By Grace Chen
HOOK: When a Wall Street Oracle Questions the Narrative
Abby Joseph Cohen doesn't scream. She doesn't need to. The woman who called the bull market of the 1990s before almost anyone else, who spent decades as Goldman Sachs' chief market strategist, speaks in measured tones. That's precisely why her recent warning carries the weight of a hundred shouting analysts.
"Uneven economy." "Unsustainable AI investing."
Two phrases. One quiet storm.
Open source isn't just about code; it's a philosophy of transparency. And what Cohen has done with these words is expose a transparency deficit in our collective market narrative—one that should concern anyone who believes markets function on fundamentals rather than faith. She's telling us the emperor's new clothes are not just ill-fitting, but woven from unsustainable threads.
CONTEXT: Who Is Abby Cohen and Why Does Her Voice Matter?
Before we parse her warning, let's establish the credibility matrix. Cohen is not a crypto enthusiast. She's not a tech blogger with a newsletter. She's a former co-chief investment strategist at Goldman Sachs, a Columbia Business School professor, and a woman who has spent nearly half a century reading macro-economic tea leaves with an uncanny track record. She was one of the first to publicly articulate the "productivity boom" thesis in the late 1990s—a thesis that seemed insane at the time but proved prescient. When Abby Cohen speaks about economic structure, the institutional world listens, even if reluctantly.
Her current role as a senior investment strategist at BNY Mellon gives her a bird's-eye view of institutional capital flows. She sees the machinery of money moving, not just the headlines. So when she says the economy is "uneven" and AI investment is "unsustainable," she's not playing to the cameras. She's reading the same data that the Fed sees, the same flows that pension fund managers see, and the same red flags that often get ignored in the frenzy of a concentrated bull market.
This warning comes at a specific moment in the current economic cycle. The equity markets are riding an AI wave that has lifted the so-called "Magnificent Seven" to valuations that resemble the late-1990s dot-com glory days—and we all remember how that ended. The "uneven economy" is not just a phrase; it's a description of a bifurcated reality: a handful of AI behemoths capturing the bulk of market cap gains while the rest of the economy—the small businesses, the traditional manufacturers, the consumer-facing sectors—feels increasingly left behind.
CORE: The Structural Anatomy of an Uneven Economy
Let me break down what "uneven" actually means in the data that Cohen and her cohort are watching. This isn't about feelings; it's about divergences that create systemic fragility.
The Earnings Divergence
The most striking manifestation of this unevenness is the earnings spread. As of the most recent quarterly earnings season, the S&P 500 ex-AI saw earnings growth that was essentially flat—zero percent growth year-over-year. But strip out the top seven AI-exposed names, and you see a different picture: the remaining 493 companies have been growing earnings at a slow, single-digit rate at best. That's not a healthy, broad-based economic recovery; that's a few monopolies lifting the average.
This is the geometry of unevenness. When you draw the line between the AI-driven top and the traditional rest, the curve looks like a staircase—a steep ascent followed by a flatline. The problem is that the flatline represents most of the real economy: construction, transportation, consumer staples, even financial services outside of the tech adjacency. These sectors are not in a recession, but they're not in a boom either. They're just stuck, feeling the weight of high interest rates and cautious consumer spending.
The Capital Misdirection Problem
When I look at where the money has gone in this current cycle, I see a concentration that is deeply reminiscent of the 2000 dot-com era—but with a twist. In the 2000s, we had a similar massive capital allocation to internet infrastructure. But today, the sheer dollar volume is staggering. The big tech giants are planning to spend over $300 billion combined on AI-related capital expenditures (CapEx) in 2024—that's more than the entire GDP of many countries. This isn't just investment; it's a bet-the-farm moment.
And here's the ethical algorithmic framing problem: that level of capital concentration is a form of societal risk. When a small group of companies has that much investment concentrated in a single technological direction, the entire economy becomes a hostage to the outcome of that one bet. If AI produces a massive productivity boom—which is a real possibility—then we're fine. But if the returns on that $300 billion in CapEx turn out to be marginal, or if the deployment time horizon is longer than the market's patience, we're looking at a potential capital waste of historic proportions.
The "Meh" Consumer, The "Wow" AI
The unevenness extends to the real economy. The consumer, who is 70% of the U.S. GDP, is not in great shape. Retail sales data has been showing a slowdown. Savings rates have dipped. Credit card debt is hitting record levels. Yet the AI sector seems to be living on a different planet, with high-end GPU chips on backorder and data center construction going 24/7.
This bifurcation is the "unevenness" that Cohen is pointing at. It's not just an economic phenomenon; it's a sociological one. When the most powerful people and companies are all-in on AI, but the average worker is still struggling with the cost of groceries and rent, you have a recipe for populist backlash, political dysfunction, and consumer retrenchment. An economy that feels like a casino for the few and a struggle for the many is not a stable economy.
The "Unisustainable" AI Investing: A Fundamental Analysis
Now, let's dig into the "unsustainable" part of Cohen's warning. This is where my technical experience and background come in.
From an applied mathematics perspective, I see AI investing as a function with two critical variables: cost of compute and price of output. The current market is pricing in an assumption that the cost of compute will continue to decrease at a historical rate while the output (AI-powered products and services) will see exponential demand. Both assumptions are dangerous.
