So in one corner of Wall Street credit is imploding. In the other corners…not so much. Baa/Treasury spreads still near all-time lows. https://t.co/SZthgCEVyc
Jim Chanos
1,711 items collected
1,711 matching items · newest first
All of these analyses of data center capital stacks(not just $IREN’s)assume that “growth” and “deals” are universally good. But that’s not necessarily the case. In fact, most of the AI DC leasing deals with good granular data disclosed appear to be mildly to severely uneconomic.
@ericjackson (2) And just this morning Goldman cut their $IREN FY 2026-28 EPS estimates further, post-announcement. https://t.co/TBH5TZiOxa
The marks, er, investors who own $IREN are convinced that an new equity ATM filing equivalent to almost half of their current market cap…means another deal is imminent. You know, like a material event that would need to be disclosed concurrent with this equity filing. https://t.co/O8WFwFkzAP (2) But never fear, in true $IREN style, they (sort of) raised ARR guidance for the December 2026 quarter, but neglected to use this opportunity to reaffirm the $500M digital revenue ARR guidance for the quarter we are currently in. Why not?
These are some pretty bullish statements about near-term developments at $NBIS. Given the substantial cuts in guidance/estimates(see 2026E Revenues below) just three weeks ago, we will see if these upbeat statements show up in 1H results. https://t.co/ZDW3W9Q8GO
People are often too confident…
Their response keeps changing. Just address the affiliated party issue once and for all with a detailed 8-K filing. It’s not that hard. $CVNA
Unassailable logic here.
A reminder that $BX already has private funds set-up to invest in data centers. With institutional investors. https://t.co/Q0RMvYZMGi
@RHouseResearch (2) And your analysis implies 7-8 year useful life for the GPU’s. Again, this is a financial services(leasing) firm, not a tech company. $CRWV
It’s hard to overstate how bad the $CRWV results/guidance were, but consider this: CoreWeave would still be losing money if they depreciated their GPU’s over 10 years! And at a conservative industry-standard 6-yr life, Coreweave would have zero EBIT coverage of their interest. https://t.co/6HiizAOeKs (2) And on top of that insanity, $CRWV seems to be having liquidity issues, as despite 6 months worth of receivables, their “Quick Ratio” has slid to 0.38x at year-end, which is a distressed level. They will need to issue more equity beyond the $NVDA sale, soon, in our opinion.
Lol, there IS another term for this.
OMG, isn’t Professor Siegel famously long the market…?! And he publishes a bullish piece on the impact of AI on the economy? This is outrageous! (See how absurd that sounds?)
Imagine that.
They keep floating this same trial balloon…
Cue the $TSLA Fanboy Chorus: “Great news, as his work is done!”
@wallstengine Then again, when you design products “beyond the limits of physics”, who knows? https://t.co/ikPOpIQb3C
I’m looking forward to catching-up with Jim Grant and Evan Lorenz later today for a rollicking “glass half-full” discussion! Details here: https://t.co/rJXgIwmvuc https://t.co/UMZLwLkA8C
Electric power costs are roughly 6-7% of CPU/GPU rental revenues. Energy is abundant, but needs regulatory/infrastructure progress. Which will happen over the next few years. Power is not a long-term bottleneck. That is also why “Data Centers In Space” is absurd.
FinTV talking heads are puzzled that future forecasts could be other than optimal. (2) Because if you are upset with an “AI Economic Armageddon” scenario wherein the S&P 500 flirts with 8,000 before dropping to a level last seen (checks notes)…ten months ago…then you should be checking your priors. https://t.co/Mq56ODSE0J
Maybe, just maybe, these five year AI forecasts are just guesses. Really big guesses. https://t.co/jXpMMXzXjH
So immigrants ARE gaming the election process…?!
#OnWisconsin
They’re literally credit instruments denominated in fiat. What’s digital is the assets, not the liabilities/preferred. $MSTR
The obvious follow-up question to this is: Do affiliated entities ultimately purchase loans originated by $CVNA, and if so, from whom? Are there third parties in-between Carvana loan originations and Bridgecrest/DriveTime purchases? Or other similar structures? https://t.co/wIm0b0cwQN
Continued legal issues w/Ticketmaster was no longer discounted in the price of $LYV shares.
Continued legal issues w/Ticketmaster was no longer discounted in the price of $LYV shares.
Looks like $CVNA 4Q Gain on Loan Sales was $315M, or 97% of Pre-Tax Profits ($326M), before extraordinary items.
This is more AI Snake Oil from the Silicon Valley promoter class. Currently annual electricity costs for AI data centers are about $600K per MW, which is only about 5-7% of revenues on recently announced deals. Launch, radiation, insurance and redundancy costs would be greater. (2) And while grid power permitting is a current problem, it is rapidly being addressed in business-friendly jurisdictions. Actual power is in abundance in the US, particularly when energy companies partner w/private power producers, as we have seen in many recent deals. (3) So before you fall for the “Data Centers in Space” hype, ask your favorite Silicon Valley pro…
Lol, aren’t we forgetting something…? #CryptoMath
Collection coverage
facebook: unchecked
linkedin: unchecked
podcast: unchecked
reddit: unchecked
substack: unchecked
web: unchecked
x: linked
youtube: unchecked