Chinks in the AI Armor

We are starting to see changes in the hyperscalers.  META is following in the footsteps of xAI, cutting its losses on its AI ambitions.  META announced they are going to be renting out their excess AI cloud capacity, and it resulted in a big jump in the stock.  Maybe AI compute capacity isn't as tight as people assume.  Judging by the reaction to the META announcement, the stock market is rewarding hyperscalers cutting costs, and punishing those that mindlessly dump more money into AI capex.  Eventually, the hyperscaler CEOs will get the message and start reducing their futures capex plans.  It could come as soon as the next tech earnings reports in late July.  If MSFT, AMZN, GOOGL, and META signal lower than expected AI capex, that could be an atomic bomb dropped on MU, SNDK, NVDA, DRAM, SOXX, AMD, INTC, DELL, and a slew of other AI hardware/semi names.  

The AI momentum trade has hit a speed bump.  It is coming at the same time that the SPX has been in a short term rally.  This is the first time this year that the AI infra/hardware trade is underperforming the market as it is going up.  The biggest trade of the year is showing a change of character.  This comes after MU had a huge gap up after a big earning beat, and then sold off hard afterwards.  It has been only a few days, so I wouldn't say its game over for the AI trade, but it makes me more interested in shorting the index this month on a further rally towards SPX 7600.  

It is interesting that CNBC seems to be more bullish on the overall market now that the non-tech stocks are rallying while the semiconductors are pulling back.  This is the exact situation where they should be more bearish.  Ever since 2020, rallies where tech stocks, especially momentum names, leading the index are much more durable than rallies that include a lot of non-tech stocks and small caps outperforming big cap tech.  

Investors have poured in record amounts into US equities so far in 2026.  

Retail has been aggressively buying dips in 2026.  They've also been buying quite a lot on up days as well.  Both are above the lofty 2021 levels.  Given the underperformance of the heavily retailed owned Mag7 this year, it looks like they are doubling down as their holdings go lower.  Not a pretty picture.  Bad things happen when retail investors refuse to cut their losses and instead add to their losers.  
 

Only halfway through 2026 and IPO proceeds already above the SPAC mania 2021 levels.  More big IPOs are lined up for the 2nd half of the year.  Supply is coming out big time to meet the demand for US equities.  

SK Hynix has pulled forward its US IPO plans to July 10.  It was originally planned for August, but it looks like they want to sell shares while the iron is hot.  

The evidence is undeniable.  Equity supply is increasing at the fastest rate since the dotcom bubble.  Add to that the reduction in stock buybacks by big cap tech.  It creates a supply-demand dynamic that requires big investor inflows to keep the bull market going.  If we just get a normalization of inflows to just the average over the past 10 years, the demand won't be able to keep up with all the new supply.  Equities are priced at the margin.  With passive investing dominating equity flows, the stock market is much more inelastic to price than before.  A small drop in demand, a small increase in supply could result in a big drop in price.  You could see some big time volatility if we see equity outflows when these big IPOs have their lockup expirations later in the year and into 2027.  

Last week, we finally heard the crowd recognize that Mag7 has been performing poorly this year.  The crowd just about nailed the bottom in the Mag7, as it has been outperforming the AI momentum names like MU, SNDK, etc. since then.  Either the Mag7 or the AI infrastructure momentum names have to lead to keep this aging bull going.  If they are lagging the indexes, the days will be numbered for this bull market.  There is no way that non-tech stocks will be able to lead this market for long.  Investors don't buy US stocks to get long low growth health care, financials, utilities, staples, or old fashioned industrials.  That doesn't drive a long term up cycle.  They are hedge fund pod shop rotation plays that have a limited shelf life.  

Bonds are trading much firmer in recent weeks as oil prices come down and the AI capex trade cools off.  CTAs are near max bearish on Treasuries and near max long on equities.  If you get equities selling off, I expect a long awaited revival of Treasuries as a negative correlation, flight to quality asset.  It won't be like the QE years, but I do expect somewhat lower bond yields on equity market selloffs.  

The next 2 weeks are a seasonally strong time period that aligns with start of the 2nd half inflows into equities.   Its almost time to put on long term shorts that could be held for months.  Seeing more signs of a meaningful top in the AI momentum trade, NDX trading weaker than SPX, and value outperforming growth.  All signs of a coming trend change for this bull market.  Holding a lot of dry powder, waiting to deploy shorts on rallies over the next 2 weeks.  Still think that non-AI infra tech is the place to be on the short side, even though the AI momentum trade looks toppy.  It may be less reward to short names that aren't up so much on the year, but its also much less risk.  There are plenty of non-AI infra tech names that are overvalued and underperforming the Nasdaq.  Lots of potential downside out there even in non AI hardware/semiconductor tech names.