rolf thrane

On issue 148 - It is my view that yes, as the setup is now, the economics will not work - because it is pitted as a low cost everyone can afford it if it - like google searches - which it is not at all. That said, there has to be an economic version of AI - we are just not seeing that yet in the market I as a consumer am willing to pay a lot of money for AI - and I mean a lot of money. Others are not. By definition, there has to be an economic equilibrium of consumers willing to pay for a certain supply/output value. On the issue of scalability, it cannot be there are no scale economies. The issue is that the market is not investing commensurately with viable scalability. I agree with Dougie, there will be a reconning and a complete restructuring of AI.

A. High-value users

People like (professionals, technologists, financial operators, legal/technical writers) who will pay $200–$1,000+/month because the marginal productivity gain is enormous.

B. Low-value or casual users

Who are not willing to pay meaningful sums and don’t generate significant economic output through the model.

Right now everyone is lumped into one subscription tier. That is a recipe for catastrophic economics.

The real equilibrium will be stratified, similar to:

• Enterprise SaaS pricing
• Bloomberg/FactSet tiering
• GPU-indexed consumption pricing
• Metered compute models (like AWS)

In equilibrium, heavy users will subsidize model development—not casual ones.

IMHO

sbickleyinreno

Good analysis as usual. I'd also state the tiering above is already happening. Here is an interesting tidbit though. If you have an O/M365 consumer subscription from MSFT, the cost has been $99.99. You will get an e-mail where they tell you how they have added AI features and whatnot and are raising your fee to $129.99/year. They don't tell you that you can opt out and keep the "classic" version for $99.99/year. Now MSFT has ~85M consumer users. This price increase alone, left uncontested, pays them back $2.5B per year on their AI investments. Most consumers will not blink at this price increase, but let's call it an even $2B a year. MSFT is in a uniquely good position with their limited investments and high ownership/profit sharing/IP access with OpenAI.

% Gainers

% Losers

Closing Volume

- NYSE volume 3% below its one-month average

- NASDAQ volume 8% below its one-month average

- VIX index: up 13.06% to 22.42

Breadth

S&P 500 Sector ETFs

% Movers

Nasdaq 100 Heat Map

"Worries restrictive monetary policy is weighing on the economy; Recent data should not deter Fed easing; Supports Dec rate cut for 'Risk management’ "

* Affordability of housing and cars poses ongoing challenge for consumers, weighing on spending growth.

* Excluding Govt shutdown, GDP likely slowed in H2. 

* The labor market is still weak and near stall speed — Underlying inflation is close to 2% target.

* Dec rate cut will provide additional insurance on the labor market — Doesn't see any factors that would accelerate inflation.

Moving to medium sized from large sized on Kimberly-Clark  (KMB) and PepsiCo  (PEP) .

At 2 p.m.:

sbickleyinreno

One could compare Dougie's MSOS experience to his crypto and AI analysis I suppose, except the MSOS trade has worse results! I will say that at least with AI and Crypto there are real metrics, supply chain data, supply/demand debates, use case debates that all center on actual data and real metrics. The marijuana trade is a fool's errand - it is based solely on politics which are as unpredictable as ever. Funny how we don't see the negative side of the weed trade in the selected tweets and articles posted here like we do on AI Tales and the pending crypto armageddon.

Dougie Kass

In the interest of setting the record correct (after 1-2 flops) I have made seven figures in 8-10 position trades in weed, net over the last number of years. Eeking out that sort of profit (buying when things look dismal and selling on rumors of improving regulatory moves) has been a feat.

PS: The AI trade is the market (cannabis market caps have always been very very small and irrelevant) - despite your sarcasm, that is an explanation why almost 150 tales of Nvidia - though I have had fairly thorough coverage of cannabis.

Haters gonna hate but my skin is pretty thick after all these decades and the above comments do not even register on my give a s#*t meter.

robbo

thanks for this...I could add that when the last maven sells his weed stks, they will have a chance to recover (or even more). Read over the weekend an ANALyst dislikes the sector, but does like GTBIF (which I have frequently called "best of breed). Unfortunately "best of breed" still is not very good, although their cash flow is actually quite good. What the sector needs is 280E tax to get revised, so they can account for their earnings like any other company. Then the stks will be fine. But not until then.

