this is a multi-year investment.
not likely, turning into paperclips would be an incredible anti-adversarial-example success itself, and would imply we’re doing well enough to probably not all die. Also, I expect we are doing that well.
Gary Marcus talked about this in his newsletter today and reached sort of the opposite conclusion: that a deal like this might indicate that OpenAI leadership are bailing out now while the getting is good because the long term prospects for what they have are not looking great. Then again, he would say that?
Either way this is indeed a big deal (if true), whatever comes of it!
If this happens, there might be a greater-than-expected push to reduce token generation latency (currently a glaring UI issue), which translates into the initial LLM AGI speedup factor.
While this might be a great way to earn money (assuming competitors won’t invest similarly in AI soon enough), but aren’t there good reasons not to invest in AI capabilities, like reducing P(doom)?
Also I assume it’s wise to mention you’re not a financial adviser and don’t bear responsibility for actions people take because of your comment (same counts for me).
If we get turned to paperclips literally by Microsoft Clippy, that will be quite ironic.
not likely, turning into paperclips would be an incredible anti-adversarial-example success itself, and would imply we’re doing well enough to probably not all die. Also, I expect we are doing that well.
FYI from a site norms standpoint, I think it’s fine for this post to exist. I’ve tentatively left it on Personal Blog rather than Frontpage though.
Fortune reported on the investment plans and I rewrote the post accordingly.
Gary Marcus talked about this in his newsletter today and reached sort of the opposite conclusion: that a deal like this might indicate that OpenAI leadership are bailing out now while the getting is good because the long term prospects for what they have are not looking great. Then again, he would say that?
Either way this is indeed a big deal (if true), whatever comes of it!
Bloomberg reported on plans to invest $10B today
Thanks, added to the OP.
If this happens, there might be a greater-than-expected push to reduce token generation latency (currently a glaring UI issue), which translates into the initial LLM AGI speedup factor.
If this goes through I would strongly advise you to buy MSFT stock. I bought a ton when the possible deal was announced.
While this might be a great way to earn money (assuming competitors won’t invest similarly in AI soon enough), but aren’t there good reasons not to invest in AI capabilities, like reducing P(doom)?
Also I assume it’s wise to mention you’re not a financial adviser and don’t bear responsibility for actions people take because of your comment (same counts for me).