Hallucinations

‍

What Happened to Cheap Compute?

‍

Soundbite: An MIT study on AI deployment reported that only 5% of all projects “achieve rapid revenue appreciation” with recent initiatives not generating profits. While this alarmed investors last week, we believe poor revenue generation is more a function of a steep learning curve than a general statement on the technology itself. As users learn more about how to properly apply AI, revenues may increase, but the increased costs of more sophisticated applications could slow down acceptance. With investor expectations sky-high, if AI rollout falls short of forecasts, it will negatively impact the AI-theme and the broad indices it has driven to record highs.

AI productivity may come at a higher price than stock bulls expect. General purpose models were extremely expensive a few years ago, then users enjoyed steep price drops as the technology developed. The price of one million tokens (the unit of data that AI models process) is only a quarter of what it was last year, but has since stopped falling. As users have become more familiar with AI agents (LLM models that reason and dynamically adjust to new information), the number of tokens used per project has increased with more complex tasks.

This higher consumption of tokens translates into higher project costs for both training and inference (the latter being the decision output of the trained AI model), resulting in higher costs than had been expected for AI projects. 

Companies find themselves at a crossroads. The AI-wave promised torrential deflation as increasing scale was meant to lead to lower expenses. Unfortunately, the more complicated tasks carry higher costs than expected, causing a pullback from the most optimistic forecasts. The hope for rapid AI deployment is at risk for smaller firms with lower budgets and may even become problematic for S&P 500 companies. With P/E ratios and profit margins precariously at peak levels, the productivity miracles that were being built into earnings forecasts may need to be reworked. Overly optimistic investor expectations could find themselves vulnerable to disappointment, and that could develop into an eventual market top.

‍

The Fed: Couples Therapy

‍

Soundbite: The July 29-30 FOMC Minutes isolated the two dissenting voters who wanted to cut rates over their concern about potentially weak employment demand. In contrast, the majority of FOMC voters at the July 29-30 meeting were focused on rising inflation, describing the labor market as at or near maximum levels. Then, Chair Powell’s Jackson Hole speech on Friday aligned with the two dissenters after the most recent nonfarm payroll report (released days after the last FOMC meeting). Echoing Governors Waller and Bowman, we believe he described a shift among Committee members toward cutting rates to ensure against rising employment risks in hopes of avoiding a sharp drop in consumer demand. The market’s joy in a more certain rate cutting campaign should be tempered by the fact that cuts may not be precautionary, but due to an actual economic slowdown.

The isolation between Bowman and Waller and the rest of the Committee can be summarized by the following passage on balancing the risks of higher inflation vs. weak employment:

“In their evaluation of the risks associated with the economic outlook…a majority of participants judged the upside risk to inflation as the greater of these two risks…A couple of participants considered downside risk to employment the more salient risk.”

We had written about the Minutes last week: ”We know that Governors Waller and Bowman dissented…We will look for how often the word 'couple' appears.”

On careful reading, the phrase “couple” was linked to remarks about slowing growth and consumer activity, inflation running close to the Committee’s two percent goal, and that “higher tariffs were unlikely to have persistent effects on inflation.” In contrast, the remaining members discussed stable employment, with inflation risks “exceeding two percent for an extended period,” increasing the risk of inflation expectations “becoming unanchored in the event of drawn-out effects of higher tariffs on inflation.”

Fast forward to Friday, and Chair Powell opened his Jackson Hole speech (that could have been written by Waller), saying “the balance of risks appears to be shifting.” Powell described the July employment report as follows:

“This slowdown is much larger than assessed just a month ago…May and June were revised down substantially...[with] a marked slowing in both the supply of and demand for workers. This unusual situation suggests that downside risks to employment are rising. And if those risks materialize, they can do so quickly in the form of sharply higher layoffs and rising unemployment.”

While most analysts are attributing Powell’s comments as his own view, we believe Powell may still be in line with the Minutes, and inflation is his main concern. Powell notably distanced himself from Waller (who thinks the Chair is mistaken when he views policy as not fully restrictive) when the Chair said Friday: “With inflation above target, our policy rate is restrictive—modestly so, in my view.” Therefore, Powell’s statement that “A reasonable base case is that the effects will be relatively short-lived—a one-time shift in the price level” may not be his own, but that of the consensus. His primary concern may still be about the size, duration, and timing of tariffs.