The cost of compute is currently skyrocketing, not decreasing. Energy prices are rising, and the power requirements for training large models are enormous. The cost of high-end GPUs (like NVIDIA's H100) remains astronomical, and supply chain constraints mean that many companies are paying a premium just to get access. The economics of AI model training are becoming less attractive at the margin, not more.
The output side is equally problematic. While consumer AI products like ChatGPT have seen rapid adoption, the monetization path remains unclear. How many subscriptions can the consumer absorb? How many businesses can genuinely justify the cost of AI integration when the ROI is still uncertain? The gap between "cool" and "commercially sustainable" is wider than the current market pricing suggests.
When I analyze the cash flows of some of the most visible AI companies, I see a concerning pattern: capital expenditures are far exceeding free cash flow. The companies are burning cash to build the AI infrastructure that they believe will be the future, but the future may be further away than their balance sheets can sustain. This is exactly what we saw with the internet infrastructure buildout in 1999-2000: the "picks and shovels" companies (like Cisco, JDS Uniphase) were spending billions, but the end consumer demand didn't materialize fast enough. The result was a catastrophic correction.
The math doesn't work in the short term. The total addressable market for AI is enormous, but the current revenue base is tiny. This is a classic "asymmetric risk" profile: the potential is huge, but the current path is not sustainable without a dramatic increase in actual user demand.
The Geopolitical Supply Chain Twist
There's another layer to the "unsustainable" aspect—the geopolitical supply chain. The AI revolution is built on chips, and chips are built on a supply chain that passes through Taiwan, Korea, and Japan. Any major geopolitical disruption in the region could create a catastrophic bottleneck, strangling the AI investment thesis overnight. Cohen's "uneven economy" may also include a geopolitical component—a fragility that the market currently is not pricing.
The 2022 CHIPS Act and the subsequent export controls on high-end chips to China have actually created an artificial shortage, which has benefited the chipmakers' stock prices but has also injected a level of volatility and uncertainty into the entire AI ecosystem. A trade war escalation could turn the "unsustainable" into an immediate "correction."
CONTRARIAN: The Pragmatism Test—Why We Shouldn't Dismiss AI (or Cohen) Entirely
Now, let's apply the contrarian lens, because I'm not here to simply fear-monger. The nuance in Cohen's warning is not that AI is a fad; it's that the investment structure is risky.
Here's where I diverge from the pure "AI bubble" narrative. The technology itself is undeniably transformative. The capability of large language models (LLMs) to write code, analyze legal documents, and even draft economic reports is real and improving. The underlying blockchain technology—and the broader digital infrastructure—is also being positively impacted by AI.
The contrarian perspective is that the "unsustainability" is not necessarily a signal of an imminent collapse. It could be a signal of a "digestion" period—a time when the market corrects from an extremely high valuation to a more sustainable growth level. This could be a slow burn rather than a crash, a "valuation decompression" where the market takes 18-24 months to adjust to the fundamentals.
Additionally, the "uneven economy" might not be a permanent feature. If AI-driven productivity gains finally reach the broader business community, we could see a second wave of investment and growth that lifts all boats. The investment in AI infrastructure today could be the prerequisite for the economic "everything" of tomorrow.
So, Cohen's warning is a "yellow flag," not a "red flag." It's a call for a more prudent approach to capital allocation, not a signal to sell everything. Her "pragmatic risk integration" is about risk management, not market exit.
My Personal Experience and the Lesson of the "Geometric Metaphor"
In my own work, I've built an education platform that teaches people about decentralized finance and blockchain. I've seen how narratives can become a self-fulfilling prophecy. When the narrative is "AI is the future," the money flows to AI, the valuations rise, and then the narrative becomes "AI is overvalued," and the money flows out.
This pattern is predictable. It's not a straight line. It's a geometric curve, with a steep ascent, followed by a period of consolidation, and then a sustainable ascent. Cohen is signaling that we're in the "over-extended" part of the curve. The key is to not confuse the "peak" of the current cycle with the "peak" of the technology's long-term potential.
TAKEAWAY: The Structure of a Sustainable Future
So, what should we do with this warning?
First, recognize the "unevenness" in your own portfolio. Are you over-exposed to AI and tech? Are you ignoring the traditional sectors? The "uneven economy" suggests a need for a more balanced approach, a "barbell" strategy where you have some exposure to the AI growth but also to defensive sectors that can weather a storm.
Second, apply the "Pragmatic Risk Integration". For every AI investment, ask yourself: "What is the sustainable growth path? What is the path to profitability? What is the risk of overinvestment?" Don't buy the "story" of AI; buy the "math" of the business.
Third, think about the "geometric metaphor". The market is not a straight line; it's a curve. The current steepness is not sustainable. A flattening or even a temporary downward slope is not a crash; it's a natural correction.
The takeaway is not "sell everything." The takeaway is "be more thoughtful." Abby Cohen is not a doom-sayer; she's a data-reader. And the data is saying: the current AI investment path is not sustainable at the current pace. The economy is uneven, and that unevenness creates risk.
The future is not AI; it's balanced. It's about ensuring that the tech revolution benefits more than just a few companies and that the economic growth is broad-based. Open source isn't just a philosophy of transparency; it's a blueprint for resilience. The market needs a little more "openness" about the real risks it faces, and a little less "closed-loop" thinking that just focuses on the next miracle. AI has a magnificent future, but only if we can build an economic structure that can sustain it.
The question is not whether AI will change the world—it already is. The question is whether the way we invest in it can adapt to the unevenness of the world it is creating. And that answer is still being written.