Jim Chanos

A good reminder that the tech leaders didn’t see the oncoming cliff twenty-five years ago, FWIW.

The usual. Cutting 2026-27 revenue estimates materially, but margins will increase. Lol, ok. $TSLA

It was very clear to @packers fans that McManus was still not healthy over the past month he was back on the field (missing important kicks). The team said he was fine. And now we find out he is doubtful for Sunday with the same original injury. That’s a bad look, Guys.

@GavinSBaker (2) But more importantly, at what point do results matter here, Gavin? $TSLA began its FSD effort in 2014, claiming big advantages in data collection/usage. Almost twelve years later they are still at Level 2 (supervised), the same as many other auto OEM’s. Efficiency…?!

What’s so ridiculous about this view that $TSLA is an AI/robotics “leader”, is that you can quantify just how little Tesla is actually spending on AI. For the LTM it’s below $3B, or less than 1% of the Big Four hyperscalers. https://t.co/THIqHDvO7T

Indeed

$CLSK doing a convert to buyback shares from certain investors…?! Lol, ok. (It’s really hard to take these Bitcoin miners seriously when they keep doing stuff like this.)

If you were crowing about your stocks being up 200-400% Y/Y a few weeks ago, but are now wondering why they are suddenly down 30-50% (with the market 2% off its ATH)…you might want to read-up on beta.

In case you missed this viewpoint on GPU depreciation. $MSFT $ORCL https://t.co/XISvgkJgd0

@mederComments @gnoble79 The issue was not the fiber optic buildout itself, which totaled only $50B cumulatively, over the five years. It was the double and triple-ordering of ALL networking equipment by enterprises, that got cancelled abruptly in 2001.

Chevron Picks Texas for First AI Data Center Power Project | Greenfield power in West Texas…in 2027. https://t.co/ZnE3cbtB9x

Starting in 2026…?! I have been reliably informed that this is not possible due to “scarcity” and “bottlenecks”. I guess we will see soon enough. https://t.co/Jsx963GubB

$AMD CEO asked about OpenAI’s losses and path to profitability on @SquawkCNBC: “Let’s talk about the vision, then the math.” (Then talks only about the “vision”, not the “math”.) (She does claim 5-6 years for depreciable life math on GPU’s)

@gnoble79 Remember that much of that capacity got built because of the WorldCom lie, “Internet traffic is doubling every quarter.” It was doubling every year(2x per yr),not every quarter (16x). Still substantial secular growth, but not enough to overcome the over-investment from 1998-2002. @gnoble79 (2) See Anthony Odlyzko’s great work/papers on this subject and the railroad buildout in the 19th Century. Must read for the AI buildout debate (and on my course reading list!).

As the AI DC bulls now try to convince you to extend depreciable lives on GPU’s today, consider this: $CRWV’s 3Q annualized Adj EBITDA was $3.4B, and annualized interest was $1.2B. Using 10-year life(!) on their $20.0B of est. GPU’s ($2.0B) means they are still barely profitable.

And yet, despite $CRWV having clients willing to “re-up” fully-depreciated GPU contracts within 5% of original contracted prices…margins are declining. Lol, ok.

The $CRWV CEO: “There’s plenty of power right now, and we believe there will be ample power.” Not the accepted current neocloud narrative.

Uzbekistan Partners with NVIDIA to Boost AI and Digital Innovation | So $NVDA is doing GPU deals with both Kazakhstan and Uzbekistan…?! https://t.co/HbecPA6RwW

Yet another $IREN bull raises his price target as he cuts FY 2026-27 numbers materially, lol. And his model is just as atrocious as Macquarie’s. https://t.co/jsOy49W4ZF

@R_and_Invest (2) Furthermore, the guys running a number of these new AI data center companies told you five years ago how much money they would make from Bitcoin mining. Bitcoin has gone up 5x since then, yet none of these companies made money. So be mindful of their “AI” guidance going fwd.

As we have gotten some inquiries, I can confirm that we have unwound our $MSTR/Bitcoin hedged trade as of yesterday’s open. https://t.co/lgrWNy35H8

@Umbisam (2) And here’s the model with no residual value for the chips. Hope this helps! $IREN https://t.co/oFV09dHuK7

If these guys are getting $NVDA chips… $FRHC #IYKYK

I’m just getting to some of the sell-side cheerleaders’ reports on $IREN’s quarter, and boy…are they a hoot. Here’s Cantor quietly cutting their FY 2026 EBITDA estimate by a cool 35%. And look at the revenue estimate cut(19%!) for the quarter we are in now. https://t.co/hmtSX4imlD (2) And here’s Macquarie telling you that despite needing over $10B externally over the next few years(they will need more), neither interest expense nor the $IREN share count will increase! That’s pretty nifty. Trust these recommendations at your own peril. https://t.co/MxNjLYMzev

So we’re all fine with this guy making these absurd claims to his shareholders? (I guess that question answers itself, given the vote) $TSLA https://t.co/cFd9aBNuKi

@Mining4Answers (2) Here’s the income statement for $DLR. Their consolidated IT load was 2,087 MW at 9/30. https://t.co/DxVMQvckNO

Good to see that some things never change, like $RUN burning another $(865)M in the latest quarter. #ScienceProject https://t.co/cAhRGLwaDt

Did $IREN management just say the $MSFT deal would have a “low double-digit” unlevered IRR…?! Sounds familiar. https://t.co/SS4vw5LRcF

So one month later we give guidance for the end of calendar 2026, but pull the guidance for the end of 2025…?! What happened to that $225M ARR figure from early-October? $IREN https://t.co/Bx3t17FWH8

@CedarStResearch So there is such a shortage of compute going forward that the biggest consumer of it is thinking about selling some of its capacity?