Commodities continue to be a successful diversifier, providing competitive returns and a negative correlation. They are in a sweet spot on the efficient frontier (which below is from the April 2025 “launching point”).

With bonds providing a port in the storm for now, it’s a reminder that bonds have become a viable asset class again, regardless of their sometimes-lacking anti-correlation credentials. They may not continue to be that port in the storm for long, but at least we are getting paid to take the risk (via high real yields).

History shows that the higher nominal yields go, the more correlated bond and equity returns become. Makes sense since the risk asset must compete against the risk-free asset. In the chart below I show the 10-year yield on the horizontal axis and the 2-year (104-week) correlation of equity and bond returns on the vertical. The size of the dots shows the 13-week correlation to highlight periods when secular regimes of positive correlations are interrupted by short bouts of negative correlations. At 5% we are currently in the “low positive” correlation regime, not unlike the late 1990’s. That suggests to me that the current growth scare (which has produced a negative 13-week correlation) will ultimately give way to the prevailing trend of positive correlation.

Note the divergence in the chart below between the models that include rates and the one that doesn’t. You can also clearly spot previous bubbles, including 1987, 2000, and 2021, when none of the P/E multiple at the time could be explained by these three variables. We are not there today, provided margins and credit spreads hold up (which they are). The point of the chart is to highlight how damaging rising rates can be if they are not offset by growing earnings.

Earnings season is off to a good start with 122 companies reporting so far and 80% beating estimates by an average 560 bps. Trailing earnings are up 33% from a year ago and the n12m estimate is up 39%. The difference between the n12m estimate and trailing earnings is 19% and decelerating. That suggests that a peak in the second derivative is forming. Investors rarely pay up for peak earnings growth, and this in part explains why the P/E-multiple is down 16% from a year ago. The other factor is of course the rising cost of capital.

With growth scares replacing rate scares for now, expectations for a Fed hike this week have all but disappeared. But based on the SOFR curve the Fed is still expected to hike two more times in the coming months.

Yields have put in what appears to be a trading high at 5.36% and are back into the yellow zone (around 5%). Real yields have dropped as well, with the 10-year down to 2.64%. It’s interesting that the current AI-related growth scare happened when the real yield poked its head above the economy’s trendline (potential real GDP growth). Perhaps the economy’s speed limit acts as a natural brake for real yields with respect to growth.

Credit worries are a thing again, from French spreads to ORCL and CRWV CDS. The bond market is regulating as it always does (or at least when not financially repressed), and the glass-half-full take on widening AI spreads is that this is a good thing, in that a more disciplined allocation of capital today will prevent a bursting bubble tomorrow. I tend to agree with this. The counterpoint, however, is that the bubble might not be in valuations but in (circular) earnings, and that a slowdown in the datacenter buildout could cause the seemingly endless backlog of semiconductor demand to slow.

It’s interesting that the S&P ex-AI basket (from GS) is now far less correlated to the S&P 500 index than the equal-weighted S&P 500 (SPW), while both are similarly correlated to bonds.

Correlations are on the move, and the pressure on AI stocks is pulling down the stocks/bonds correlation, even though the stocks/oil correlation remains deeply negative. Growth concerns are overshadowing inflation/rates concerns for now.

The stock market is broadening out again and the “jaws” are closing between the cap-weighted and equal-weighted S&P 500. This has not been a bullish broadening so far, given the zero-sum dynamic of the AI space selling off more than the ex-AI stocks are rallying.

For the DCF model, the battle between the numerator (cashflows) and denominator (cost of capital) has morphed into a battle between rate scares (higher rates force valuations lower) and growth scares (fears about the AI train slowing down and causing the earnings “bubble” to burst). The former produces a positive correlation between equity and bond returns, and the latter produces a negative correlation. Price discovery is always happening, and usually the climb up the stairs takes longer than the climb down. While earnings and margins continue to boom, the cap-weighted index has struggled to get out of its own way since early June as the AI space continues to correct. The market is now broadening nicely, with 59% of stocks above both their 50-day and 200-day moving average, but you wouldn’t know it from looking at a price chart of the SPX. For now, worries about the AI wave slowing down and becoming systemic to the economy have brought the 10-year yield back down to 5.0% and has reduced the prospect of a Fed rate hike this week. Rate worries have been replaced by credit worries (at least in high yield), so the cost of capital hasn’t really changed that much. A big question for asset allocators continues to be the value proposition of Treasuries. They are attractively priced yes, but will they provide negative correlation when we need it the most?

No matter how farfetched the idea might be of the Treasury Department repricing its gold to create cashflows out of thin air, even a remote possibility that desperate governments will do desperate things is a reminder to always have real assets in the portfolio. On that note, I continue to be impressed by Bitcoin’s resilience. I think it will be eating gold’s lunch for a while, although I remain bullish on both.

And Bitcoin? A new 4-year cycle bull market could be underway, based on its power law patterns. There doesn’t seem to be a narrative, but then again, bull markets rarely start with one.

Gold and Bitcoin both remain in holding patterns. Below is gold, which has seen steady inflows but continues to be held down by a deceleration in global liquidity growth (per the tightening in financial conditions mentioned earlier).

Both commodities and managed futures have continued to shine as diversifiers, namely 1-5 year TIPS. As an asset class, they are much less correlated to bonds than the regular TIPS index and have little correlation to equities.

For now, I am giving this cyclical (and secular) bull the benefit of the doubt, assuming (hoping?) that the stellar earnings momentum is not about to mean-revert. I suspect that the “melt-up” phase of this bull (since April 2025) may have ended in early June, but perhaps that is just a pause in the momentum during the seasonally weakest timeframe (July-October). I’ll stick to my 60/20/20 playbook, which has held up quite well, even though it diversifies away from the mega cap theme.

For the bull markets that fill the space, the average duration has been 39 months, and the average price gain has been 101% nominal and 80% real. The average multiple expansion has been 59%. As of September 2026, this bull market is 48 months old and has produced a price gain (using monthly data) of 119% and a real gain of 80%. So, a little longer and steeper than the average, but not an earth-shattering difference. Below is an attempt to visualize a few attributes of bear markets: price, valuation, and credit spreads. The price and P/E drawdowns are below the axis, and the spread widening is above. The big bear cycles produced outsized drawdowns and big credit dislocations.

Perhaps unrelated (or not), we have now reached the climax of the 1998-2000 internet bubble analog. Price analogs tend to fail sooner or later, and I won’t be surprised if this one fails sooner (given the difference in valuations), but it has been a good roadmap.