[ekjhgkejhgk]

In reference 14 we read

> However, what’s become clear is that OpenAI plans to pay for Nvidia’s graphics processing units (GPUs) through lease arrangements, rather than upfront purchases

I wish someone here could explain it to a dummy like me. Nvidia tells OpenAI: heres some GPUs, can you pay for them over 5 years. How is this an "investment" by Nvidia? That reference keeps calling this an investment, but what they describe is a lease agreement. Why do they call it an investment? What am I missing?

[sbuttgereit]

That Nvidia has to front the costs of the product at the beginning and arguably the risk that the allocation of assets end up not being paid off (bankruptcy, etc.) By carrying those costs early and the associated risk, Nvidia expects a return on that. If the risk is realized they'll lose but otherwise they'll gain. That has all the hallmarks of an investment.

[ekjhgkejhgk]

Thanks for your reply, but:

NVidia could protect itself against OpenAI bankrupcy by adding a clause to the lease saying that if OpenAI goes bankrupt, Nvidia gets its GPUs back. So the risk would only be that the lease would be aborted sooner than expected.

[tome]

> So the risk would only be that the lease would be aborted sooner than expected.

That is, in fact, the risk.

[ekjhgkejhgk]

But Im saying something different than the person I was responding to. They said that the risk was due to company going bankrupt and therefore Nvidia losing its "investment" - read: the GPUs that it leased. Whereas Im saying that the risk is due to company going bankrupt, Nvidia getting its GPUs back, but now they have too many at hand than they can usefully deploy/sell. The two are risks triggered by the same event, but the former is about 1 order of magnitude greater than the latter. The former is lost capital, the latter is lost opportunity - read: return on the capital.

[sbuttgereit]

In leasing to OpenAI, they aren't deploying those GPUs to other productive uses.

The costs are not just the raw cost of the physical asset, but also the opportunity costs of doing "thing a" vs. "thing b".

The GPUs will also depreciate during the time they're in OpenAI's possession, so again, if that investment doesn't pay off they aren't just getting GPUs back, they're getting "used GPUs" or "more used GPUs" back, not the original or full value of the originally leased asset. Naturally, the lease has to have those costs built in to be a good deal, but the lease has to fulfill to terms for that to happen.

In the end, leasing them means they carry the early risk of the lease not being fulfilled... but they should gain more in the than if they just straight up sold them.

[ProjectArcturis]

In the world in which OpenAI goes bankrupt, their used GPUs would be worth perhaps 10% of their value today.

[xnx]

How so? If OpenAI when bankrupt tomorrow, what percent of their value do you think their GPUs would be worth?

[foundart]

Speculation by a non-accountant:

At the end Nvidia retains ownership of what are probably very low value assets.

Contrast that with car leases: there is a robust market for used cars.

Nvidia is in effect financing the GPUs by not requiring the full payment up front.

Do the lease payments add up to the total cost?

[ekjhgkejhgk]

The GPUs do/might lose value over time, but perhaps Nvidia is betting that they wont? Going by recent years, their value went up in some cases.

EDIT: Interesting note

> CPUs historically have 5-10 years of useful life , while GPUs in AI datacenters last 1-3 years in practice , despite 6-year accounting assumptions.30,31 Evidence from Google architects shows GPUs at 60-70% utilization survive 1-2 years , with 3 years maximum.31 Meta’s Llama 3 training experienced 9% annual GPU failure rates , suggesting 27% failure over 3 years.31