AI Bubble Peak? | Systematic Investor | Ep.388

Episode Overview

• This episode explores the complex interplay between macroeconomics, asset bubbles, and investor psychology, focusing on how regime shifts change the rules of investing.
• Mark Rzepczynski breaks down the mechanics of "risk" versus "uncertainty," explaining why traditional financial models fail when applied to new technologies like AI or shifting geopolitical landscapes.
• The conversation frames the current economic environment through the lens of history, comparing post-pandemic inflation to post-war economies and analyzing the tension between fiscal and monetary policy.
• Listeners will learn why the definition of a "safe asset" is changing, why machine learning struggles in finance, and how to construct portfolios that survive in a world where stocks and bonds may no longer diversify each other.

Key Concepts

• The Anatomy of Bubbles: Bubbles require two ingredients: excess liquidity (money) and a powerful narrative. They cluster most aggressively in "hard-to-value" assets like crypto or AI tech because there are no concrete cash flow models to disprove the optimistic stories. In contrast, commodities have hard supply constraints, meaning euphoria eventually hits a physical wall that narratives cannot overcome.

• Risk vs. Uncertainty: This is a critical mental model. "Risk" is measurable based on historical distributions (like volatility). "Uncertainty" applies to new phenomena (like the future of AI) where no history exists. When facing uncertainty, investors abandon math and rely on narratives, which makes markets prone to wild behavioral swings.

• The Shifting Definition of "Safe Assets": Safety is now relative, not absolute. Due to the weaponization of the US dollar (sanctions) and US fiscal irresponsibility, the dollar's "risk-free" status is eroding. Central banks in at-risk jurisdictions (like China or Poland) are diversifying into gold not just for inflation protection, but as a geopolitical shield outside the US system.

• Fiscal Dominance & The Price Level: Inflation is increasingly viewed as a fiscal phenomenon, not just a monetary one. Under "Fiscal Dominance," debt levels get so high that the Treasury effectively forces the Central Bank to keep rates low to prevent default. This leads to "financial repression," where inflation is allowed to run hot to erode the real value of debt, punishing bondholders and pushing capital toward real assets.

• Non-Stationarity in AI Models: A major limitation of applying Machine Learning to finance is that markets are "non-stationary." Unlike a picture of a dog (which always looks like a dog), market rules and correlations change constantly. AI models trained on past data often fail because the underlying "subject" they are analyzing transforms completely during regime shifts.

• Conditional Correlations: The traditional 60/40 portfolio relies on stocks and bonds moving in opposite directions. However, this correlation is conditional on inflation. When inflation rises above ~3%, stocks and bonds often move together (positive correlation), failing to protect the investor. This necessitates a "Global Macro" approach, using alternatives like commodities or managed futures to provide true diversification.

Quotes

• At 0:05:38 - "You're going to need sort of the fuel, which is excess money... but you also need a narrative and a story associated with a bubble. What we find with a lot of bubbles is that it's for those assets that are very hard to value." - Explaining why bubbles cluster in tech/crypto rather than traditional sectors.
• At 0:07:05 - "People sometimes forget that a lot of commodities... [have] supply constraints... So if you have some optimistic euphoria... where are you going to get the supply? The supply is constrained so you get these demand shocks." - Differentiating commodity price spikes from equity bubbles.
• At 0:11:32 - "A risk is something that's measurable, I can count it... But then there's an uncertainty... which is not countable. I don't have any past events I could look to." - Defining the framework for analyzing unknown market drivers like AI.
• At 0:16:20 - "What do you think of the US inflation relative to Emerging Market inflation?... If you look at a basket of EM countries... their inflation is actually lower than US inflation even now." - Highlightng a major reversal in global economic structure where EM debt is becoming more attractive.
• At 0:26:28 - "I'm spending more time thinking about the concept of a safe asset as a relative concept as opposed to an absolute concept." - Explaining why capital flows are rotating out of the US despite it being the global reserve currency.
• At 0:29:08 - "There is price-driven regimes, there is policy/structural regimes, there are sentiment regimes... When you look underneath the surface... you have to decompose this into a lot of different types of regimes. And that's where you can be able to create your edge." - Explaining that regime analysis requires dissecting specific drivers, not just looking at volatility.
• At 0:30:05 - "If the stock/bond correlation changes... that has a big impact on what type of alternatives you want to buy... But when that correlation starts to go positive... that's a great time to own a lot of hedge fund strategies." - Highlighting the practical portfolio construction implications of inflation-dependent correlations.
• At 0:35:40 - "The pandemic had all the characteristics of a war. We had constrained demand, we had excess monetary policy... so what happens when we came out of that? That's why we had a sort of inflation burst." - Contextualizing the recent inflation spike by comparing the economic mechanics of COVID-19 to a war economy.
• At 0:46:20 - "It's a fiscal theory of inflation, it's not a monetary theory... you're going to get inflation because people are going to take your money out of... debt assets... [and] start to buy real assets as opposed to financial assets." - Explaining why debt sustainability concerns directly fuel inflation and change asset allocation preferences.
• At 0:58:50 - "We show 100,000 pictures of dogs... after a while it learns how to find a picture of a dog in a photo. Well, if the dog is changing through time... its characteristics are changing... then we find that it's a lot harder to do this." - A powerful analogy explaining why AI struggles in financial markets compared to other fields.

Takeaways

• Monitor Inflation for Portfolio Defense: Do not rely on bonds to hedge stocks if inflation is structurally above 3%. In this regime, correlations flip positive. You must actively allocate to "alternatives" (commodities, managed futures) to achieve actual safety.
• Distinguish Narrative from Value: When investing in "uncertain" technologies like AI, recognize that current prices are driven by stories, not math. Be prepared for volatility that reflects sentiment shifts rather than business fundamentals.
• Watch for Fiscal Dominance Signals: If the Central Bank seems to lose independence or coordinates too closely with the Treasury to keep rates low despite high inflation, expect "financial repression." In this scenario, rotate out of long-term bonds and into real assets (gold, real estate).
• Beware the "Factor Zoo": Be skeptical of complex quantitative strategies or AI-driven funds that claim to have found a "new edge." These are often over-fitted to past data (stationary) and will likely break when the market regime shifts (non-stationary).
• Follow the "Safe Asset" Migration: Acknowledge that gold is no longer just a "bug" trade but a geopolitical necessity for central banks. Follow the smart money: if central banks are diversifying away from the dollar due to sanctions risk, consider similar diversification for your own portfolio.
• Simplify Your Models: In high-uncertainty environments, simple, robust strategies (like trend following) often outperform complex predictive models. Don't try to predict the specific future; instead, use systems that react quickly to whatever trend actually emerges.