AI Carnival Barkers

The Anthropic CEO came out pumping for the upcoming IPO, putting a $30T total addressable market (TAM) for his company.  Getting a scent of Musk in those numbers, when SPCX was pumping the crowd for its IPO in June.  Musk stated that the TAM for SPCX was $28.5T.  These carnival barkers don't let reality get in the way of a good story.  So Dario managed to one up SPCX's TAM by $1.5T.  Never mind that these numbers come out of thin air with unbelievable projections.  Apparently, Anthropic is looking for a $2T IPO valuation, once again one upping SPCX which got priced at $1.6T.  

$2T for a company that already has serious competition from China, a place well known for cutthroat competition, leading to price wars and massive overproduction.  These LLM companies are losing loads of money as they sell compute way below cost.  With the much cheaper Chinese models catching up fast to the frontier models, the price competition will only get more heated.  The CDS of some of the companies financing this AI boom are rising noticeably as Wall St. begins to sniff out that all this debt could result in an AI bust if hyperscalers stop malinvesting.

Yet the stock price for NVDA has recently been closing in on all time highs, despite the rising CDS.  There are two different pools of buyers for AI stocks and AI bonds, and there is a divergence growing.  The CDS for AI companies is rising with the growing amount of debt issuance and guarantees.  The stocks are staying strong with equity inflows providing continuing support.

This AI mania is testing the sanity of bears who see an overhyped technology with much fewer practical use cases than what the carnival barkers proclaim.  It all comes down to how much the financial markets will facilitate the continued spending on AI.  Rising credit spreads for AI companies is a warning sign.  If the markets continue to take down gobs of debt and equity without a further rise in CDS, then this mania can continue for longer.  Anthropic and OpenAI are at the center of the circle.  If they have trouble raising capital in the future, that would be a certain death knell for the AI trade.  Before that happens, you probably see some smoke before the fire.  Slowing growth rates, less demand for AI at higher prices, and losing market share to cheaper but very capable Chinese models.

During the last 5 months, we have had the biggest equity fund inflows ever over a 5 month period.  Market cap adjusted, they are even higher than the 2021 flows.  Consumer sentiment on stocks is extremely bullish.  When retail gets this bullish on stocks, a big correction/bear market is not far away.  

Household equity allocations are at an all-time high.  SPX 10 year forward real return vs. household equity allocations.  

When you are looking for a correction off an extended bull market uptrend, you have to give yourself time for the move to play out.  Local tops are harder to time than local bottoms. Its taken longer than expected, due to the lower trading volumes in August.  The lower volatility provides support for the market as vol control funds add equity exposure and lower IV on puts force dealers to reduce their shorts as they delta hedge.  But there is a downside, as lower volatility make investors complacent about the risks out there as they look at the past few weeks and extrapolate it out to the next several weeks.  

The bear catalysts are still ahead of us, perhaps too far ahead for investors to really care about in August.  But as we get closer to Anthropic's IPO and the midterm elections, you should see volatility pickup.  With investors having a big overweight to US equities after an extended rally, its not going to take that much to get that 5+% correction.  

I don't put too much weight into seasonal tendencies, but we are entering historically the most bearish part of the year after having rallied for almost 5 months without a 5% correction since the late March bottom.  There has been studies showing that there is a seasonal pattern to investor risk aversion and risk seeking behavior.  

Even in the most bullish markets, after a significant bottom, there is usually a 5+% correction within 4 to 6 months.  The last time you had a significant bottom, in April 2025, it took about 7 months to get that 5% correction.  Unlike 2025, you didn't get a huge capitulation drop.  The February-March drop didn't even manage a 10% correction, less than half of February-April 2025.  And after the March correction, investors were extremely eager to pile back into the highest beta stocks, so they got their full allocation in June, within 2 months of the bottom.  

COT data for NDX shows small specs increasing to their largest net long position.  Commercial traders have drastically cut their long positions and are now down to net short levels similar to January.  

NDX Futures Small Spec and Commercial Net Position

One of the foundations of my short thesis is NDX underperforming SPX.  While SPX blasted to a new all time high in August, almost 3% above the June highs, NDX was unable to make a new all time high, and has been an underperformer this summer.  In the dotcom bubble, NDX continued to outperform SPX throughout 1998 and 1999, up until March 2000.  It never had a 3 month stretch where it underperformed the SPX before the final top.  Only after the March 2000 peak did you see SPX outperform NDX.  From April 2000 to August 2000, SPX and NDX traded in a range until the start of the bear market in September 2000.  

The NDX lagged the SPX during the final 2 months of 2021, which was right before the final top in January 2022.   Without NDX leadership, you cannot sustain a bull market for long.  Investors don't buy US stocks because they are excited about health care, financials, or consumer discretionary/staples.  They buy for tech exposure and growth.  Without AI, there is no real growth in the US, its all just inflation.  

As I have repeated many times in previous posts, this market is nearly a carbon copy of what happened during the dotcom bubble.  It is surprising to see the NDX lagging SPX for so long, even though the AI capex story remains strong, as everyone out there keeps repeating.  Based on the price action and what I read on Fintwit and wallstreetbets, it seems like a lot of investors are already all in on the AI trade. 

A quick word on the bond market.  The Fed and Treasury continue to get a lot of attention.  Like 2000, the bond market is basically a non-factor.  Clearly, inflation is much stickier than the investment pros will have you believe.  The guy on the street knows that CPI and PCE are bunk, that they vastly underestimate inflation.  Commodities like wheat and corn are making new 52 week highs during a seasonally weak time period right before harvest.  The fiscal deficits continue to grow with an aging population, defense spending, tariff refunds, and OBBA tax cuts.  The Fed is too dovish considering the environment we're in.  But at current 10 year yields, I would say its neither stimulative or restrictive.  You would need at least a 50+ bps move in yields for bonds to become a factor again, and I doubt that happens.  Any significant market reactions to a dovish or hawkish Fed are overreactions.  

This week, some were thinking that Jackson Hole would be a market mover, and we did get some intraday volatility, but nothing major.  In the past, the Fed was the only game in town.  Not anymore.  Its fiscal and AI.  Fiscal policy is basically used up, with Trump already passing his tax cut bill and with Democrats all but a lock to take the House, ensuring gridlock.  The Dems are definitely more anti-AI than the Repubs, so not likely to get much AI friendly legislation passed in the next 2 years.  

Maintain my full position short SPX and NDX.  This market is testing the patience and conviction of the shorts as it has gone nowhere over the last 3 weeks.  If you are a long term believer in the AI theme, then you shouldn't be short.  If you are not a believer, then this is probably the best time to be short since the bull market started in October 2022.