Dougie Kass

I expect a lot of tax loss selling pressure in the weeks ahead, until year end - as I have written.

Besides my index shorts and cover, my only trade today has been a short of Live Nation  (LYV) .

I effected this trade following a report that the U.K. plans to ban the resale of live events tickets above their box office price.

I suspect such an aggressive crackdown on resellers might creep into other geographies.

* I wonder when people will figure out that due to all the capex and borrowing, big tech no longer has the money to buy back stock and the equity dilution is really going to start showing up? That will be combined with the coming depreciation hit.

* Look at the alt-coin index of junk crypto. It has gone from about 200 to start the year to zero now.

Volume

- NYSE volume 12% below its one-month average

- NASDAQ volume 9% below its one-month average

- VIX index: up 5.50% 20.92

Breadth

S&P 500 Sectors

% Movers

Nasdaq 100 Heat Map

From Peter Boockvar:

NY mfr'g improves in November

The volatile November NY manufacturing index picked up the pace to 18.7 from 10.7 and above the estimate of 5.8. It was -8.7 in September. The rise in new orders to 15.9 helped but it was not followed by backlogs which remained negative at -5.8. There was a rise of inventories to 6.7 and increase in shipments to 16.8. Employment was positive at 6.6 vs 6.2 in the month before and the workweek went positive after two months below zero. Delivery Times rose (implies slower supply chains) while prices paid and received slipped after the October increase.

After rising to 30.3 from 14.8 in October, the 6 month business outlook fell to 19.1. Capital spending plans did bounce to +11.5 from -2.9 and I’m guessing the tax expensing incentives in the OBBB helped.

Bottom line, after almost a 3 yr manufacturing recession we search for any clues of a bottom and today’s print from the NY Fed looked good but it hasn’t been correlated with the Philly index, its close district and the national number too. So, we’ll wait to see more information from other regions before jumping to conclusions. No real market response to this figure, ever.

NY Manufacturing Index

6 Month Business Outlook

I like the price action of my two most recent long purchases -  (KMB)  and  (PEP) .

I should probably add on the strength.... for a change!

Following up to my previous piece where I said: 

“Here is Jensen Huang trying to explain why this space is not a bubble, but unwittingly explaining why it is a bubble":

What he lays out is precisely why the economics do not work. This is why it is terribly dis-economic, and everyone is losing money to NVDA’s benefit, including people who are struggling to put food on the table but having to pay their elevated power bills. Gen AI does not scale, is too resource consumptive, and does not work well. The one-sided trade to NVDA’s benefit cannot continue forever. NVDA needs to fund their own customers for a reason. Open AI needs to beg for government money for a reason. Bubble.”

The economics of this industry make zero sense. It takes a massive capital investment and massive opex to train a generative AI model. This is with the benefit of being able to take content without consent, which they have done to date, and it may not be sustainable. This entire process has ZERO revenue associated with it. It is a money loser right off the bat. Then the problem, per Huang's comments which he presented as bullish, but are really bearish, is every query (inference) is also tremendously compute and power intensive with different infrastructure that also requires massive capital investment. Inference alone, even without the training cost, seems to be a massively money-losing proposition. In third-grade language, the revenue from inference is substantially less than its cost. Every click costs money.

I am not fully certain, but estimates I have seen are that 10%-20% of a model’s cost comes from its training, and the remaining 80%-90% of its cost comes from inference over its lifecycle. I think the CAPEX mix is somewhat similar, although my guess is somewhat less biased towards inference. Point being almost all of the cost of Generative AI is in its end use, which makes it hard for it ever to be economic. Every single query is incredibly resource consumptive. There is a massive cost of compute to every query. Each query loses money, even absent the massive upfront investment in training.

It does not pencil out in my view.

Related to this issue, below are some guesses as to what OpenAI's revenue really is. The first link is a summary, the second link is a more detailed article:

https://x.com/kshaughnessy2/status/1989646014156783957

https://www.wheresyoured.at/oai_docs/

I was recently skeptical about Sam Altman’s claims of $20 billion of revenue (which is laughable in light of the fact that they need to spend $1.4 trillion, no wonder they are begging the government). $13 billion was more commonly referenced. The author thinks the number is even lower.

