J.P. Morgan's OpenAI loan is strange
marketunpack.com
marketunpack.com
A company that has revenues and is extremely well-capitalized gets debt finance. That is not news. That is totally commonplace. "Shouldn't all their capital come from investors?" No. Companies at all stages typically use a mixture of debt and equity finance.
His EV calculation is completely flawed also. Debt finance is typically senior to equity in recovery at bankruptcy, so when JPMC do this analysis (and believe me they did this analysis) they are not assuming 0% recovery. They are thinking it is most likely in a bankruptcy that they get some x>0% recovery.
Finally, banks don't think about their relationship with a multi-billion-dollar company in terms of the ROI on a single revolving credit. (even though this will in all likelihood be very profitable for JPMC). They think about how giving this revolving credit makes it more likely they get advisory on any future bond issuance and I-banking work when OpenAI want to do takeovers, and a foot in the door at IPO time etc.
Comparatively speaking, I don't know why investors won't touch DOCN with a 10-foot pole, but I will gleefully reap my returns from dividends and stock buybacks when DOCN laps AWS in 20 years with better services maintained by better engineering staff.
I think we're well beyond the point where OpenAI can be called a "startup."
Different dictionaries provide different definitions.
A common one is that it's small and recently started business, but it's a very vague boundary.
Debt is almost always cheaper than equity. Particularly if you can collateralise.
Well-capitalized companies rejecting debt is more of a Silicon Valley outlier in the global economy. (It likely stems from dot-com trauma.)
Not really. Tech, including low fixed-cost software, has been tremendously capital intensive for decades. Early-stage start-ups lack the cash flows to fund debt. But post-Series B companies raising equity are generally doing so for idiosyncratic reasons, e.g. capital sponsors being concentrated in equity for historical reasons, valuation escalators and capital denial to competitors.
I may have missed something in my career on Wall Street, as a founder and in VC.
(Being wrong is fine. Being confidently wrong is dumb.)
> Many tech companies are started out of their founders apartments for essentially 0 startup cost
You’re mixing up fixed costs and capital. Both fixed and operating costs consume capital. (We call the latter working capital.)
(You may also be mixing up PP&E and capital.)
> a reason that tech founders get so much richer than founders in other industries and that reason is because the minimal capital requirements allow them to sell off so much less of the company before reaching massive scale
This is wrong. Obviously wrong.
Tech founders get richer because their companies get bigger. Apple, Tesla, Google and Saudi Aramco have massively different capital requirements. Their owners (and founders/founding lineages) are in the same ballpark.
Similarly, most family businesses never sell equity until they sell the business. And most tech founders don’t have a majority of equity after a couple rounds.
IMO it's because CFO is treated as a second-tier C-suite role in Silicon Valley and a lot of the CFOs are completely substandard as a result.
A lot of CEO have the mentality that if the product and the tech works the money sorts itself out, so your CFO is already behind the CTO and the CPO.
This Forbes article has a good overview: https://www.forbes.com/sites/amyfeldman/2023/03/19/silicon-v...
The $4B revolver will likely sit undrawn. When it gets drawn, there usually a specific plan to reduce it back to zero. It's not for building data centres, a revolver typically used just for timing differences like a credit card is used (and the lenders will be paying attention). Also, when things get bad, there are covenant triggers which would allow lenders to renegotiate.
> OpenAI is going to generate $3.6 billion in revenue this year, but the costs will lead to a loss of more than $5 billion.
We know that actually:
> OpenAI generated around $4.3 billion in revenue in the first half of 2025... OpenAI said it burned $2.5 billion
> OpenAI looks to meet its full-year revenue target of $13 billion and a cash-burn target of $8.5 billion, the report added.
https://www.reuters.com/technology/openais-first-half-revenu...
Well I am personally, and the last 2 places I've worked have too.
Worth every cent!
2) It's a revolver, it's not all being used
3) If things go great and OpenAI ends up buying smaller guys or getting bought out (probably MSFT?) then JPMC will be right in there with those young bankers who don't sleep. They will pull in many many millions in fees with very little expense.
