Goldman Says AI Trade Isn't Dead — It's Just Rotating Into Storage and Data Centers. t check.

Pump, dump, debug. Repeat. That's the rhythm of this AI trade, and Goldman just threw a wrench into the narrative that everyone's favorite momentum play is bleeding out. Their latest note isn't a eulogy for AI. It's a sector rotation call dressed up in quant jargon. And honestly? The data's got teeth. Let's cut through the noise. The AI对冲组合 dropped 10% in five days. High-beta momentum got slapped with a 12% drawdown. Retail's screaming bloody murder. But Goldman's sitting there with a straight face saying the AI trade isn't over. Typical. Because underneath the panic, the momentum factors are quietly reshuffling. Software just overtook semiconductors as the heaviest weight in the three-month momentum long book. Semis and the whole "AI complex"? They're now in the short book. That's not a crash. That's a rotation. Now, context. We've been here before. I've audited enough Solidity code during the 2017 ICO madness to know that when the crowd piles into one narrative, the smart money starts looking for the next exit. This isn't 2017, but the pattern's identical. Everyone's chasing the same GPU names, same AI ETFs, same "buy anything with AI in the ticker" strategy. And then the music slows. Goldman's saying the party's not over — just moving to a different room. The room with storage cabinets and cooling towers. Here's the core of their thesis, and it's worth breaking down with a debugger's eye. They're pointing at storage and data centers as the most attractive tactical plays right now. Their reasoning? "Profit recovery hasn't fully reflected in stock prices." That's a valuation gap. The market's been so busy staring at Nvidia's revenue hockey stick that it forgot the companies actually housing and serving all that compute. Micron, Dell, Supermicro — these aren't sexy. But they're the picks-and-shovels of the AI gold rush, and their earnings expectations are still lagging reality. That's a classic contrarian setup. Let me pull from my own experience here. Back in DeFi Summer 2020, I spent weeks dissecting Uniswap and Compound. The yields were insane, but the real money was in the infrastructure — the oracles, the aggregators, the gas-efficient contracts. Same thing now. Everyone's obsessing over the latest model release, but who's selling the HBM? Who's renting out the racks? That's where the margin lives. But here's the contrarian angle that Goldman's note doesn't spell out. They mention money rotating into "previously ignored areas" — European and Japanese banks, gold miners, copper miners. That's not just diversification. That's a hedge against AI's power consumption becoming a real bottleneck. Copper, specifically, is the quiet tell. AI data centers are power-hungry beasts. Every GPU cluster needs massive electrical infrastructure, and that means copper for wiring, transformers, and cooling systems. The market's starting to price in the physical constraints of AI expansion. That's a narrative shift from "digital bits" to "physical atoms." And it's happening while everyone's still staring at the chip chart. Now, let's get to the meat — the actual investment thesis. Goldman's recommendation hinges on two catalysts: Nvidia's Q2 earnings (due around late August) and industry conferences in September. These events will either validate or break the current rotation. If Nvidia crushes earnings and raises guidance, the whole AI complex could rally again, pulling money back from storage and data centers. If they disappoint — even slightly — the de-leveraging continues, and the storage names become the safe harbor. The key metric to watch isn't the headline revenue. It's the capex guidance from hyperscalers. If Microsoft, Google, and Amazon keep spending on data centers, then storage demand is locked in. If they blink, we're in trouble. But here's what Goldman didn't say — and what my code-first instinct keeps nagging at. The storage and data center "profit recovery" they're betting on is largely dependent on AI inference, not training. Training is a one-time cost. Inference is recurring. If the market's finally shifting from model building to model deployment, then the demand for high-bandwidth memory (HBM) and NVMe storage becomes sticky. But if the AI hype cools and companies stop deploying models, that storage demand evaporates. The question is whether the hyperscalers' capex is building for future inference workloads or just finishing current training clusters. That distinction matters more than any momentum factor. Let me give you a concrete scenario from my 2026 AI-agent experiments. I deployed autonomous agents to trade stablecoins. The friction points weren't in the model — they were in the data pipelines and storage layers. Every transaction needed fast, reliable state management. That's the same infrastructure challenge enterprises are facing now. The software layer gets the headlines, but the data layer gets the orders. That's why I'm cautiously bullish on the storage names — but only if they're tied to inference workloads, not just training. Now, the risk matrix. Top risk: Nvidia's earnings trigger a second wave of de-leveraging that drags down everything, including storage. That's a real possibility. The market's fragile. One bad print and the rotation becomes a rout. Second risk: the "profit recovery" in storage is already priced in by the time retail catches on. Goldman's not the only one reading these momentum signals. Third risk: macro liquidity tightening. If the Fed surprises, all risk assets get hit, and AI's not immune. But here's the opportunity that most people are missing. The rotation into non-AI sectors — banks, miners — isn't a rejection of AI. It's a recognition that AI's benefits are spreading beyond the core tech sector. AI-driven automation in banking, AI-optimized mining operations, these are real use cases. The market's finally pricing AI as a horizontal technology, not a vertical sector. That's a mature perspective. And it opens up a whole new set of value plays that don't have "AI" in their name. So what's the takeaway? Don't panic-sell the AI trade. But don't blindly buy the dip either. The game has changed from "own everything AI" to "own the right AI infrastructure." Storage and data centers look like the smart money's next stop. Nvidia's earnings are the pivot point. If you're not already positioned, wait for the earnings reaction. If you are, tighten your stop-losses. And keep an eye on copper — that's the real tell for whether this AI buildout is for real or just another cycle of hype. Pump, dump, debug. Repeat. The debug stage is where the real money gets made. t check. Goldman's note is a rare moment of honesty from a sell-side desk. They're not saying "sell everything." They're saying "sell the crowded trade, buy the overlooked one." That's the kind of advice that actually works. But remember, their clients might be the ones holding the storage stocks. Conflict of interest? Maybe. But the data's on their side this time. The momentum shift is real, and the valuation gap is measurable. I've seen enough bull markets to know that the second phase always favors the picks-and-shovels. This is phase two. Don't miss it. Final thought: The AI trade isn't dead. It's just taking a coffee break. And it's ordering a storage upgrade while it's at it. Keep your eyes on the earnings calendar, keep your position sizing tight, and for god's sake, don't chase the green candles without checking the red flags underneath. t check.