Kinda how we built DigitalOcean in the early days, the software was venture funded but the HW was funded as lease lines and credit against the HW we owned originally + a decent business plan based on the growth. It's a good(/slightly scary) strategy! Highly recommend more folks look at credit facilities.
It's very different types of money. Both are dangerous but vc money is considerably less dangerous (imo). If you take someone like Fortress as your capital group (credit line) you're going to lose the company if you f up, period. With VC, you have more latitude to have a conversation. We ran a very difficult path to build DO, I'm sure a lot of people lost sleep many nights, but if you're very sure about the business, credit facilities can be helpful, just know they're not joke. Neither is better or worse than the other, they're different but both useful. My point was if you are sure about the business credit facilities can be really helpful and are under explored for many startups (granted HW is often a factor in credit)
VC money is super expensive. Banks are the cheapest option, especially if you own property you can just use that that get a bunch of working capital. Any capital for equity should if you can swing it be done as convertible loans so that you at least have the option to pay it back if that is in your best interest. It also keeps the cap table clean and you (the founders) solidly in control until you decide to change that.
I've seen more than one case where a company was looking to do a round but they had not even considered the option of taking out a bank loan against their inventory. Especially for short term liquidity that's much, much better than VCs buying stock.
it's expensive in the long term, if your company is successful. In the short term, it is cheap, because you don't pay interest on VC money, which makes your cashflow much better.
Bank loans/bonds are expensive up-front, because you are obligated to service interest. It's cheap in the long run because once your business generates revenue, you can use them to pay off the loan, and slowly wean it off, or find an even lower interest loan to refinance off.
Of course, a bank will only lend if you have collateral, while VC will "lend" if the business idea is good.
I think you have that reversed. For the past few years, credits if you could get them were basically free money given the very low interest rates and the tax shield. Meanwhile VC wants shares. You are actually selling parts of your company, it’s not only about control.
That works as long as you maintain that growth. Too long a bump and you're waiting tables ;) But at least that didn't happen to DO. Being create with financing is probably a very important component in starting companies that have a lot of capital expenses to make, and it is the kind of thing that your typical tech founder usually has little experience with.
It's a routine financial transaction. Often these are structured as leases, where the lender remains the owner. It's common for aircraft, locomotives, etc - things with long, predictable lifespans, insurable against damage.
The problem for the lender is that it's a bet against progress in the field obsoleting the thing. If the user can cancel and send the thing back because there's a more cost-effective model, the lender has a problem. That mistake hurt Lloyds of London in the 1970s.[1]
I don't know the details of this lease, but it is likely a capital lease, not an operating lease. The lessee will not have the option of returning it. It will be recognized as an asset on the lessee's balance sheet. The lessee backs the lease not just with the collateral, but with the full value of the capital structure up to any debt senior to it.
It's a bet against progress in the field only to the extent that they depend on the collateral if the lessee fails. Lessors hate getting collateral back. So they've probably looked at the lessee's capital structure and decided that, along with whatever interest rate they're getting, it's a reasonably priced risk. They're likely mapped out their estimated risk of default by year along with a declining recovery value of the collateral by year.
Why does it feel like this story will be depicted in a future Adam McKay film, and involve Steve Carell or Ryan Gosling on the phone, talking to a loan officer, maybe interposed with a cutscene of Selena Gomez explaining what the hell an Nvidia H100 is to the audience?
The funny part to me is they'll likely use that money to buy more H100's.
> Last month, Inflection AI built a supercomputer worth hundreds of millions of dollars powered by 22,000 NVIDIA H100 compute GPUs.
What? InflectionAI has relatively little brand recognition, do I not know something about them, are they just the first in line, or is nvidia just selling this many cards in general? I can only imagine all the typical ones like lambda, AWS, openai, apple, <insert large tech company here> are buying even more?
EDIT: Wait if the collateral is for a 2.3B USD loan - how many cards do they have?!
It really depends on the type of financial institution that has given the loan. A bank for this type of transaction is very unlikely nowadays, so I guess it is a private credit fund. Given that they have probably applied some sort of traditional Loan-to-value ratio, it's very likely that the $2.3 billion is only a portion of the total value. Meaning the cloud provider has received e.g. 50 or 80% of the total value of the H100's as a loan. It surely is not very much since electronics have a high depreciation rate from an accounting perspective (3-5 years max).
Next week's news: Nvidia prohibiting collateral loan agreements as part of their partner agreements. The last thing they want is a glut of these hitting the market due to bankruptcies.
But their sales team or other execs who have incentives to keep sales prices inflated for as long as possible will fight against such a move.
Even if a few companies go bankrupt, the demand for AI training hardware is absurd and likely won't go down any time soon - and unlike all the shitcoins, there is no (viable) threat of ASICs outcompeting GPUs.
When you use the fixed pipeline to render triangles they are, but when you use the general purpose compute pipeline to do stuff like multiply tensors and apply arbitrary functions on their elements, they aren't application specific.
Possibly longer than some stocks against which rich people take loans. Much stupider things have happened in the credit market.
Also, the Reuters article says that they have a depreciation schedule built into the contact. Compute hardware can depreciate pretty quickly, if I remember correctly major tech companies have around 3 years planned for the depreciation.
> if I remember correctly major tech companies have around 3 years planned for the depreciation.
Historically yea, though now I think it's being stretched to 5+ years as they see hardware last longer in production (but you're atiop correct, it's very quickly relative to a lot of other loan collateral, still a short term deal even if it's 7 years)
I expect they'll lose value in the step-function shape that's typical of technology, where each new generation decreases the value of the previous one. But that's fine, as long as each chip generates more revenue over its lifetime than it cost. And the lifetime can extend through multiple generations, as long as each successive generation is only a marginal improvement over its predecessor.
Personally I think the bigger risk is software innovations making CPU training (and/or inference) sufficiently viable that it's cheaper to train models on a commodity CPU cluster than on some proportionally expensive GPU cluster. I don't know enough about the space to say whether that's likely, but it seems like a low risk, since pretty much any parallel algorithm will always be faster on GPU than CPU - it's just a question of the marginal benefits and cost (e.g. maybe it takes more CPU to train same model in same time, but cost of CPU is so much lower that it's worth buying more of them).
Even the larger V100 instances are still heavily used at AWS/GCP/Azure. The reality is that demand does not disappear if there are no H100/A100, it just finds another way.
There's also the fact that if you are not training LLM, you can get a better deal using some older hardware.
But some value as long as computing power can be sold above running costs(power, cooling, etc.). Good question is when the newer model is so much more efficient it makes sense to replace them.
Lease deals like this tend to be structured as an amortizing loan over 3-5 years. They also probably did not finance the entire value, but the entire value will be collateral. If they financed, say, 70%, then a couple years in the remaining exposure may be well under 50% of original price; four years in, they likely need only a low recovery to become whole.
That's precisely my point - even very conservative financing you're looking at collateral value that falls off a cliff in what a year vs a 3-5 year loan.
Worse the collateral side of the equation is currently in max AI hype bubble frenzy pricing while the loan is well what it says.
Very much feel like we're getting half the story...specifically the part that makes for good PR "Look at us we have lots of H100s and they're really valuable".
surprised they couldn't get asset-backed financing for the new build. wonder if they just took a better rate this way. ceo says cheap way to access debt so i guess so
