Interesting to think that roughly $4 trillion in capex, debt and leases rests on the performance of 300ish people ($13 billlion per person). Realistically all of those people need to outperform all of the openweight/open source models by at least a 2 sig move for ROI to appear on time (or at all).... Frontier models need not only to be great, but also continuously benchmark higher than free/open (which is measurable) and by a quantifiable margin.
- Token costs are down 37% from May peak $NVDA
- What is the threshold for a CFO to justify 90% as good for free over paid? $SNDK
- LLM labs pivoting to 'enterprise sales' $META
- How long does Frontier API premium last over open source self hosting? $CRWV
- And by task tier (and task value)? $IREN
#RedOctober
New article will cover:
- 'First time in history people have borrowed more against their stocks than they did in the peak of dot-com... and it just started to unwind' $CRWV
- 'One private company loses ~$14b/year, owes ~$1.15 trillion to mega caps...
- ... bought ~$7.9b 3x levered chip ETF (most of any security on Earth) and lost $36b in eight weeks. Smart money was selling the entire way down
- Current tech 'moat' is a people premium problem, where the people can and do leave (and why that really, really matters) $META ...
How 300 people driving trillions in CapEx (quantitative and qualitative value personell in Frontier Labs), why they're leaving/starting their own, and wouldn't it just be cheaper to in house solution and pay lawsuit later? Go open weight anyway right? $META
How 300 people driving trillions in CapEx (quantitative and qualitative value personell in Frontier Labs), why they're leaving/starting their own, and wouldn't it just be cheaper to in house solution and pay lawsuit later? Go open weight anyway right? $META
The sign of the Kingdom of God present among us necessarily prioritizes that the poor and vulnerable be defended from the machinations of the powerful. This is not complicated; though often deliberately obfuscated.
Ps 82,3-4
Do justice for the weak and the orphan; give justice to the poor and afflicted.
Rescue the weak and the needy;
set them free from the hand of the wicked.
Low of the day so far is $712.61. $QQQ 50 day sma at 713. Currently at 716.82. Implied move into end of day is $1.67 or 23bps. VWAP essentially here at 716.67. Interesting setup. Bought the 720/713 strangle for .11 looking for .22+.
What to watch; in a few weeks the next ABS will price and the pre-sale to final price arguments settled. Sub 4% excess spread will get the attention it deserves. Public tests always win.
Extremely reputable and respectable publication Hunterbrook (being serious) came out with a long thesis on $CVNA which I agree with in part; Walter's shares are pledged to $C and therefore $C most likely will not let him sell them..
Because if first lost tranches (mystery buyer(s)) go no bid, then the senior tranches won't either (for the same GOS economics). Therefore, these subpoenas could reveal that and alot more..
Sometimes you just get a perfect storm and everything will coalesce into pricing parity $CVNA. Mark Jenkins (CFO) has sold alot of his stock (no surprises, it's his to sell and his comp). Really interesting last few weeks for Jenkins:
August 12: Jenkins doted out at JPM conference (lead underwriter)
August 13: Bloomberg story on Walters emergency loans (with double digit yields, ouch)
August 14: Carvana pledged everything for floating rate term loan
August 16: four intermediaries story lands
Coming weeks: not sure how Jenkins stays on in the same capacity. SEC investigation is still live. May or may not be correlated. Take a sabbatical.
The subpoena will look into Jenkins architecture of the gain on sale structure he built no doubt. He was just on stage with $JPM speaking to the other day. Hopefully he didn't allude to cleanliness of mystery buyers (Jenkins was a Wharton professor prior).