Great writeup. There's one concern I have though. Let's assume progress speeds up even further and demand inflects more. Full bull case. Will Coreweave be able to benefit from near term demand changes? From what ChatGPT tells me, almost all of Coreweave's customers are on long term fixed price contracts. They have little spot price exposure. But I'm not sure if this is accurate. Have you looked into it?
I agree they will appreciate if demand rises, of course. I think the main source of increased revenue would be renewals. I'm unsure exactly how many contracts expire before 2028 though. In the S1 it said most contracts were 4 years in length I think. But that's now potentially out of date. I'll try to look for more recent figure later.
Yeah DM me if you find anything interesting, I'm still learning too. No upside is fine imo these are all likely hoppers and the main worry bears have is that their pricing goes down not up. New capactiy of rubins and blackwells should command several X hopper pricing so the scarcity will show up there when the new contracts are signed and will be massively larger than the old ones.
My understanding is that the typical LTA that the memory cos have been signing guarantee volume and include a price floor, but the price actually paid is not fixed. Leaving them open to the upside as prices rise. From what I can gather about Coreweave, they've signed mostly fixed price long term leases in order to lower the risk and get debt financing. So I think they're less levered to the upside than the memory names.
Did you just let chatgpt let you talk yourself into a degen yolo? Lots of words to justify a low probability strat. You know a pitfall of high iq is the ability to use elaborate reasoning to argue pretty much any point you like. It sounds like s possible outcome but I'm taking the other side of this, I think that AI capex is so far ahead of revenues, there will be a major reckoning as soon as venture capital (in a general sense of the word) dries up and they need to start paying debt and capex with revenue. A lot of the upstream suppliers have real revenue, but it is spending from the downstream co's that arent generating enough revenue yet and are instead relying on debt, venture capital, circular financing, or cashflow from other business units.
That all said if you are right, this is probably the highest leverage play. Im no expert, mostly just going off of vibes, theory, and the same data everyone else has.
Great writeup. There's one concern I have though. Let's assume progress speeds up even further and demand inflects more. Full bull case. Will Coreweave be able to benefit from near term demand changes? From what ChatGPT tells me, almost all of Coreweave's customers are on long term fixed price contracts. They have little spot price exposure. But I'm not sure if this is accurate. Have you looked into it?
New contracts, uncontracted capacity, and renewals. It's similar to the LTAs of the memory players. They will always benefit in the end.
I agree they will appreciate if demand rises, of course. I think the main source of increased revenue would be renewals. I'm unsure exactly how many contracts expire before 2028 though. In the S1 it said most contracts were 4 years in length I think. But that's now potentially out of date. I'll try to look for more recent figure later.
Yeah DM me if you find anything interesting, I'm still learning too. No upside is fine imo these are all likely hoppers and the main worry bears have is that their pricing goes down not up. New capactiy of rubins and blackwells should command several X hopper pricing so the scarcity will show up there when the new contracts are signed and will be massively larger than the old ones.
My understanding is that the typical LTA that the memory cos have been signing guarantee volume and include a price floor, but the price actually paid is not fixed. Leaving them open to the upside as prices rise. From what I can gather about Coreweave, they've signed mostly fixed price long term leases in order to lower the risk and get debt financing. So I think they're less levered to the upside than the memory names.
Yeah coreweave is a bet on the beta rerate for AI. Oracle is also a good secondary option here.
Even if it is a true commodity, it’s still going up. What happens when oil prices go up? Oil companies go up, yes?
So what happens when the demand for compute goes up and thus the cost? The compute suppliers go up.
Oracle is worth watching because they might lag the inevitable rerate. Even now they have stayed flat while nebius, Amazon and coreweave are up.
Did you just let chatgpt let you talk yourself into a degen yolo? Lots of words to justify a low probability strat. You know a pitfall of high iq is the ability to use elaborate reasoning to argue pretty much any point you like. It sounds like s possible outcome but I'm taking the other side of this, I think that AI capex is so far ahead of revenues, there will be a major reckoning as soon as venture capital (in a general sense of the word) dries up and they need to start paying debt and capex with revenue. A lot of the upstream suppliers have real revenue, but it is spending from the downstream co's that arent generating enough revenue yet and are instead relying on debt, venture capital, circular financing, or cashflow from other business units.
That all said if you are right, this is probably the highest leverage play. Im no expert, mostly just going off of vibes, theory, and the same data everyone else has.