4) If things don't go great, OpenAI will be looking for more financing. Guess who will help them?
5) It's really only in case there's an Enron things are terrible for JPMC. Like if it turns out the whole thing was a bunch of guys in India answering every ChatGPT query, something like that. If there's actually an AI business, and despite JPMC's history of due diligence misses (Javice case) that's probably the case, then there's deals to be done.
Who? I can only think of the Saudis/UAE and SoftBank.
As we've seen from the NVidia/Oracle and AMD deals, there is more than one way to structure investments. Financing doesn't just mean "deposit cash into my bank account:.
If OpenAI folds, there are two basic scenarios:
- one: the LLM crash has come, and OpenAI barely has any material assets. Microsoft isn't putting more money into it, and they won't take it over -- people with seven figure salaries will be looking for six-figure jobs.
- two: somehow, only OpenAI crashes, and the rest of the LLM boom continues. This likely involves OpenAI being extraordinarily outcompeted, so it's a long slow decline as contracts run out and are not renewed. 40% is probably high, but if JPM can retract the revolving debt before it all goes out the door, not ridiculously high.
They have claims on the use of assets via their leases.
More directly, if AI crashes in the next 2 years, they get bailed out. Between OBBA, tariffs and immigration enforcement, the American economy less AI is probably already in a recession. Trump and the GOP would get desperate if the political leash the AI boom has granted them is shortened. Borrowing another few trillion dollars to fix it would be worth the gamble.
Bluntly: nobody in the Trump administration has the ability to soft-land the economy. When the ship is undeniably sinking, the rats will flee.
The problem is that AI is ultimately a luxury good. Even at it's currently ultra-subsidized rate, it's hard to justify. If it gets any more expensive, no consumers will buy it.
Which might be fine, if we turn to enterprise. Except: a problem. Companies produce goods, which people then buy. AI can replace labor maybe, but it can't replace consumers.
This means the number is wrong.
Depends what other creditors OpenAI has, and the priority on their claims.
>For some reason I have yet to comprehend, when I was sitting on a credit desk pricing CDS, they always used 40% as the recovery rate.
That doesn't mean its the actual amount they will recover, just that it is sonething like the expected average for a large class given the criteria JPMC uses when entering such a position and the anticipated range and distribution of outcomes when those firms fail.
A quick Google search tells me:
> Generally, a U.S. company can sell assets below cost, as long as the transaction is transparent, conducted in good faith, and serves a legitimate business purpose.Which is exactly the same argument that the author made, no?
> Cost: $1,000 Case 1 (90%): OpenAI goes bankrupt. Return: $0 Case 2 (9%): OpenAI becomes a big successful company and goes 10x. Return: $1,000 + 5% interest = $1,050 Case 3 (1%): OpenAI becomes the big new thing and goes 100x. Return: $1,000 + 5% interest = $1,050
The actual math is that if OpenAI succeeds, then there's a nod and a wink that JPM will land the lead role in the IPO or any mergers/acquisitions, which translates into huge fees.
This isn't a financial transaction. This is a "relationship" transaction.
This is the easiest money and best relationship JPM could imagine
Wow, that's slightly more than Yahoo has. Well, had.
We don't know how ripe OpenAI's industry or market position is, yet. Yahoo knew what they had lost pretty early onto its spiral.
I’ve worked for enough startups that even if your company folds and goes bankrupt with no business plan the ip generally can easily cover the outstanding loans.
They have no moat. So, how do you get to $4B?
I think the models are wrong way too often for relatively simple queries, so unless they give a secret prompt like "be wrong a lot in the free version" to users, it's basically worthless.
This is a huge assumption.
I am not sure how differentiated OpenAI's IP is, so I don't have a strong opinion, but it seems to me that OpenAI is worth a lot more as a package than it would be as a collection of pieces.
Well maybe the employees will be worth something.
