Over the past seven days, the AI hedge basket has shed 10% of its value. The high-beta momentum portfolio has dropped 12%. On the surface, this looks like the beginning of the end for the AI trade that has dominated markets for eighteen months. But Goldman Sachs is telling a different story—one that deserves closer scrutiny, especially for those of us who have spent years tracing the hidden vulnerabilities in code and capital flows alike.
The report, which crossed my desk this week, contains a counter-intuitive signal buried beneath the headline numbers. While semiconductor and AI complex stocks have entered the short basket, software has quietly replaced semiconductors as the largest weight in the three-month momentum long portfolio. And perhaps most tellingly, Goldman is explicitly recommending storage and data center names—not because they're cheap, but because their profit recovery hasn't been fully priced into the market.
This is not a retreat from AI. It's a rotation within it.
The Context: What Goldman Is Actually Measuring
To understand why this matters, we need to step back and look at the mechanics of how institutional money tracks the AI theme. Goldman's momentum factor analysis isn't a prediction—it's a lagging indicator that reflects where capital has already flowed over the past three months. When the report says "software has replaced semiconductors in the momentum long basket," it's telling us that the marginal buyer has shifted their attention from the picks-and-shovels of AI (chips) to the application layer (software) and the physical infrastructure that makes AI deployment possible (storage, data centers).
This is a classic pattern in technology adoption cycles. In the early innings, capital floods into the most visible beneficiaries—the chip makers, the cloud providers, the companies with "AI" in their ticker. But as the technology matures and the market becomes more discerning, the focus shifts to where the actual bottlenecks are. Right now, that bottleneck isn't compute. It's storage, power, and the physical infrastructure required to run AI workloads at scale.
The key insight here is that Goldman is not saying AI is over. They're saying the undifferentiated buying phase is over.
The Core Analysis: Reading Between the Lines of the Momentum Shift
Let me break down what this rotation actually tells us, based on my experience auditing infrastructure projects and watching capital flows through the crypto and AI ecosystems.
The Semiconductor Short Signal
When Goldman says semiconductors and the AI complex have entered the short basket, it's not a fundamental indictment of Nvidia or AMD. It's a statement about positioning. The market has become crowded in these names, and the marginal buyer has been exhausted. This is similar to what we saw in crypto during the 2021 bull market peak—the moment when everyone who wanted to own Bitcoin already owned it, and the price could only go down as leveraged longs were forced to unwind.
The 10% drawdown in the AI hedge basket over five days is the equivalent of a liquidation cascade. It's not a fundamental repricing—it's a positioning reset. And Goldman's recommendation to look at storage and data centers suggests they believe the reset is creating opportunities in areas that were overlooked during the semiconductor mania.
The Storage and Data Center Opportunity
This is where the report gets interesting. Goldman's thesis is that storage and data center companies have a "significant valuation gap" because their profit recovery hasn't been fully reflected in their stock prices. In plain English: the market has been so focused on GPU supply that it's forgotten that you can't run an AI workload without storage, cooling, power, and physical facilities.
Based on my audit experience, this is a classic infrastructure lag effect. When a new technology wave hits, the upstream components get priced first—the chips, the specialized hardware. But the downstream infrastructure—the mundane stuff like hard drives, power distribution, and cooling systems—takes longer to be repriced because it's less exciting and more capital-intensive.
The companies Goldman is likely referring to—Dell, Super Micro, Micron—are not glamorous AI names. They're the workhorses of the data center economy. And their profit recovery is tied to a simple fact: AI data centers need massive amounts of storage and compute infrastructure, and that demand is only now starting to show up in earnings.
The Non-AI Rotation Signal
Perhaps the most underappreciated signal in the report is the mention of capital flowing into European and Japanese banks, gold miners, and copper stocks. This is not a rejection of AI—it's a hedge against AI concentration risk. When institutional investors have been overweight AI for eighteen months, they naturally start looking for uncorrelated assets to balance their books.
But there's a deeper signal here. Copper stocks being mentioned alongside AI infrastructure is not a coincidence. AI data centers are massive consumers of copper—for power distribution, for cooling systems, for the physical infrastructure that supports GPU clusters. The fact that copper miners are being mentioned in the same breath as AI suggests that the market is starting to price in the physical resource requirements of AI at scale.
This is the kind of cross-sector signal that gets lost in the noise of daily market commentary, but it's exactly where the real information is.
The Contrarian Angle: What Goldman Isn't Telling You
Now, let me put on my risk-first hat and look at what's missing from this analysis. Because there are several blind spots in Goldman's report that deserve attention.
The Interest Rate Elephant
Goldman's recommendation of storage and data center stocks is predicated on the assumption that profit recovery will drive stock prices higher. But this ignores the elephant in the room: interest rates. Data center companies are among the most capital-intensive businesses in the market. They borrow heavily to build facilities, and their profitability is directly tied to the cost of capital.
If the Federal Reserve doesn't cut rates as quickly as the market expects, the "profit recovery" Goldman is betting on could be delayed or diminished. The valuation gap they've identified might not close—it might widen.
The AI Capex Cliff
Here's a question Goldman doesn't answer: what happens when the hyperscalers—Microsoft, Google, Amazon—decide they've built enough data centers? The current AI infrastructure buildout is unprecedented, but it's not infinite. At some point, capital expenditure will normalize, and the companies selling storage and data center equipment will see their order books shrink.
The market is pricing in a multi-year AI infrastructure boom. But the history of technology cycles suggests that infrastructure buildouts are lumpy, not linear. We saw this in the dot-com era, where fiber optic companies overbuilt and then spent a decade digesting the excess capacity.
The Concentration Risk
Goldman's report is a sell-side document, which means it has inherent biases. The firm's clients include companies in the storage and data center sectors. The recommendation to buy these stocks could be influenced by investment banking relationships that aren't disclosed in the report.
This doesn't mean the analysis is wrong—it means it should be treated as one data point, not gospel. The same way I approach a smart contract audit: I verify the claims independently, I stress-test the assumptions, and I look for the edge cases that the original author might have missed.
The Takeaway: What This Means for the Next Six Months
So where does this leave us? Let me offer a forward-looking perspective based on the signals in this report and my experience watching technology adoption cycles.
The AI trade is not over—it's rotating. The next phase of the AI market will be characterized by differentiation, not indiscriminate buying. Companies that can demonstrate actual AI-driven revenue growth will be rewarded. Companies that are merely "AI-adjacent" will be punished.
For the storage and data center sector, the window of opportunity is real but time-limited. The profit recovery Goldman is betting on will likely materialize over the next two to four quarters. But by the time it's fully reflected in stock prices, the opportunity will be gone. The key is to identify which companies have the strongest order books and the most efficient capital structures—the ones that can weather a potential rate shock or capex slowdown.
The critical catalyst to watch is Nvidia's Q2 earnings and the September industry conferences. These events will provide the clearest signal on whether AI capital expenditure is accelerating, stabilizing, or decelerating. If Nvidia's guidance is strong, the rotation into infrastructure will likely continue. If it disappoints, we could see a second wave of deleveraging that hits storage and data center stocks just as hard as semiconductors.
The quiet truth is that infrastructure always lags the hype cycle. The chips get the headlines, but the storage, the power, and the physical facilities are where the real value accumulates over time. The question is whether you have the patience to wait for the market to recognize it.
In the meantime, I'll be watching the momentum factors, the capital flows, and the earnings revisions with the same attention I give to smart contract audits. Because in markets, as in code, the vulnerabilities are always hidden in the details. And the details are telling us that the AI trade is far from over—it's just changing shape.