Brian

I grill and smoke a lot of foods, and I haven't found a reasonable substitute, not even the indoor smoker boxes. If you find one, I'd like to know.

Feeding some vegetarians this weekend, it's supposed to snow, and I want to make baingan bhartha, which depends on grilled flavor.

Ok, well, I was off today for the federal holiday, but the gym was open, and this picture shamed me into deadlifting 365.

The fairy dust is thick in the Silicon Valley air these days...

That assclown is gonna be Fed chair. Jesus.

I tried to back down off that peak intensity a little, take more gym time for myself, try to impose better balance. Turns out they didn't like that, management wanted ogre geeks in the nerd fighting pits. That's the culture.

I don't think there's a good way to find out definitively. Problem is, most data centers aren't public companies, and most that are don't break out their AI-specific workloads.

A lot of the people I knew who cashed out did things like buy an old farm in the mountains to grow blueberries, or be kindergarten teachers.

My floor was mainly core debuggers. It was core dumps where you were the last hope to fix a problem. I think 5 of 30 people one year had heart attacks.

Jesus Christ, I worked there a long time, and a Principal in any role is a hard, hard gig, much less being one twice. I made it to one step short of Principal, and the hill just got too steep to take another step.

Coreweave said the same thing today. Need government involvement in power generation.

Yeah, because they're bagholding tons of depreciating GPUs bought with borrowed money for data centers that haven't even broken ground yet.

Remember who would approve a bailout. Trial balloon for Trump. Without the largest data center public company, OpenAI can't even think about meeting its $1.4 trillion spending commitments.

You know the same thing is likely true across the entire data center ecology. They're not charging enough.

Kind of his brand at this point, given how credulous the tech and financial press have become.

Next, Adam Jonas at Morgan Stanley bumps Tesla price target by $1000 based on robot guardian sales worldwide.

What horseshit. Charlotte is one of the safest, most chill places you could be.

That was a disastrous earnings report from CRWV.

Yes, revenue more than doubled, but operating expenses nearly tripled, operating income fell by 2/3rds and interest expenses tripled. Debt service is 6 times operating income.

Kind of wondering if this was the source of the freakout last week, wanting debt backstops and government 'involvement' in data centers.

Data centers are the weakest of several weak links in this value chain. That was frankly an awful earnings report.

Very mixed quarter from Coreweave.

Revenue more than doubles, but operating expenses nearly triple, operating income falls by 2/3rds and interest expenses triple.

Debt is 3.81 times equity. Debt service is now 6 times operating income. Things have to break just right for them.

Translation:

"I'm drowning in high-interest debt and stuck holding these depreciating GPUs bought with borrowed money for data centers that haven't even broken ground yet. HELP ME!"

Translation:

"I have too much debt and I'm stuck holding all these depreciating GPUs bought with borrowed money for data centers that haven't even broken ground yet."

My take is they're swamped in debt. Debt is 3.81 times equity. Debt service is now 6 times operating income. Things have to break just right for them.

I'm short them from that little lower-high on 10/31.

Very mixed quarter from Coreweave.

Revenue more than doubles, while operating expenses nearly triple, operating income falls by 2/3rds and interest expenses triple.

But, hey revenues are higher.

I guess we'll see how long they can keep this thing in the air at this pace.

Wen profit?

Yeah, I've tried. Futile to try to convince people. They want mushy canned green beans, canned artificial fried onions and gloppy canned soup. S&P only, not even garlic.

They never heard of a prequel?

Jesus, that's amazing. I've been lifting a minute or ten, and I'd never aspire to *deadlifting* 500.

That would be the best-case scenario. Many things have to go far better than expected for that to happen.

OpenAI has made $1.4 trillion in spending commitments by 2030. I strongly suspect the $12b annual revenue they tout is exaggerated; it's more like $4.2b through Q3, losing $12b last quarter.

This *currently* has little in common with '08-09.

Not much leverage, nor the leveraged securities used then. Half the loans are from private credit, which is outside the banking system.

The rest is hyperscaler cashflow, and now, IG debt and other structured instruments. The latter are the concern

No, not at all. They've just overextended, trying to compete with hyperscalers with far better finances, while taking on too much debt, while cashflow is going negative. A lot of their price action is predicated on OpenAI paying them $300b before 2030, and I'm betting that won't happen.

Disingenuous gaslighting.

It's clear centrist Democrats are just another op for the 1%.

NVDA is a designer. It owns no fabs. TSMC actually makes the GPUs under contract.

They publicly claim the proceeds will be used to buy more OpenAI.

I've come to the overall view that data center centers are the weakest link in the AI value chain.

An AI Addendum Last month, I chose to strip away all the hubris around AI and ask one simple question, one that oddly no one had really bothered to ask; how much revenue is needed to justify the current level of cap...

This piece got me thinking about the depreciation issue, FWIW. Burry is very right to talk about all this.

An AI Addendum - Praetorian Capital share.google/1SObrzG6i5O7...