Naa that takes too long to get any value from. If openAI goes pop, their IP isn't going to be worth much, because it'll die from competition, or the economy going to shit.
The article linked[0] is from last year.
A recent article[1] from this year says "OpenAI looks to meet its full-year revenue target of $13 billion and a cash-burn target of $8.5 billion, the report added."
[0]https://www.reuters.com/technology/artificial-intelligence/o...
[1] https://www.reuters.com/technology/openais-first-half-revenu...
No one is OpenAI’s financing while anthropic et al. keep raising. The big risk is that future innovation fails to live up to the hype and they can't afford full priced GPUs or the proposed datacenters.
They've made 8 billion this year with 800 million users, or $10 ARPU (average revenue per user). They have committed to spend a trillion dollars over the next 5 years. I'll call it $200 billion/year, with a (rounded-up) billion active users, they would need to make $200 ARPU . For comparison Meta has about $50 ARPU. I'm having a harder time finding Google's ARPU, but with $350 billion in revenue last year if they made $200 ARPU they would have less than 2 billion users (I can't find how many user's they actually have but I would bet money it's a lot more). They would have to be making 3-4x more money per user than two of the largest companies in the world for this bet to work out.
I don't see how this is going to work out for them.
The loan value in case OpenAI goes bankrupt depends on the details of the deal.
I'd loan them the money for that opportunity. It's possible that the IP will somehow turn out to be worthless; I don't know what will happen, but I am confident ruling out worthless. Could the lender could refuse Microsoft's payment - from a minority shareholder? OpenAI could accept Microsoft's money and pay the bank; with that offer, OpenAI would have leverage to play Microsoft and the lender against each other.
OpenAI could go bankrupt for many reasons having nothing to do with the value of their IP. For example, they could overinvest in developing it - a positive outcome for the lender.
You do realize other countries are working on this too, right? The biggest one rhymes with INA.
Anyway, J.P. Morgan's OpenAI loan isn't strange, it's calculated exposure to Microsoft.
The bank's risk seems quite a bit lower than the VC's risk.
Plus the post seems to only include 1 year of interest.
Unless we know the terms, I don't think we can necessarily calculate EV from JP Morgan's perspective. I would say that they aren't usually carelessly giving away money though... They probably have terms where they can get out early if OpenAI's position weakens etc.
I agree but had different questions. TFA mentions the consideration of whether failure cases are correlated, but of course if OpenAI wins big, there's a good chance this directly or indirectly creates much instability and uncertainty in many other loans/partners. What's the EV on whether that is net-positive considering this is a loan at 5% and not an investment?
On the other side, if OpenAI crashes hard, is it really such a sure thing that Microsoft will be the on the hook to pay off their debts? Setting aside whatever the lawyers could argue about in a post-mortem, are they even obligated to keep their current stake / can they not just divest / sell / otherwise cut their losses if the writing is on the wall?
Google, Amazon, Apple, MSFT, Meta & Nvidia didn't face "China risk", but AI firms do. As Tik Tok saga has proven, the days of running large losses for years, to recover it all back in IPO, and monopolize thereafter, may be over.
China might win the AI race and force the hand of USG for yet another national security imperative, to expropriate foreign assets.
- existing models are profitable if cutoff future training and focused on inference
- debt is senior to equity
- if my life depended on one company not going bankrupt over the next decade I’d pick OAI over Citibank
- banks use revolving credit as a break even or loss leader for higher fee business
- high fixed cost businesses use debt and equity to scale
- lead investors would very rarely pay down the debt of an investment, that’s not the backstop
- unlikely for revolving credit, but a convertible structure could mitigate any perceived asymmetric downside
Even for a very simple model, this seems unrealistic. Isn't the whole point of a loan that your liable to pay it back if you're in any way able to? (very roughly)
So the bank would have claim to some of OpenAI's existing assets even if they went bankrupt. The probability that they not only go bankrupt but also have no assets left at all would be much lower.
> 0 of 0 matches
Some analysis.
