The AI Trade Is Deleveraging: Goldman's Rotation Signal and What It Means for Capital Allocation

Most analysts are still chasing the AI narrative. They're late. The data from Goldman's latest positioning report tells a different story. The high-beta momentum portfolio lost 12% in a week. The AI hedge basket dropped 10% in five days. Leverage is coming off the table. This isn't a crash. It's a structural rotation. And if you're not reading the order flow, you're going to get run over. Let me be clear about what I see. This is not the end of the AI trade. Goldman explicitly says that. But the era of buying the whole sector and getting paid is over. The beta is gone. What's left is alpha. And alpha requires a different playbook. I've been through these rotations before. In 2020, I watched the DeFi summer turn from a broad rally into a brutal grind. The same pattern is playing out here. The first phase rewards everyone. The second phase rewards only those who understand the structural shifts underneath the surface. Goldman's report is a map. You just have to know how to read it. The Context: From Beta to Alpha The market structure has changed. For the past eighteen months, the AI trade was simple. Buy semiconductors. Buy anything with a GPU in the supply chain. Collect the gains. The tide lifted every boat. But that tide is receding. The numbers are stark. The high-beta momentum portfolio, the vehicle that captured the explosive upside of the AI rally, just suffered its worst week in recent memory. A 12% drawdown is not a blip. It's a signal. The AI hedge basket, a more curated collection of AI-related equities, fell 10% in five days. Leverage, which had built to extreme highs, is now being unwound. This is the classic signature of a deleveraging event. The crowd is selling. The question is: what are they selling, and what are they buying instead? Goldman's answer is revealing. Semiconductors and AI complexes have moved into the short portfolio. That's not a tactical hedge. That's a statement. The market is now betting against the very names that led the rally. Meanwhile, software has taken the top weight in the three-month momentum long portfolio. The rotation is clear. Capital is moving from the picks-and-shovels of the AI gold rush to the miners themselves. From infrastructure to application. The Core: Dissecting the Order Flow Let's break down the order flow. This isn't just a story about sector rotation. It's a story about where the market believes value is being created and captured. First, the semiconductor short. This is the strongest signal in the report. The market is pricing in a slowdown in AI training demand. Or perhaps a shift in the competitive landscape. Nvidia's dominance is no longer a foregone conclusion. AMD's MI series is gaining traction. Custom ASICs are eating into the general-purpose GPU market. Cloud providers are designing their own silicon. The moat is narrowing. But there's a deeper issue. The market is also pricing in the risk of export controls. The US restrictions on advanced chip sales to China have created a ceiling on the total addressable market. This is a structural headwind, not a cyclical one. It's a permanent reduction in demand potential. Second, the software move. Software is now the top weight in the momentum long portfolio. This tells me the market believes AI is about to start generating real revenue. Not just promises. Not just user growth. Actual income. AI coding assistants. AI agents. Enterprise AI SaaS. These are the categories that are starting to deliver. The shift is from training to inference. From building the model to running the model. From the one-time cost of training to the recurring cost of serving. This is a fundamental change in the demand profile. And it favors different companies. Third, the storage and data center call. Goldman explicitly identifies storage and data centers as the most tactically attractive sectors. Their logic is simple: the profit recovery has not yet been fully reflected in the stock prices. This is a valuation gap. Let me dig into this. The AI inference wave requires massive storage infrastructure. Model weights. Training data. Inference caches. All of this needs to live somewhere. The demand for HBM and enterprise SSDs is exploding. The supply is concentrated in three players: Samsung, SK Hynix, and Micron. That's an oligopoly with pricing power. Data centers are similar. The shift to inference requires distributed deployment. Not just a few massive training clusters, but many smaller, geographically distributed facilities. This drives utilization rates and rental prices. The operators with scale and efficiency are set to benefit. Based on my experience auditing infrastructure projects, I can tell you that the storage and data center story is more than just a narrative. The revenue is real. I've seen the contracts. The demand is not speculative. It's backed by actual deployment plans from hyperscalers and enterprise customers. The Contrarian Angle: The Crowd Is Wrong About the Crash The retail narrative is simple. AI is a bubble. It's popping. Run for the hills. This is wrong. What's happening is not a bubble bursting. It's a reallocation of capital within the AI ecosystem. The total investment in AI is not decreasing. It's shifting. From hardware to software. From training to inference. From the US to other markets. Look at where the capital is flowing. Goldman notes that funds are rotating into European and Japanese banks, gold miners, and copper stocks. This is not a retreat from AI. It's a search for value outside the crowded AI trade. Copper is particularly interesting. The AI data center buildout requires enormous amounts of copper for power transmission and cooling systems. This is an indirect play on the AI infrastructure buildout. The market is so crowded in the direct plays that it's finding value in the secondary beneficiaries. But here's the blind spot. The retail crowd is focused on the headline numbers. The 12% drop. The 10% decline. They see these as signs of a crash. They're missing the structural shift underneath. Let me be blunt. The AI trade is not dead. It's maturing. The first phase was about speculation. The second phase is about fundamentals. The market is now demanding that companies show actual earnings from AI. Not just promises. Not just roadmap. Actual revenue. This is a healthy correction. It's the market doing its job. Weeding out the weak hands. Rewarding the companies that are actually executing. The risk is not the AI trade ending. The risk is being on the wrong side of the rotation. If you're long semiconductors without a clear thesis on the competitive landscape, you're exposed. If you're long software without a clear view on revenue growth, you're exposed. The market is now differentiating. And differentiation means some names will get left behind. The Takeaway: Actionable Levels and Forward-Looking Signals The market is telling you something. The question is whether you're listening. The semiconductor short is a warning. The software long is an opportunity. The storage and data center call is a tactical play. Here's what I'm watching. Nvidia's Q2 earnings are the next catalyst. If the guidance is strong, we could see a relief rally in the semiconductor names. But if the guidance is weak, or even just in line, the selling could accelerate. The market is pricing in a slowdown. Nvidia needs to prove that the slowdown is not happening. September industry conferences will provide additional signals. Watch for announcements about data center deployments, storage demand, and inference infrastructure. These will confirm or deny the rotation thesis. For the storage names, the key is the HBM supply-demand balance. Micron's earnings, typically released in late August or early September, will be a critical data point. Watch for guidance on HBM pricing and volume. If the numbers are strong, the valuation gap Goldman identified could close quickly. For data centers, the focus is on utilization rates and rental prices. The shift to inference is driving demand for distributed infrastructure. Operators with scale and efficiency will benefit. The ones with legacy assets and high debt could struggle. For software, the question is revenue growth. The momentum shift is a signal, but momentum can reverse. I need to see actual revenue acceleration. AI coding tools. AI agents. Enterprise AI deployments. These need to show up in the income statement. The bottom line is this: The AI trade is not over. But the easy money has been made. The market is now rewarding precision over enthusiasm. Structure over story. Earnings over expectations. I've been through enough cycles to know that the moment the crowd declares a trade dead is often the moment the smart money starts building positions. The question is: are you on the right side of the trade? Let me leave you with a thought. The market doesn't pay for vision. It pays for delivered earnings. The AI trade is entering its earnings phase. The companies that deliver will be rewarded. The ones that don't will be left behind. The rotation is happening now. The question is whether you're positioned for it. The market is a discounting mechanism. It's already pricing in the next phase of the AI trade. The question is whether you're paying attention to the signals. I am. And I'm acting accordingly.