$JPM: "....a year ago it would have taken 6-12 months to onboard and start running models on AMD GPUs. The process was expensive and cumbersome given engineering required. Now, using their latest AI model to do the engineering, they can onboard new models on AMD GPUs in just 3 days."
https://t.co/2fdxXcv8Li
I built a free site with the letters of the 23 value managers I read.
750 letters, each linked to the original source and updated every week. You can also search by company and see which managers have written about it.
https://t.co/lSVGPaUkym
An interesting tidbit on BNSF: at first glance, it looks like BNSF reduced its dividend to BRK parent to just $115 million from the usual $1-1.5 billion per quarter.
However, this quarter, it booked a $1.385 billion “non-cash distribution of investments”, which I’m guessing are T-Bills bought at BNSF and then transferred to headquarters rather than letting them sit in lower-yielding cash until quarter-end.
Read the footnotes, kids.
@BNSFRailway
Next time you think that short sellers just make shit up - have a look at this SIX MONTH OLD tweet - which basically alleges the affair that is now news.
Voici AI Bubble Tracker 🫧
Depuis que je suis fortement investi dans l’IA et les semis, j’ai vibecodé ce site qui permet de prendre du recul sur les narratifs en les confrontant à des datas en temps réel : prix de location des GPUs, CapEx, construction de data centers, etc.
https://t.co/wxo9PMElFN
Morgan Stanley maps out estimated hyperscaler and neocloud capex to supplier revenue for 2027.
Gives us a good snapshot of companies in the AI universe.
I always intended Stratechery to have a sponsorship component; it just took me 12 years to get back to it. That I could do so with the company that made the subscription business possible makes me so happy. There is nothing better than endorsing a product you whole-heartedly believe in.
Copart $CPRT new CEO pay package looks on the surface like textbook alignment, but there are some caveats.
The structure is basically...
- Jay Adair takes a $1 salary, no bonus and no new equity for 10 years.
- He gets instead one 20M-share option in four tranches, with strikes at $26.27, $35.03, $43.78 and $52.54. The option has a 5-year cliff, a clawback if he leaves early, and his gains are locked up until year 10.
So basically...
- Stock flat → he earns $0.
- Stock 2x → he earns ~$263M.
My "concern" is the timing of the option issue, which evidently impacts the existing holders more than it does the recent holders.
Jay Adair came back as CEO on July 31, with the stock near multi-year lows around $28, down from ~$52 in mid-2025. The strikes are anchored to the depressed price in which he (opportunistically?) became CEO.
This ultimately means that...
- Tranche 1 pays on any rebound, so many shareholders might still be underwater while he makes millions (this has no simple solution as setting the bar at ATH would also not be fair)
- The “ambitious” top strike of $52.54 is roughly where the stock already traded in 2025...
- “Doubling the stock” mostly means getting back to where it was without any new value creation
- Over 10 years, a 2x is roughly ~7% a year
Now, a founder-family CEO taking $1 and betting everything on the stock is rare and definitely better than nothing, but the design matters too and reality is more nuanced than it appears.
As designed, Jay Adair could earn hundreds of millions simply by returning Copart to where it was before he came back.
Still subject to shareholder approval!
Nomad Letters: The Full Collection of the Nomad Investment Partnership Letters to Partners, 2001–2014, now available for preorder: https://t.co/oNIxT4wyH8
👀Berkshire is quietly buying one stock almost every single day. Nearly $600 million in just over two weeks.
It's in the most hated corner of the market right now.⤵️
Before defending a position, write the strongest case against it. Then name the evidence that would distinguish a temporary setback from a broken thesis. That is a practical application of Ed Wachenheim’s discussion of contrarian research and imperfect information.