AI returns and risk in a nutshell

Stock returns moderated this year across many key AI industry sectors. Hyperscaler and software returns have moderated sharply. Semiconductors, hardware, and power still show strong performance but have also moderated to some degree. Earnings remain strong, except in the power sector. Sustained earnings have required more capex which has cut into free cash flow for the hyperscalers. If they cannot scale as fast as expected, it will spill over to the other AI sectors. The best returns may have passed in this industry, yet that does not mean we are in for a reversal; nevertheless, moderation is not what many investors expect. 

 

The why for higher yields.

Don’t act surprised that yields are moving higher. Inflation is not being tamed. Real rates are rising alongside stronger economic growth. The term premium is rising because of risk and bonds' lower diversification value. And we haven’t even mentioned the huge increase in Treasury supply that the Fed isn’t buying. 

What would you expect to happen? Nevertheless, this shall pass. Rates are not going back to the days of financial repression and the Fed flooding the market with liquidity. Bonds will reflect the underlying drivers of yield, and if those are controlled, yields will moderate.  

 

Economics as art - ML does not solve the problem

Economics is not an exact science, it's a combination of an art and elements of science. And that’s almost the first and last lesson to be learned about economics: that, in my judgment, we are not converging toward exactitude, but we're improving our databases and our ways of reasoning about them.   

- Paul Samuelson 

Every time we try to make a forecast, we run into the problem that economics isn’t an exact science. There is an art to looking for and weighting different factors. These factor weights also change as economic agents' behavior changes. 

We are improving our data or at least have more it, yet data needs analysis and data science without theory can only sort or rephrase the data but not always providing useful insights. Machine learing has pushed the boundaries of data analysis but that does not mean we are better interpreters at understanding the causal links within the economy. 

Is it time for rotation into bonds?

One of the key drivers of portfolio return is the allocation between stocks and bonds. If you can get this right, you will achieve stronger overall portfolio performance, but it is not easy. Timing will not be perfect, so you have to play the odds when evidence suggests a tilt toward bond relative value. We may be at a point where equities are overvalued versus bonds.  Can stocks get more expensive? Yes, the overvalued market can last for another year, yet a high current coupon suggests that you are being paid to take risk off the table.