Tracing the alpha from the mint to the melt: Goldman's momentum data reveals a sector rotation that crypto traders should be watching closely — because the same de-leveraging playbook is now running in AI equities, and the spillover into digital assets could be brutal.
The numbers hit like a flash crash. Over five trading days, Goldman Sachs' AI hedge basket dropped 10%. The high-beta momentum basket? Down 12%. For anyone who lived through May 2022 — watching leveraged longs get liquidated in cascading waves — the pattern is sickeningly familiar. But here's the twist: Goldman isn't calling the top. They're calling it a rotation.
The firm's latest note, which crossed my desk this morning, is a masterclass in deconstructing the terraformed logic of collapse — except this time, the terraforming happened in equities, not in algorithmic stablecoins. The headline: "AI trade is not over." The subtext: the easy money phase is dead, and the market is now entering the differentiation phase where fundamentals matter more than narrative.
Let me break down what Goldman is actually seeing — and why this matters for crypto, especially for AI-focused tokens and DePIN narratives that have been riding the coattails of the AI equity rally.
The Context: What Goldman Actually Found
Goldman's quant team has been tracking momentum factors across US equities, and the signal is unambiguous. Software has overtaken semiconductors as the largest weight in the three-month momentum long portfolio. Semiconductors and the "AI complex" — their term for the cluster of AI-exposed stocks — have shifted into the short portfolio.
This is not a small shift. When Goldman's momentum factors flip this decisively, it means systematic funds — the algos that trade on these signals — are now actively shorting the very names that led the market higher for 18 months. Think about that: the same institutional flow that pushed Nvidia to a $3 trillion valuation is now mechanically shorting it.
The sector rotation is equally telling. Goldman is explicitly recommending storage and data center names — Dell, Super Micro, Micron — arguing that the "valuation gap is the most significant" and that "profit recovery has not yet been fully reflected in stock prices." This is the classic value-plus-catalyst play: buy infrastructure before the earnings catch up to the narrative.
And where's the money going? Into European and Japanese banks, gold miners, and copper stocks. The AI trade's capital is being recycled into traditional sectors that have been ignored for years.
The Core: De-leveraging Without the Panic
Here's the part that should make crypto traders sit up. Goldman's data shows the AI equity complex went through a violent de-leveraging — but the firm explicitly states this is a healthy correction, not a structural breakdown. They're framing it as profit-taking after an extreme run, not as a repudiation of the AI thesis.
The key catalysts they're watching: Nvidia's Q2 earnings (due late August) and the September industry conferences. These events will determine whether the de-leveraging continues or whether the momentum factors flip back.
Now, let me translate this into crypto terms, because the parallels are uncomfortable.
Remember the NFT minting frenzy of 2021? I spent three weeks analyzing on-chain wallet clustering for BAYC's 15,000 mints and found that 30% of the initial supply was controlled by five interconnected entities. The narrative was "community ownership." The reality was concentration. When I published "The Illusion of Decentralization in PFPs," it went viral — not because I was right, but because I was early.
This is the same dynamic. The AI trade's "community" — retail and institutional alike — has been buying a narrative of infinite growth. Goldman's momentum data is the on-chain analysis of equities: it reveals the concentration, the leverage, and the fragility beneath the surface.
The storage and data center recommendation is the equivalent of finding the infrastructure plays that haven't been priced for the AI buildout. These companies — Dell, SMCI, Micron — are the picks-and-shovels of the AI gold rush. Their earnings haven't caught up to their stock prices, which means there's a lag between narrative and fundamentals.
But here's the trap: the same logic that makes storage attractive makes the broader AI complex vulnerable. If Nvidia's earnings disappoint — if the company guides lower on data center revenue — the entire infrastructure trade gets repriced downward. Storage and data centers are not immune; they're just lagging.
The Contrarian Angle: The Blind Spot in Goldman's Thesis
From viral mint to structural reality — the AI trade's leverage hangover is worse than the headline numbers suggest.
Goldman's note acknowledges the de-leveraging but doesn't quantify the residual leverage. From my experience auditing on-chain data during the LUNA collapse, I know that the first wave of liquidation is rarely the last. The Anchor Protocol withdrawal rates I tracked in May 2022 showed a slow bleed before the final capitulation. The same pattern applies here.
The momentum factor data is a lagging indicator. It reflects the past three months of flows. It doesn't predict what happens when Nvidia reports. If the earnings call reveals softening demand — or worse, inventory buildup in the AI supply chain — the second wave of de-leveraging could be more violent than the first.
And here's the part Goldman doesn't say: the rotation into banks, gold, and copper isn't just a value play. It's a hedge. Smart money is positioning for a world where the AI trade doesn't deliver on its promises — or where the broader market corrects and these defensive sectors outperform.
The copper mention is particularly telling. AI data centers consume enormous amounts of power and require massive copper infrastructure. If investors are buying copper stocks, they're betting on the physical buildout of AI infrastructure — not just the digital layer. This is the same logic that drives DePIN narratives in crypto: the physical infrastructure of decentralized networks.
The regulatory whisper here is deafening. As AI equities de-leverage, the narrative shifts to "real value" — profits, cash flows, earnings. This is the same shift that happens in crypto during bear markets. The "utility" narrative replaces the "revolution" narrative. Projects that can demonstrate actual revenue and usage survive; everything else gets repriced to zero.
The Takeaway: What to Watch Next
Speed is the only moat in noise — but precision is the moat in de-leveraging cycles.
The AI trade's rotation into storage and data centers is a signal, not a destination. It tells us the market is hunting for the next leg of the AI narrative — the infrastructure layer that hasn't been fully priced. But it also tells us the leverage hasn't been fully flushed.
For crypto traders, the implications are twofold. First, AI-related tokens — whether they're GPU marketplaces, data storage networks, or compute protocols — will face increasing scrutiny on fundamentals. The narrative-driven pumps are over. Second, the rotation into "real assets" (banks, gold, copper) mirrors the flight to stablecoins and BTC during crypto market stress. The flight to quality is real.
My watchlist for the next 30 days: Nvidia's Q2 earnings (the single most important catalyst for both AI equities and AI tokens), the September industry conferences (for new demand signals), and the momentum factor data from Goldman's next weekly update.
The question I'm asking myself: if the AI equity complex de-leverages further, does the AI token complex follow — or does the decentralized nature of crypto infrastructure offer a hedge that centralized equities can't?
That's the trade I'm watching. The AI trade isn't over — but it's entering its most dangerous phase, where the difference between narrative and fundamentals becomes a chasm. And in that chasm, only the prepared survive.