Writing at about AI: https://xitter.cf/t.co/wfSaW4n514 Book a call to talk compute markets: https://xitter.cf/t.co/9XQJFrdeB1
NY, NY
Joined June 2008
- Tweets71.1K
- Following1.6K
- Followers6K
- Likes208K
Incentives drive outcomes.
The hope is the borrower defaults, the lender acquires title to the GPUs, and the lender suddenly owns scarce GPUs that it can sell on the secondary market, rent to other people, use for its own purpose, etc.
Incidentally, this suggests that at least some lenders are not interested in residual value insurance. Why buy residual value insurance if the play is to let borrowers default on loans so you can acquire title to your collateral?
This reads as someone who doesn't understand how this kind of quote will be received by others. Which, frankly, seems to be a recurring pattern with Mayer.
Agreed with this; she was correct and prescient.
Surprised at the amount of heat her comment generated. My bet is it has something to do with the unreconstructed optimism that pervades startup/AI world, which optimism I think tends to overweight 'the future is so bright I gotta wear shades'-style epistemology over, well, realism.
Hey chat, I think @Melt_Dem is owed an apology after she was lynched for saying there's a chance we might have a compute credit apocalypse coming soon.
Good overview of some of the financial issues that neoclouds are running into.
What
Nicki Minaj Floated As Press Secretary Replacement By White House Staffers tmz.com/2026/10/25/nicki-min…
Yeah this makes sense. Locking in forward compute prices makes lenders rest easy since neoclouds’ customer base (ai startups) are not themselves creditworthy counterparties. For a lender, quality of contracted revenue matters more than contracted revenue.
If you really want to annoy the Bay Area AI contingent, tell them that AI is physical infra and commodities with a thin software layer on top.
The statement has the virtue of not only being correct, but of being a rebuttal to the Andreessenean software-eats-the-worldism that has become so au courant amongst the silicon set.
This is fascinating.
(I don't have an informed view either way, and even if I did, I am not much interested in participating in betting contests. *That said* I do think bets are a good way to price events.)
Two greats join our bet:
- @dwarkesh_sp on the explosive-growth side
- @Thom_Wolf on my side
benjaminmoll.com/growth_bet/
Total pot now $700k 💰
Took a ~2 week break from the gym. Back today and I’m stronger than ever. Recovery and recuperation is a thing…
I see this is as a correlation risk issue.
Most of the neoclouds have startups as customers.
Startups tend to rise and fall in unison.
So if most of the neoclouds' customers all fail around the same time, the neoclouds' debt all goes bad at the same time.
And it's not like the lower-tier neoclouds can turn around, call Google and say "hey guess what we have spare capacity, come buy some!" CoreWeave and a few others can sell to hyperscalers.
The hyperscalers are flush with cash, obviously, and want to get as much compute capacity as possible, but they also have requirements that most (all?) lower-tier neoclouds can't and won't meet.
Net result: a lot of bad loans coming down the pike, bankruptcies, consolidation...