With "Scammy" Altman, who knows. Slowing growth, share losing, incredibly diseconomic, highly questionable management, constant dilutive money raising, no idea what anyone finds compelling beyond betting on the next greater tulip-picking fool.

Upside:

-ZYME +34% (positive HERIZON-GEA-01 Phase 3 Results Support Ziihera (zanidatamab-hrii) as HER2-Targeted Agent-of-Choice and Ziihera Combination Regimens as New Standard of Care in First-Line HER2-Positive Locally Advanced or Metastatic Gastroesophageal Adenocarcinoma)

-SSP +25% (Sinclair confirms 8.2% stake in Scripps; considering possible combination)

-JAZZ +20% (positive HERIZON-GEA-01 Phase 3 Results Support Ziihera (zanidatamab-hrii) as HER2-Targeted Agent-of-Choice and Ziihera Combination Regimens as New Standard of Care in First-Line HER2-Positive Locally Advanced or Metastatic Gastroesophageal Adenocarcinoma)

-DFLI +11% (earnings, guidance)

-SQM +5.7% (lithium strength)

-GOOGL +3.8% (Berkshire discloses new stake in GOOGL)

-GDRX +3.7% (launches new $39 Per Month Weight Loss Telemedicine Subscription)

-LAC +3.1% (lithium strength)

-SE +2.9% (a nnounces up to $1B share repurchase program; hearing Phillip Capital Raised SE to Buy from Neutral, price target: $170)

-GAP +2.4% (Barclays Raised GAP to Overweight from Equal Weight, price target: $30)

Downside:

-ARMK -10% (earnings, guidance)

-DELL -6.0% (Morgan Stanley Cuts DELL to Underweight from Overweight, price target: $110)

-PLUG -5.1% (hearing price target cut at Susquehanna on hydrogen demand concerns)

-XPEV -4.8% (earnings, guidance)

-HPQ -3.6% (Morgan Stanley Cuts HPQ to Underweight from Equal Weight, price target: $24)

-NUVL -3.5% (announces Topline Pivotal Data from ALKOVE-1 Clinical Trial of Neladalkib for TKI Pre-treated Patients with Advanced ALK-positive NSCLC)

-HPE -3.4% (Morgan Stanley Cuts HPE to Equal Weight from Overweight, price target: $25)

-SEE -3.3% (confirms to be acquired by CD&R at $42.15/shr for $10.3B cash deal)

-LB -2.5% (files to sell 2.5M class A shares with 30-day option for 375K shares)

-MSTR -2.2% (acquired 8,178 BTCs for $835.6M during Nov 10-16th)

Fed Speakers:

9 a.m.: Fed Bank of New York President Williams (Voter) gives welcome remarks before the 2025 Governance and Culture Reform Conference organized by the Federal Reserve Bank of New York (No text. No Q&A);

9:30 a.m.: Fed Vice Chair Jefferson (Voter) speaks on the economic outlook and monetary policy at event hosted by the Federal Reserve Bank of Kansas City, Kansas City, MO (Text and livestream available. Q&A from moderator);

1 p.m.: Fed Bank of Minneapolis President Kashkari (Non-Voter) moderates a fireside chat and audience Q&A with Christophe Beck, CEO of Ecolab, at the November luncheon for the Economic Club of Minnesota, MN (No media Q&A. No prepared/embargoed text. Livestream at minneapolisfed.org/live);

3:35: Fed Board Governor Waller (Voter) speaks about the economic outlook at the Society of Professional Economists Annual Dinner, London

Economic Calendar:

Charts from 7:49 a.m. 

From Peter Boockvar:

Just a few things to note but of relevance

The Citi Panic/Euphoria index measuring sentiment on the stock market is now at .79, almost 2x the Euphoria threshold of .41. The chart had to be widened out to include the new print seen on Saturday. With this disclaimer from Citi, we must take note:

“The panic/euphoria model is a gauge of investor sentiment. It identifies “Panic” and “Euphoria” levels which are statistically driven buy and sell signals for the broader market. Historically, a reading below panic supports a better than 95% likelihood that stock prices will be higher one year later, while euphoria levels generate a better than 80% probability of stock prices being lower one year later. Currently the reading at or above 0.41 indicates euphoria and anything at or below -0.17 indicates panic.”

