The numbers hit my terminal at 14:32 Stockholm time, and they didn't make sense. Not immediately. Goldman Sachs' equity strategists have been running the numbers on the so-called 'AI trade' โ the momentum-fueled basket of semiconductors, power utilities, and everything Nvidia-touched that printed money for the past eighteen months. The report confirms a structural fracture. The 'AI trade' is not over โ Goldman is explicit on that point โ but the era of buying the whole sector and watching it lift is gone. What we are looking at now is a rotation, a violent, capital-driven re-pricing of who actually benefits from AI infrastructure spend. The momentum signal has flipped. Software has replaced semiconductors as the largest weight in the three-month momentum long book. Semiconductors and the broader 'AI complex' have entered the short basket. If you are still positioned for a straight-line rally, you are now the exit liquidity for the algorithmic funds that have already re-priced this trade.
This is not a story about AI failing. This is a story about AI capital entering its 'Show Me The Revenue' phase. The market has stopped paying for potential and is now demanding earnings per share. Goldman's data shows that AI hedge fund portfolios dropped 10% in five days, and high-beta momentum portfolios took a 12% hit. That is a liquidation event, not a slow drift. It smells like a systematic margin call on a crowded trade. But the deeper signal, the one that actually matters for survival, is where the money is being re-routed. Goldman's note explicitly flags storage and data centers as the highest-conviction tactical trade, citing a 'valuation gap' โ meaning the profit recovery in these sectors hasn't yet been priced into the stock prices. This is the alpha hiding in the rubble. The market is realizing that the entire AI edifice is built on physical, power-hungry, storage-heavy infrastructure. The chips get the headlines, but the data centers are where the money is going to be made when the dust settles.
Let me bring in my own forensic lens here. Over the past 48 hours, I have been stress-testing the implied logic of Goldman's call against the on-chain and market microstructure signals I track. The recommendation to go long storage and data centers is not a bet on AI innovation. It is a bet on physical, industrial, and electrical engineering. This is the most important signal in the report. When a global macro powerhouse like Goldman starts recommending Dell, Super Micro, and Micron over Nvidia, they are not saying Nvidia is a bad company. They are saying the beta is bad. The risk-reward is wrong. Nvidia is a zero-sum game now; it is a crowded trade where every earnings whisper is a risk event. Storage and data centers, on the other hand, have a more predictable growth curve because they are directly tied to the buildout of physical capacity. I have been tracking the capital expenditure commitments of the major cloud providers, and the trend is undeniable. The bottleneck has shifted. The bottleneck is no longer the GPU design; it is the physical plant โ the power, the cooling, and the storage to hold the training data.
This sector rotation is the market's way of admitting that the AI buildout is a real-world infrastructure project. The initial 'vaporware' phase, where the market was buying AI dreams on a slide deck, is over. Now we are in the 'steel in the ground' phase. The copper miners are getting a bid. The gold miners are getting a bid. European and Japanese banks are getting a bid. This is not a risk-off flight to safety. This is a rotation into the physical world. It is a bet that the AI economy will require massive amounts of physical resources โ copper for data center wiring, electricity for compute, and bandwidth for data transmission. The contrarian angle here is not the rotation itself. The contrarian angle is that this rotation is a lagging indicator, not a leading one. The momentum funds are now buying the picks-and-shovels trade. But if you are a true forensic analyst, you know that the storage and data center trade is already semi-crowded. The 'valuation gap' Goldman cites will close quickly as more capital piles in.
The question you should be asking yourself is not 'Is AI over?' The question is 'Has the market already priced in the next stage of AI infrastructure?' My signal filter tells me that the market is pricing in a steady-state AI economy. But what happens if the Nvidia Q2 earnings report, due August 28, throws a wrench into the works? If Nvidia management guides down or hints at a GPU surplus, the entire infrastructure trade gets hit. The sell-off in AI equities, the aggressive deleveraging, is not a statement about AI technology; it is a statement about market structure. It is the market clearing out the 'weak hands' โ the retail traders and ETF buyers who were blindly buying the index. It is a painful, necessary process. The true signal here is the decoupling of the stock price from the underlying earnings. As Goldman notes, the sector has had a dramatic re-rating that has outpaced the actual profit growth. That is the definition of a bubble in a narrow sense โ not the entire sector, but the specific stocks that have run too far too fast.
The core insight from the report is not the rotation out of Nvidia. It is the specific identification of the next vector. Goldman's call on storage and data centers is a bet on the distribution of AI, not the training. The training model was the past; the inference is the future. The inference requires data to be stored and served at high speeds. It requires infrastructure. The market is betting that we are moving from the 'Epoch of the Model' to the 'Epoch of the Data Center'. This is the framework that the retail investor needs to grasp. We are shifting from a narrative-driven market to a fundamental-driven one. The sell-off in the AI basket is a feature, not a bug. It is a healthy market removing the excesses that were built up during the AI hype cycle. The stock price is finding its technical level.
But here is the contrarian angle. I want to deconstruct the 'Goldman say storage' trade. The report assumes that the 'profit recovery' in storage is not yet priced in. I question the timing. The earnings recovery in storage is contingent on the incremental demand from AI inference workloads. If the AI application layer does not materialize as fast as the infrastructure, the storage will be over-built, and the 'recovery' will be a one-time inventory restock, not a multi-year growth cycle. Due diligence is just paranoia with a spreadsheet. My spreadsheet tells me that the capital being dumped into AI data centers is at a cyclical high, and if the demand does not meet the supply, we will see a storage glut. The play is not to buy the storage basket; the play is to identify the storage company with the best AI-specific content. The commodity memory makers might get a valuation multiple, but the custom, AI-optimized storage solution providers are the ones with real pricing power. That is the nuance that is missed in the broad stroke of the Goldman call.
Finally, I am going to stress-test the 'AI trade not over' thesis. Goldman is telling you the market is not over. But what is the baseline for the Nvidia stock? The market has priced Nvidia for a flawless execution. The 'Q2 earnings and September industry conference' as catalysts. That is the most dangerous phrase in a market analyst's vocabulary. If the earnings are good, the stock might trade flat, because it is already priced in. If they are great, the market might rally for a day before realizing that the capex cycle has peaked. The AI trade is not a trade anymore; it is a referendum on the state of global capital expenditures. The smart money is not positioned for the Nvidia earnings; it is positioned for the aftermath. I am watching the credit spreads for tech companies. I am watching the copper futures. I am watching the energy futures. These are the leading indicators. Nvidia's report is a lagging indicator of capital flows. If copper is holding and credit is stable, the AI trade is just rotating. If the credit starts to break, we are in a different bear market entirely.
Here is the takeaway: The AI trade is not over, but it is no longer a 'buy everything' trade. It is a 'pick your poison' trade. The market is in a massive, massive transition from a beta phase to an alpha phase. The next 30 days are the most critical. Watch the August 28 Nvidia earnings, not for the EPS, but for the management's tone. Then watch the September industry conferences, not for the tech announcements, but for the procurement forecasts. The market is going to find the next winner. The window for the 'free money' trade has closed. Due diligence is just paranoia with a spreadsheet. And right now, my spreadsheet is telling me that the easy money has been made. The next move is for the analysts who can read the physical infrastructure buildout, not just the spec sheet. The signal is in the wires, not the chips.
