The signal came from an unexpected corner. Abby Joseph Cohen, the woman who called the 1990s bull market from a desk at Goldman Sachs, stood in front of a room in May 2024 and said what most portfolio managers refuse to whisper: the economy is uneven, and AI investment is unsustainable.
Let me translate that into the language I speak. She’s saying the tape is lying. The indices are green. The narrative is euphoric. But underneath, the order flow is fragmented, the liquidity is concentrated, and the fundamentals are diverging faster than a stablecoin depeg. This is not a prediction. This is an observation from someone who has read the balance sheet of America for forty years.
I’ve spent the last decade in the crypto trenches, auditing smart contracts and hunting yield. But I started in traditional finance, and I know a structural warning when I hear one. Cohen isn’t talking about a recession. She’s talking about a rotation that hasn’t happened yet. A repricing that the market hasn’t agreed to.
Let me show you what she’s actually seeing. Because if you think this is just another macro pundit hedging her bets, you’re missing the trade.
The Context: A Two-Speed Economy
Cohen’s core thesis is simple: the US economy is not one economy. It’s two. The first is powered by AI infrastructure, semiconductor fabs, data centers, and the relentless demand for Nvidia’s H100 GPUs. This economy is booming. Capital is pouring in. The second economy is everything else: traditional manufacturing, consumer discretionary, small business, commercial real estate. This economy is flatlining.
You don’t need a macro PhD to see this. You just need to look at the earnings reports. The S&P 500 is being carried by a handful of mega-cap tech names. The equal-weight index is going nowhere. The breadth of the market is terrible, and that’s not a bullish signal. It’s a sign that the tide has stopped lifting all boats. It’s only lifting the ones with the most expensive engines.
From my seat in DeFi, this looks familiar. It’s the same pattern we saw in 2021 when total value locked was surging but only in a few yield farms. The narrative was “DeFi is taking over.” The reality was that 80% of the yield was coming from 5% of the protocols, and most of that yield was printed, not earned. When the printing stopped, the farms collapsed. The same thing is happening in the AI trade. The revenue is real, but the expectations are astronomical. The gap between what Nvidia is selling and what the market is pricing is the gap between a bull market and a bubble.
Cohen is not anti-AI. She’s anti-misallocation. She’s saying the capital being poured into AI is disproportionate to the current economic reality. It’s pulling resources away from other sectors, creating an uneven recovery that is fragile by definition. And when the AI trade wobbles, the whole market will feel it, because there’s no second engine to catch the fall.
The Core: What “Unsustainable” Actually Means
Let’s get technical for a second. When a macro strategist says “unsustainable,” she’s not making a moral judgment. She’s making a balance sheet calculation. She’s saying the current rate of capital expenditure on AI cannot be maintained without a corresponding increase in end-user demand. In other words, the supply side is running ahead of the demand side. That’s a recipe for overcapacity.
I’ve seen this movie before. In 2017, I was auditing the 0x protocol smart contract, and I noticed a similar pattern in the ICO market. Projects were raising tens of millions of dollars based on whitepapers that promised decentralized everything. The code was often a mess. The use cases were speculative. But the money kept coming because the narrative was strong. I spent six weeks manually auditing the v2 contract, found three critical reentrancy vulnerabilities, and published them. The market didn’t care. The token price kept rising. But when the music stopped, the projects with no revenue and no users went to zero. The ones that survived were the ones that had actually built something.
The same logic applies to AI. There are companies building real things: Nvidia, Microsoft, maybe a few others. But there are also a hundred companies that are just “AI-powered” versions of existing software, with no moat and no revenue. They’re riding the narrative, not the fundamentals. And when the narrative shifts, their valuations will compress faster than you can say “overbought.”
Cohen’s warning is a reminder that capital flows are not permanent. They are driven by incentives, and incentives change. If the Fed has to keep rates higher for longer to fight inflation, the cost of capital goes up. That hurts high-multiple, long-duration assets the most. AI stocks are long-duration assets. They’re priced on the expectation of exponential growth ten years out. If the discount rate goes up, that growth is worth less today. The math doesn’t care about your conviction. Code doesn’t care about your feelings.
Let’s talk about the transmission mechanism. Cohen mentioned the economy is uneven. That’s a fancy way of saying the Fed’s tightening is working too well in some sectors and not at all in others. Housing is frozen. Commercial real estate is struggling. But AI capex is still booming because the private market is flooding it with cash. This is a classic sign of a policy divergence. The central bank is trying to cool the economy, but the private market is running the opposite way. This creates a tension that has to resolve. Either the AI trade cools, or the Fed has to accept that its policy is ineffective and change course. Either way, the volatility will be significant.
