Hook
The numbers hit my terminal at 09:00 UTC, and they were ugly. AI hedge fund portfolios down 10% in five days. High-beta momentum baskets off 12%. The kind of violent de-risking that usually precedes a full-blown narrative collapse. Retail sentiment was already in panic mode, screaming that the AI trade was over.
But Goldman Sachs just published the counter-thesis: the AI trade hasn't ended. It's rotating.
The ledger doesn't care about your conviction. What matters is where the institutional flows are moving next. And based on the data in this report, the answer is clear: storage, data centers, and the overlooked infrastructure layer that makes AI actually function. The market isn't abandoning AI — it's repricing which layer of the stack deserves the premium.
I've spent 14 years tracking these rotations, and I can tell you this pattern is textbook. The question isn't whether AI is a bubble. It's whether you're positioned for the second act.
Context
Let me break down what's actually happening on the tape. Over the past three months, Goldman's momentum factors have undergone a quiet but seismic shift. Software has displaced semiconductors as the largest weight in the three-month momentum long portfolio. Meanwhile, semiconductors and the AI complex have moved into the short basket.
Read that again. The same funds that were chasing Nvidia and its peers to all-time highs are now actively shorting that exposure while piling into software names. That's not a market that's lost faith in AI. That's a market that's decided the easy money in chip stocks has been made.
The report points to a specific catalyst for this shift: the violent deleveraging we just witnessed. When AI hedge fund portfolios drop 10% in five days, that's not fundamentals breaking. That's positioning being forced. Leveraged longs getting liquidated. Risk desks cutting exposure before the weekend. This is mechanical, not fundamental.
But here's the part that most retail investors miss. Goldman explicitly states the AI trade is not over. They're recommending storage and data centers as the most attractive tactical opportunity, citing the most significant valuation gap in the market. The logic is straightforward: these companies' profit recovery hasn't been fully priced into their stock prices yet.
Core
Now let me get into the data that matters. Based on my experience auditing market structure during the 2020 DeFi liquidity panic, I've learned that when institutional flows rotate, they don't do it quietly. They do it with conviction. And the signals here are unmistakable.
First, the momentum shift. Software replacing semiconductors as the top momentum long is not a minor rebalancing. It's a statement. The market is saying that AI value creation is moving up the stack — from the chips that train models to the applications that monetize them. This aligns with what I'm seeing in on-chain data and corporate earnings: the infrastructure buildout is largely complete at the chip level, but the software layer is just beginning to generate meaningful revenue.
Second, the specific recommendation of storage and data centers. Goldman identifies this as the sector with the most significant valuation gap — profit recovery not yet reflected in stock prices. This is a classic "value + growth catalyst" setup. The companies in this space — think Dell, Super Micro, Micron — have been beaten down because investors assumed their AI exposure was cyclical. But the data suggests otherwise. AI data center buildout requires massive amounts of storage, both for training data and inference workloads. HBM (high-bandwidth memory) demand alone is projected to grow exponentially over the next two years.
Third, the capital flow diversion. The report notes that money is flowing into previously ignored sectors — European and Japanese banks, gold miners, copper stocks. This is the most misunderstood signal in the entire report. Most retail investors will read this as "AI is dead, money is rotating to safety." That's wrong. What this actually represents is the AI trade maturing. Copper, specifically, is an AI trade — data centers require enormous amounts of copper for power infrastructure and networking. The market is discovering that AI is not just a semiconductor story. It's an everything story.
Fourth, the catalyst timeline. Goldman flags Nvidia's Q2 earnings and the September industry conferences as the key events to watch. This is critical because it tells us the market is waiting for confirmation. If Nvidia delivers strong guidance — which I believe they will, based on supply chain data — the AI complex could re-rate higher. If they disappoint, we could see a second round of deleveraging.
Contrarian
Here's where I diverge from the consensus reading of this report. Most analysts are framing this as "sell semiconductors, buy storage." That's a surface-level interpretation that misses the deeper signal.
The real story is that the AI trade is transitioning from a beta trade to an alpha trade.
During the first phase — which we saw through 2023 and early 2024 — you could buy any AI-adjacent stock and make money. The rising tide lifted all boats. That phase is over. The Goldman report is essentially confirming that we've entered phase two, where stock selection matters more than sector allocation.
But here's the contrarian angle that I haven't seen anyone discuss: the money flowing into European and Japanese banks, gold miners, and copper stocks isn't just a hedge against AI volatility. It's a bet on the real-world infrastructure that AI requires. Data centers need power, and power needs copper. AI-driven automation in banking is improving margins at traditional financial institutions. The "non-AI" sectors Goldman mentions are actually AI plays in disguise — they just haven't been labeled that way yet.
The second blind spot is the assumption that storage and data center profits will recover. I've been tracking capital expenditure trends across the hyperscalers, and the data is mixed. Microsoft, Google, and Amazon have all signaled continued AI infrastructure spending, but the pace is slowing. If that spending decelerates faster than expected, the storage thesis breaks down. The profit recovery Goldman is betting on could be delayed by a quarter or two, which would test investor patience.
Third, and this is the one that worries me most: the leverage hasn't fully cleared. The five-day deleveraging we saw was violent, but it may not have been complete. If Nvidia's earnings disappoint, we could see a second leg down that takes the storage and data center names with it, despite their attractive valuations. Timing matters in this market, and the difference between buying before and after the catalyst can be 20% or more.
Takeaway
The AI trade is not dead. It's just getting surgical. Goldman's report confirms what my own data has been telling me for weeks: the easy money in broad AI exposure is gone, but the next wave of returns will come from precision — picking the right subsector, the right companies, and the right entry points.
Watch Nvidia's earnings like a hawk. Watch the September conferences for any signals about AI capex sustainability. And most importantly, watch whether the storage and data center names start seeing upward EPS revisions over the next two quarters. If they do, the Goldman thesis is confirmed. If they don't, the entire AI complex could face a reckoning.
Panic is a luxury for those who didn't do their homework. The rest of us are watching the rotation. Position accordingly.
