The reunion had the feel of a long-overdue diagnosis. David Rosenberg of Rosenberg Research and Richard Bernstein of Richard Bernstein Advisors1, now under Janus Henderson, sat down on the Excess Returns podcast with host Matt Zeigler to do what they have always done best: connect the macro to the markets, call things by their right names, and resist the crowd. Between them, they have navigated the dot-com implosion, the housing collapse, and every cycle of monetary excess in between. Their conclusion today is blunt: the AI trade is a bubble, the rest of the world is being ignored at the investor's eventual peril, and credit markets are already writing the warning in real time.
The Fed, the Taylor Rule, and a Question Nobody Wants to Answer
The conversation opens on a genuine disagreement, which is itself instructive. Bernstein raises the Taylor Rule with the precision of a man who has been quietly watching a room full of people pretend not to see something obvious. "The Atlanta Fed has 30 different versions of the Taylor Rule," he notes, "and they do that on purpose because if we have one version, people say, oh, you're using the wrong version. So the Atlanta Fed has 30 different versions. All 30 suggest that the Fed should be hiking rates." His point is not strictly that the Fed should hike, but that a generation of conservative economists who once waved the Taylor Rule like scripture have gone conspicuously silent now that it no longer fits the narrative they prefer.
Rosenberg disagrees on the hiking question but makes a tightly reasoned case for why. He points to the Dallas Fed's trimmed mean inflation measure, running at 2.2% and decelerating from 2.7% a year prior, as the real signal on underlying price pressure. "Unit labor costs at last count, and that's productivity-adjusted wages, is running 0.5% year over year," he argues, adding that "real disposable income in the United States is actually running fractionally negative year on year." The wage-price spiral of 2021 to 2022 lasted 18 months and is gone. What exists now is a K-shaped economy where corporate profits absorb all the income growth while the consumer quietly softens. Strip out AI investment and the four-quarter trend in real GDP is running below 1.5%, well under the Fed's own estimate of potential supply growth. The disinflationary forces, in Rosenberg's view, are structural, not transitory. And the Fed does not need to do anything about AI, because the credit market is already doing it for them.
The Credit Market Is Already Doing the Work
This is the most important signal in the conversation, and both men converge on it with conviction. Rosenberg draws a direct line from the housing bubble to today: "The problems in ABS spreads and mortgage bonds was happening long before October 2007, which is when the equity market peaked. I think that the credit market will lead the ultimate rolling over of this AI trade. I don't think the Fed has to do anything." CDS spreads are widening, credit spreads are moving, and financing costs for the very companies driving the AI rally are rising. The equity market, as it did in 2006 and 2007, is still celebrating while the bond market quietly reprices the risk.
Capital Misallocation and the Housing Parallel
Bernstein's contribution to the bubble diagnosis is analytical and historical. He draws the through-line from the dot-com era, when capital flooded into TMT at the expense of energy infrastructure, to today, where the same dynamic is playing out between data centers and residential housing. "If you look at construction spending on data centers compared to construction spending on residential housing, you can see how the bubble is misallocating capital relative to housing affordability in the United States." Housing stocks are between a correction and a bear market, exactly where they should be given the cost of capital. But nobody is asking that question because the data center build-out is the only story being told. "I think the Fed does not appreciate that abnormal pricing in the financial economy, which we would call a bull market, when taken to an extreme, is just as damaging to the economy as is real inflation."
Rosenberg agrees but sharpens the 1990s comparison to make it more precise. Back then, the AI analog was TMT, but the entire economy was running above potential. Consumer strength was broad, housing was solid, non-residential construction was healthy. Today, "when you strip out the AI spend, the economy is actually very weak." A 1.5% second-quarter GDP print, he notes dryly, is "all you get for your money" despite hundreds of billions in AI capital expenditure. Without that spending, there is likely a recession underneath the surface.