The Contrarian Angle: The Smart Money Is Already Rotating
Here’s the part the retail crowd doesn’t see. While everyone is FOMOing into AI stocks, the smart money is quietly hedging. I’ve been tracking institutional flows since the Bitcoin ETF approval in January 2024. I ran a delta-neutral arbitrage between the spot ETF and the futures market, capturing a 12% spread over three months. That trade worked because the institutions were buying the ETF for exposure, but the futures market was pricing in a different reality. The basis was too wide. That’s a signal.
What does that signal say? It says institutions are not as bullish as the headlines suggest. They’re using derivatives to hedge their downside. They’re buying puts on the Nasdaq. They’re rotating into defensive sectors. They’re looking at the same uneven economy Cohen is describing, and they’re positioning for a slowdown.
Panic sells, liquidity buys. That’s the rule. The panic is coming, and the liquidity is already positioning for it.
Let me give you a concrete example. In 2022, when FTX collapsed, I moved $2.5 million to self-custody hardware wallets within 48 hours. I also shorted USDT during its brief depeg, profiting $300,000. I did this because I trusted the market signal over institutional loyalty. The signal was clear: the centralized exchange model was broken. The same kind of signal is emerging now in the equity market. The concentration in AI is a centralized risk. When it breaks, it will break hard.
Cohen is not a crypto person. She doesn’t care about Bitcoin or Ethereum. But her warning applies directly to the crypto market. The same capital misallocation that’s happening in AI is happening in crypto. Look at the Layer 2 wars. Look at the cross-chain bridge hacks. Over $2.5 billion has been lost to bridge exploits, and yet the industry still depends on them. Why? Because the incentives are misaligned. The VCs want to fund new narratives, not fix old ones. They want to sell you a new token, not make the old one work.
This is the structural arbitrage I look for. The gap between what the narrative promises and what the code delivers. In crypto, that gap is often massive. In AI, it’s getting there. Cohen is pointing at the same gap from the macro side. She’s saying the economy is not delivering the broad-based growth the market is pricing. She’s saying the AI trade is not delivering the productivity gains that justify the valuations. And when the gap becomes too wide, the market corrects.
The question is timing. And that’s the part nobody knows.
The Takeaway: What This Means for Your Portfolio
Let’s be clear. I’m not saying sell everything and go to cash. That’s a coward’s trade. I’m saying you need to understand the structure of the market you’re in. If you’re holding AI stocks, you need to ask yourself: is this company generating real revenue, or is it just riding the narrative? If it’s the latter, you’re holding a bag that’s about to get heavier.
If you’re in crypto, the same question applies. Are you holding a token with real utility and real yield, or are you holding a narrative? Yield is the bait, rug is the hook. The market is full of traps, and the biggest trap right now is the assumption that the trend will continue forever.
Here’s what I’m doing. I’m maintaining my DeFi yield strategies, but I’m adding hedges. I’m shorting the most overvalued AI-adjacent tokens. I’m increasing my allocation to stablecoins and real-world assets. I’m paying attention to the signals Cohen highlighted: the core PCE, the ISM manufacturing PMI, the weekly jobless claims. These are the data points that will trigger the next move.
I’m also watching the yield curve. When the 10-year and 2-year spread starts to steepen, that’s a sign the market is pricing in a policy mistake. That’s when you want to be in cash or short-duration assets. That’s when the panic will hit.
But here’s the thing. The panic will also present an opportunity. When the AI trade breaks, the money will rotate. It will go to the sectors that have been neglected: value stocks, defensive plays, maybe even crypto if it can prove its utility. The key is to be positioned before the rotation happens. You don’t want to be the last one selling the AI bubble. You want to be the one buying the dip in something that actually has a future.
Cohen’s warning is not a sell signal. It’s a risk management signal. It’s a reminder that the market is not a monolith. It’s a collection of bets, and some of those bets are better than others. Your job is to figure out which ones are which.
I’ve been through 2017, 2020, 2022, and 2024. I’ve seen bubbles form and burst. I’ve seen narratives die and new ones emerge. The one thing I’ve learned is that the market always finds a way to humble the overconfident. The current AI trade is the most overconfident trade I’ve seen in a decade. It will be humbled. The only question is whether you’ll be on the right side when it happens.
Code doesn’t care about your feelings. The economy doesn’t either. The sooner you accept that, the better you’ll trade.
Abby Joseph Cohen is telling you the truth. The market is uneven. The AI trade is unsustainable. The smart money is already positioning. Are you?