Do we know they're selling their services below cost? I'm pretty confident they're making money on inference and burning through large piles of cash on capex and research.
You can basically throw the entire analysis out on a single point:
>Case 1 (90%): OpenAI goes bankrupt. Return: $0
It won't be $0. Creditors have liquidation priority and banks make very sure they're confident in the quality of the collateral before handing out billion dollar facilities.
OpenAI is massive, fairly risky, associated with Microsoft, etc. all true. What matters to JPM is potential future business. There’s potentially an enormous IPO in the future. The credit line is just good business. They are fostering the relationship.
(Also, a revolving credit facility is not a loan, which this discussion misses)
Taken from the About page:
Who are you? Hi, I'm the author. I've been dabbling in investing since 2015 and I decided to get more serious about it in 2023.
Is this financial advice? No. In fact, everything you read here is be half-baked by design. If it were fully-baked, then I wouldn't have felt the need to write about it in order to distill my thoughts.
I would not be happy with this trade if I had any $JPM.
While, yes, it's overhyped and takes up too much of the spotlight, it isn't going anywhere and is going to continue to be a staple in our lives moving forward, and will continue to increase in impact as it improves.
Blockchain will always be blockchain, and it was absolutely a bubble. Will it improve? Sure. Will it ever be as impactful to the everyday individual as AI? Maybe, but most people will not know or care. It will mostly be a passive experience.
AI will/has fundamentally changed how we work in the matter of 3 years. So, financial system jumping on board this, is much safer than jumping onboard crypto hype trains.
(I'm not saying crypto is bad / is a bad investment, but lots of crypto and ai startups are both trash, well, most startups are trash, I say that as a founder who has had many trash ideas)
The past 3 years or the next 3 years or 1.5 years both ways?
It hasn't really done that yet. The work from home zoom meetings has made a bigger impact. Maybe in a few years? Its on par with the metaverse at this point.
You do remember that by the time the blockchain bubble burst, literally everyone and their mother were "interacting" with it?
> AI will/has fundamentally changed how we work in the matter of 3 years
For people in bullshit jobs creating workslops, yes its probably an absolute blessing. Enjoy it while it lasts.
Duh. Basically the author wrote a word salad to say the bank calculated the probability of them losing money would be lower than what the author pulled out of their ass at the beginning of the article, refuting their own hypothetical point.
In the 1% scenario, JPM could be looking at tens of billions on the upside, if this loan secures them the lead.
People have to stop saying this with NO evidence to back it up. And by evidence I do not mean random investors opinions, anonymous "insider" infomation, etc. Give numbers. Saying the same thing 1000x doesn't make it true.
Show me how a company setting never before seen piles of money on fire is profitable. Until someone can i'll happily keep claiming they're unprofitable.
Obviously we cannot assume every business is profitable unless proven unprofitable. That's why reporting and audits exist.
Note that they totally could be! I'm not asserting one or the other. But unprofitable is the default, absent evidence.
Indeed, the other poster seems right: "The burden of evidence is on the side of the person making the claim" is simplistic and reductionist. "The burden of proof lies on the more unbelievable side" is more appropriate.
Taking an example: If I said the earth was not flat, and someone else told me to prove it, that proof would be unnecessary, because the earth being flat is more unbelievable than the alternative.
Oh, so you are just trolling. Got it.
of course normal people edit their post within the first couple minutes of making it.
maybe wait longer before replying? try 2 hours if "ninja edits" are a problem for you.
at that point, the post parenting your reply will be un-editable, so you need not worry.
but anyhow, back on topic: no evidence of profitability implies unprofitable by default.
>A recent article[1] from this year says "OpenAI looks to meet its full-year revenue target of $13 billion and a cash-burn target of $8.5 billion, the report added."
[1] https://www.reuters.com/technology/openais-first-half-revenu...
If it turns out that OpenAI becomes wildly profitable, they’ll just shift to some other argument about the environment or the specialness of human thought or whatever.