Valuations, Sentiment, and Extra Innings
The sentiment picture Rosenberg paints is striking in its extremity. Mutual fund portfolio managers are carrying 1% cash. Net speculative long positions among hedge funds are stretched. Household equity allocation, per the Fed's flow of funds data, sits at a record 73%. And only 2% of global portfolio managers surveyed by Bank of America expect an economic downturn in the next year, against a historical average probability of roughly 15%. "When all the forecasts and experts agree," Rosenberg quotes from Bob Farrell's rules, "something else is going to happen."
The Shiller CAPE multiple has crossed 41, matching the highest reading since 2000 and representing a 3-sigma event versus the full historical record. Rosenberg is direct: "The US market is the only one where the CAPE has only been this expensive in the past 1.5% of the time. Think about that. We're in insanity land." Jamie Dimon called it a classic bubble in October. Warren Buffett recently described the stock market as resembling a gambling society. Both men respect the weight of that collective judgment, even if the timing remains unknowable. "You're not in the 8th or 9th inning," Rosenberg says, "you're in extra innings."
What They Own Instead
The portfolio response from both men is clear: diversify internationally, own gold as structural ballast, reduce beta, and raise the Sharpe ratio. Bernstein's firm now carries its largest non-US overweight in the history of the firm. His data shows that non-US stocks have outperformed venture capital over the last five years, a fact almost nobody is tracking. More provocatively, his analysis of the ACWI index finds roughly 200 companies globally with projected long-term earnings growth of 25% or more. "Only one of the Mag 7 is in that group," he says. "There is growth all around the world. It's just that people are unwilling to look at them."
Rosenberg adds that only two sectors within the S&P 500 carry a genuinely low correlation to the AI trade: healthcare and consumer staples. Everywhere else, from utilities to industrials to financials, the sector has become an AI derivative. "Diversification is not a dirty 15-letter word," he says, and the practical implication is to look at Southeast Asia ex-Korea and Taiwan, Japan, and select parts of Latin America, markets where the CAPE reads 50% to 60% of the time rather than 1.5%.
On gold, Rosenberg holds his $6,000 per ounce long-term target unchanged despite the recent correction, which he attributes to dollar strength during the US-Iran conflict and rising real rates, not a structural change in the demand picture from central banks. Bernstein frames it as a spare tire: "You can't predict when you're going to get a flat tire. So we always have a spare tire on hand in our car. And it's the same in a portfolio. You can't predict uncertainty." He adds a useful sentiment footnote: nobody is asking him about silver anymore, which he takes as a constructive signal after the froth of a year ago.
Five Key Takeaways for Advisors and Investors
1. The credit market is the leading indicator. Widening CDS and credit spreads in AI-related names are the signal to watch, not the equity price. The equity market lagged the credit market by roughly a year in 2006 to 2007. That same sequencing appears to be unfolding now.
2. Strip out AI spending and the US economy is fragile. Housing is contracting, non-residential construction ex-data-centers is negative year over year, and real disposable income is slightly negative. The AI boom is masking the weakness, not eliminating it.
3. Concentration risk is broader than it looks. Only healthcare and consumer staples carry low correlation to the AI trade within the S&P 500. Advisors who think their clients are diversified across sectors may be carrying far more AI exposure than the label suggests.
4. International equities carry both relative value and growth. Bernstein's data shows 200-plus companies globally with 25%-plus projected long-term earnings growth, and only one is in the Mag 7. The narrative that you must own US mega-caps to access growth is simply no longer accurate.
5. Sentiment extremes are a tactical and structural warning. A 73% household equity allocation, 1% mutual fund cash, 2% recession probability among global managers, and a CAPE multiple in the top 1.5% of all historical readings are not individually decisive but are collectively consistent with what both men identified and acted on before the two largest crashes of the modern era.
Footnote:
1 Excess Returns. "David Rosenberg and Rich Bernstein on What Ends the AI Trade — And What They Own Instead." YouTube, 8 Aug. 2026, www.youtube.com/watch?v=l81SgC8sSqo.