A big mistake investors make in private investments is believing it is less correlated with publicly traded assets just because it’s not marked to market each day. The FT today is highlighting the reality that the price of private assets actually moves around every day in a piece titled “Blue Owl private credit fund merger leaves some investors facing 20% hit.” They are merging a private credit fund of theirs called Blue Owl Capital Corporation II into their publicly traded business development company OBDC which is trading at a 20% discount to NAV. “If the mooted deal were to be approved by shareholders and completed at current prices, Blue Owl Capital Corporation II shareholders would see the value of their investments fall by about 20%.” And, those in the private fund “will be restricted from pulling money from the fund until the merger with OBDC closes in early 2026, at which time they will permanently lose the ability to redeem cash at the fund’s NAV.” Redemptions in the fund by the way are up 20% y/o/y. Link: ft.com/content

Bottom line, the only major differences between a private and a publicly traded investment are liquidity and how/when they are marked.

Excessive debts and deficits matter in Japan as the 10 yr yield rose to a fresh 17 yr high overnight, up 2.6 bps to 1.74% in response to the fiscal spending proposals that are coming from the administration of the new PM. The 30 yr yield was up by 4.5 bps to 3.27% and the 40 yr yield was higher by 5.2 bps to 3.61%. The 10 yr inflation JGB breakeven was up by 5 bps to 1.60%. Also of relevance is that the move higher in the long end is happening irrespective of what the BoJ is doing with short term rates, which currently remains nothing, though I think they hike next month. Sounds familiar as the Federal Reserve has cut rates by 150 bps and the 10 yr yield is barely lower since. And in Europe, the ECB has cut its deposit rate by 200 bps since June 2024 and the German 10 yr bund yield and the French 10 yr oat yield are higher over this time frame. The same can be said for long end gilt yields since the BoE started cutting its base rate. I remain a long duration bond bear.

10 yr JGB Yield

German 10 yr bund yield

* More on rising Japanese yields...

Meshuganah Is Trump (Craziness Is Three-Fold in Yiddish)

* Things we see at a top... 

Bonus — Here are some great links:

Option Expiration Week - A Mixed Outlook

Offense vs Defense

Deja Vu

Do You Have Insurance?

The Bitcoin Shock 

* More things that occur at a top...

Sixty-minute man, sixty-minute man

Look a here girls, I'm telling you now

They call me "Lovin' Dan"

I rock 'em, roll 'em all night long

I'm a sixty-minute man (yeah, yeah)

If you don't believe I'm all that I say

Come up and take my hand

When I let you go you'll cry, "Oh yes"

"He's a sixty-minute man"

There'll be 15 minutes of kissing

Then you'll holler, "Please don't stop" (don't stop)

There'll be 15 minutes of teasing

And 15 minutes of squeezing

And 15 minutes of blowing my top

- Billy Ward and The Dominoes, Sixty-Minute Man

I have covered my index shorts from about 4:30 AM now for a very quick $2/share gain:

*  (SPY)  $672.97

*  (QQQ)  $610.92

This is interesting below from Michael Hartnett re-shorting hyperscaler bonds. The thing I cannot figure out, if he believes this, aren’t the equities by definition a better short? Also if you look historically, at every CAPEX cycle, the equities peak well before CAPEX peaks.

Hartnett: The Best Trade Into 2026 Is Shorting Hyperscaler Bonds

"The bulk of central bank easing is behind us… which is why bitcoin is < $100k, why markets are nervous about credit cracks and financing capex booms, and why the best short is AI hyperscaler corporate bonds." - Hartnett

The S&P Short Range Oscillator is at 0.07% vs. 0.77% — at neutral.

Back shorting the indices on the +35 handle rise in S&P futures (I view this mostly as a Google mark):

*  (SPY)  $675.34

*  (QQQ)  $613.57