Everyone wants AI to be the next cloud boom until the bill arrives. Betting tens of billions on one customer whose own business model is still being debated is bold. If demand keeps exploding Oracle looks brilliant. If not, this could become the case study every finance class uses.
It’s not unusual, but it’s kind of a big deal for a corporation as large as Oracle to be doing it because it means they’re hiding tens if not hundreds of billions from showing up on the bad side of their account charts.
So they’ve got negative cash flow so that they could show epic growth, but the growth itself is hella juiced because the GPUs are only relevant for about 3 years till the new ones are out and make more AI for less power. And they depreciate them over 7 years. More than twice as long as they can or should use the GPUs for.
the growth itself is hella juiced because the GPUs are only relevant for about 3 years till the new ones are out and make more AI for less power. And they depreciate them over 7 years. More than twice as long as they can or should use the GPUs for.
We don’t actually know this for sure, yet. I had expected the A100 generation (released in 2020) to no longer be profitable to run by now, but the backlog in new data centers being turned on and the high demand from Anthropic and OpenAI still leaves those chips useful for inference. You can rent those 2020 chips out today at some price above what they cost to continue running (300W, so electricity prices of USD $0.20 per kWh would translate into about 6 cents per hour. Prevailing spot prices appear to be about $2/hour right now.
But just because I was wrong on 2020 chips, originally sold for about $15,000 in a low interest rate environment, doesn’t mean that I’m wrong about 2024 chips, the B100s that use 1000W and were sold for $35,000, requiring a ton more specialized cooling, power, and network infrastructure. Or the 2026 R100s that use 2000W, and whose prices I can’t seem to find published anywhere, but were set after the memory companies basically locked in their record breaking prices for their HBM. That’s an unsustainable path and at some point, data centers start struggling to find users willing to pay the bare minimum necessary to continue turning a profit on GPU usage.
I doubt the 2024 chips stay in service to 2031. And I’m really, really skeptical that the 2026 chips stay in service to 2033, especially after NVIDIA switches to yearly release cycles next year.
Speculating on someone who themselves are speculating that their customer’s speculation will be fruitful is exactly what low interest rates encourage. Everyone likes the extra activity that produces on the macros, but it just makes the market less accurate.
Interest rates basically set how permissive the market is with its todo list. The problem is when you let yourself do anything, and then also let yourself get distracted doing something that might help something that you maybe shuldn’t be doing anyway. This is not a functional life. It’s a busy life, but not functional.
If you want to kill AI, raise the interest rates. It will cause a recession, which sucks. But frankly, humanity will survive a recession and it will be temporary. A higher interest rate market solves real problems that benefit actual people within a near horizon. What we don’t need is people working 14 hour days, slaving to keep their heads above water, working for companies engaged in far-off speculative race for world dominance; we wouldn’t want them to succeed at anyway. That’s 100% what we’re all doing, and it is interest rates that modulate that.
Raising interest rates will kill all of those jobs involved in that, and sadly a few more from the shock. But those people weren’t doing anything productive for humanity anyway. It’s better to have them lose those jobs and then later redirect their labor towards something productive. In the long term they are digging themselves out of having a job with their current path anyway.
Everyone wants AI to be the next cloud boom until the bill arrives. Betting tens of billions on one customer whose own business model is still being debated is bold. If demand keeps exploding Oracle looks brilliant. If not, this could become the case study every finance class uses.
They’re also juicing their numbers. They’ve got obsolete GPUs on the books for years after they’re irrelevant.
Corporations lying on their books to decieve investors? Shocking!!!
That only happens on days that end in Y.
It’s not unusual, but it’s kind of a big deal for a corporation as large as Oracle to be doing it because it means they’re hiding tens if not hundreds of billions from showing up on the bad side of their account charts.
So they’ve got negative cash flow so that they could show epic growth, but the growth itself is hella juiced because the GPUs are only relevant for about 3 years till the new ones are out and make more AI for less power. And they depreciate them over 7 years. More than twice as long as they can or should use the GPUs for.
We don’t actually know this for sure, yet. I had expected the A100 generation (released in 2020) to no longer be profitable to run by now, but the backlog in new data centers being turned on and the high demand from Anthropic and OpenAI still leaves those chips useful for inference. You can rent those 2020 chips out today at some price above what they cost to continue running (300W, so electricity prices of USD $0.20 per kWh would translate into about 6 cents per hour. Prevailing spot prices appear to be about $2/hour right now.
But just because I was wrong on 2020 chips, originally sold for about $15,000 in a low interest rate environment, doesn’t mean that I’m wrong about 2024 chips, the B100s that use 1000W and were sold for $35,000, requiring a ton more specialized cooling, power, and network infrastructure. Or the 2026 R100s that use 2000W, and whose prices I can’t seem to find published anywhere, but were set after the memory companies basically locked in their record breaking prices for their HBM. That’s an unsustainable path and at some point, data centers start struggling to find users willing to pay the bare minimum necessary to continue turning a profit on GPU usage.
I doubt the 2024 chips stay in service to 2031. And I’m really, really skeptical that the 2026 chips stay in service to 2033, especially after NVIDIA switches to yearly release cycles next year.
I blame low interest rates.
Speculating on someone who themselves are speculating that their customer’s speculation will be fruitful is exactly what low interest rates encourage. Everyone likes the extra activity that produces on the macros, but it just makes the market less accurate.
Interest rates basically set how permissive the market is with its todo list. The problem is when you let yourself do anything, and then also let yourself get distracted doing something that might help something that you maybe shuldn’t be doing anyway. This is not a functional life. It’s a busy life, but not functional.
If you want to kill AI, raise the interest rates. It will cause a recession, which sucks. But frankly, humanity will survive a recession and it will be temporary. A higher interest rate market solves real problems that benefit actual people within a near horizon. What we don’t need is people working 14 hour days, slaving to keep their heads above water, working for companies engaged in far-off speculative race for world dominance; we wouldn’t want them to succeed at anyway. That’s 100% what we’re all doing, and it is interest rates that modulate that.
Raising interest rates will kill all of those jobs involved in that, and sadly a few more from the shock. But those people weren’t doing anything productive for humanity anyway. It’s better to have them lose those jobs and then later redirect their labor towards something productive. In the long term they are digging themselves out of having a job with their current path anyway.