Why is that important? While unlikely, it opens the door for Powell to lead a spirited discussion about keeping rates unchanged if more than one-third of the items in this Friday’s PCE report are inflating by more than 5-10%, or that the September 11 CPI report shows an earlier-than-expected increase.

Friday’s rally was driven by investor belief that rate cuts will provide support to avoid economic weakness, with soft labor demand preventing workers from demanding higher wages, so we are back in the sweet spot of deflationary growth. This buys time for lagged fiscal stimulus to bolster the economy in the long run.

On the other hand, the fact that the market is not pricing in even a hint of recession turns our focus to the Employment-to-Population ratio updated on the September 5 Nonfarm Payrolls report. Further deterioration in this series, which rolls over before recessions, would give the market its expected rate cut. However, if easier monetary policy is driven by a slowdown in labor demand, that is not equity friendly.

‍

To Forecast Stock Direction, Look to the Credit Markets

‍

Soundbite: High Yield credit spreads are near record lows, showing confidence that profit margins and cash flow at companies remain strong, keeping default rate forecasts at negligible levels. Until credit spreads widen out, bulls will stay complacent. We will be watching that spread for any negative reaction at the upcoming economic releases prior to the September 16-17 FOMC meeting: PCE price index, Nonfarm Payrolls, PPI, and CPI.

The VIX volatility index closed Friday at its lowest level of 2025. Investor optimism is high in equities as we approach the most seasonally challenging period of the year. Citadel’s internal data shows retail positioning peaking at Labor Day, and equity seasonals traditionally show September returns as being the worst of the year. Most recent history is particularly rocky: For the S&P 500, September has been negative 4 of the last 5 years, with the average loss being 5.72%.

The backdrop feels as though we are inching up on the roller coaster right before the drop as trend following CTAs and volatility target funds are increasing their positions, and shorts were blown out on Friday’s rally. This is occurring as corporate buybacks are becoming less of a driving factor as earnings season winds down.

Therefore, the risk/return outlook is deteriorating. Still, it is best to wait for a signal in this key spread between the lower rungs of the credit ladder and treasury bonds to begin positioning for a trend reversal. The ICE BofA Single-B High Yield spread is trading at levels not seen since H1 2007, when the S&P 500 was back toward the 2000s all-time highs. It gave a warning by sharply widening in Q3 before the equity index peaked in October of that year. If that fixed income gauge widens on any of the inflation or employment reports coming out over the next three weeks, then there is cause for concern. Until then, stock bulls do not see a reason for worry.

‍

What to Look for This Week

(All times E.S.T.)

‍

1. Friday, August 29 at 8:30 a.m. Personal Consumption Expenditures Price Index for July. The series has risen over the last two months, and July is forecast to be unchanged at 2.6%. We will wait for the Dallas Fed to release their Trimmed-Mean PCE July data at 9:00 a.m. and look for the percentage of items in the Price Index that is inflation greater than 5%. The July report revealed that 30% of the items were in that excessive category, and there has been only one occurrence over the past two years that has been higher. Therefore, a rise in July could bet a mention at the September 16-17 FOMC meeting.

2. Tuesday, August 26 at 10:00 a.m. The Conference Board U.S. Consumer Confidence for August. Our focus is on the Labor Differential (Jobs Easy to Get minus Jobs Hard to Get), considering the weaker July payroll out earlier this month. The percent of respondents who thought jobs were “plentiful” rose, but those who thought jobs were “had to get” rose a large 1.7% from June, which caused a contraction in the Labor Differential, which can lead the unemployment rate. We will watch both components, with an emphasis on the “hard to get” number.

3. Friday, August 29 at 9:45 a.m., Chicago PMI for August. The July result was at the high end of the range that has been established since December 2023. It is forecast to fall by 1 to 47, but a 48 print would get investors excited about a potential manufacturing recovery. However, a true expansion requires readings above 50, something we haven’t enjoyed in this survey since Q3 2022. We will be looking at the Prices Paid subcomponent in this regional survey for any reflection of tariff pressure coming from the auto sector.

FOMC Voters Speaking: Monday, September 25 at 7:15 p.m. NY Fed President John Williams Thursday, September 28 at 6:00 p.m. Governor Christopher Waller speaks on monetary policy.

Earnings: Light calendar, with two notes: 1) corporate buyback demand ending, taking one pillar of support away, and 2) big fish: NVIDIA earning on Wednesday August 27, after market close. As an afterthought, DELL after market close on Thursday, August 28.