Goldman's AI Trade Is Deleveraging: The Beta Party Is Over, But Storage and Data Centers Are the New Alpha Play

The numbers hit like a gut punch. Goldman's high-beta momentum basket? Down 12% in a single week. Their AI hedge portfolio? Off 10% in five days. The leverage that powered the AI trade is bleeding out, and the crowd that was screaming "number go up" is now whispering "what's the exit?" This isn't a crash. It's a recalibration. And if you're still trading AI like it's late 2023, you're already late to the next move. I've seen this movie before. In 2020, I was knee-deep in the DeFi yield farming frenzy, hosting Discord listening parties to gauge sentiment while YFI and SUSHI whipsawed portfolios. The pattern is identical: when the leverage unwinds, the narrative doesn't die—it just gets pickier. The difference now is that the smart money isn't asking if AI is real. They're asking which parts of the stack actually generate revenue. Goldman's latest note, dated August 23rd, is a masterclass in reading the tea leaves of a maturing trade. The headline is simple: the AI trade isn't over, but the era of buying the whole sector and praying is done. The subtext is where the real signal lives. They're telling you to stop chasing the beta and start hunting for the alpha hiding in plain sight. Here's the core of their argument, stripped of the institutional gloss. The broad-based rally that defined the AI narrative is transitioning into a phase of differentiation. The low-hanging fruit—the semiconductor names that went vertical on hype—are now being repriced. Goldman has actually moved semiconductors and the broader AI complex into their short portfolio. That's not a tactical hedge. That's a statement. Meanwhile, software has taken over as the largest weight in their three-month momentum long book. The baton has passed. The market is no longer paying a premium for the "picks and shovels" of the AI gold rush; it's starting to reward the "gold miners"—the application layers that are finally converting AI capabilities into actual revenue. But the most intriguing call, the one that most retail traders will gloss over, is the recommendation on storage and data centers. Goldman flags these as "tactically the most attractive" sectors, with the clearest valuation gap. The logic? Profit recovery hasn't been fully priced into the stocks. In plain English: these companies are making money again, but the market hasn't caught on yet. This is where my own experience kicks in. I've audited enough balance sheets to know that "profit recovery" in the storage space isn't just about AI. It's a confluence of factors. You've got the HBM (High Bandwidth Memory) boom, which is directly tied to AI training and inference. You've got the enterprise SSD upgrade cycle, which is being accelerated by the need to store and retrieve massive datasets for model fine-tuning. And you've got the data center operators who are seeing utilization rates climb as inference workloads—not just training—start to dominate compute demand. Algorithms smell fear, but they respect speed. The market is fast to punish narratives that don't deliver earnings. But it's equally fast to reward sectors where the fundamentals have quietly improved while the stock price lagged. That's the setup in storage and data centers right now. Let's dig into the contrarian angle, because that's where the edge is. The conventional take on Goldman's note is "sell semiconductors, buy software." That's a surface-level read. The deeper implication is that the AI infrastructure buildout is shifting from a training-centric model to an inference-centric one. Training requires massive, concentrated compute clusters. Inference requires distributed, efficient, and storage-heavy infrastructure. This shift is a structural tailwind for data center REITs and storage manufacturers that most investors haven't fully modeled yet. But here's the blind spot. Goldman's call is based on the assumption that the "profit recovery" in storage is AI-driven. What if it's not? What if a chunk of that recovery is just the cyclical rebound in traditional enterprise IT spending? If that's the case, the AI premium on these stocks is a mirage. I've seen this happen before—narratives get attached to cyclical upticks, and when the cycle turns, the narrative breaks. The key is to watch the earnings calls. If management teams are explicitly citing AI-related demand for HBM and enterprise SSDs, the thesis holds. If they're vague and pointing to "broad-based demand," be cautious. The other risk is the Nvidia earnings catalyst. Goldman flags the Q2 report as a key event, and it's a double-edged sword. A blowout quarter with strong guidance will validate the AI demand narrative and could pull the whole sector up. But a miss—or even just conservative guidance—could trigger another leg of deleveraging. The market is positioned for perfection, and perfection is a high bar. I remember the Terra/Luna collapse in 2022. The human cost of leverage is real, and it's not just about numbers on a screen. It's about the anxiety in the chat rooms, the fear in the DMs, the quiet panic of people who leveraged their savings on a narrative. That's why I focus on sentiment as much as fundamentals. The current sentiment is cautious optimism. No one is screaming from the rooftops, but no one is running for the exits either. It's a waiting game. Yield is a drug; exit liquidity is the cure. The market is currently in a state of withdrawal, and the cure is going to come from the sectors that can prove they're not just riding the AI wave, but actually profiting from it. So, what's the takeaway? The next few weeks are binary. Nvidia's earnings and the September industry conferences will set the tone. If the data points are strong, the storage and data center names could see a significant re-rating. If they're weak, the deleveraging continues, and the valuation gap Goldman identified will just get wider before it gets better. Chaos is just data waiting for a narrative. The narrative is shifting from "AI will change the world" to "AI is changing these specific P&Ls." The winners will be the companies that can show real, dollar-denominated progress. The losers will be the ones still selling a vision without a revenue model. I didn't get to where I am by following the herd. I got here by watching where the smart money moves and understanding the psychology behind the move. Goldman is telling you the herd is breaking up. The question is: are you going to be a sheep or a shepherd? We don't get to choose the market's direction, but we do get to choose our positioning. The data is clear. The leverage is unwinding. The rotation is real. The opportunity is in the overlooked corners of the AI stack. Storage and data centers are the quiet beneficiaries of a shift that most people are too busy watching Nvidia to see. The market is a story, and the best stories are the ones that haven't been told yet. This one is just beginning.