The ledger remembers every trembling hand. Over the past five trading days, a carefully constructed AI hedge basket lost 10%. A high-beta momentum portfolio shed 12%. Logic chains break where greed connects, and right now, the greed that powered a 200% run in AI-linked equities is being replaced by something far more dangerous: doubt.
This is not the end of the AI trade. It is the end of the easy AI trade. And for anyone still holding a broad-based AI ETF, Goldman Sachs' latest analysis should read less like a market summary and more like a tactical warning: the index-level sugar rush is over, but the infrastructure hangover has just begun.
Context: The Great Unwinding
The story of 2025 was simple: buy AI, beat the market. The story of 2026 is more complicated. Goldman's data shows a decisive shift in momentum factors — the algorithmic scorecards that track which sectors are accelerating and which are decelerating. Software has overtaken semiconductors as the heaviest-weighted sector in the three-month momentum long portfolio. Simultaneously, semiconductors and the broader "AI complex" have slipped into the short portfolio.
For those of us who lived through the 2021 rotation out of pandemic winners, this pattern is hauntingly familiar. The first phase of a disruptive technology trade is a liquidity-driven phase — everyone piles in, valuations detach from fundamentals, and the market rewards narrative over nuance. The second phase is a differentiation phase, where the market begins to separate the builders from the pretenders. That's where we are now. The "AI trade" is not dead; it's just being repriced from a sector-wide phenomenon into a sub-sector-specific one.
Core: Where the Money Is Actually Moving
Let's dig into the specifics, because that's where the real signal lives. Goldman's tactical recommendation focuses on storage and data centers — a sub-sector where the profit recovery is real but not yet fully reflected in equity prices. The valuation gap between where these companies should be trading (given their earnings trajectory) and where they currently trade is the most significant across the entire AI complex.
I've spent the last eighteen months running my own audits on AI infrastructure names, and I can tell you this: the gap is real. The market is still pricing these companies as if they are merely providers of commodity hardware, not as the foundational layer of AI compute expansion. But the data tells a different story. Memory pricing has been on a steady upward trajectory, and data center utilization rates remain at record highs. The hyper-scalers are not slowing down their capital expenditure commitments — they are simply being more selective about what they buy. That's a tailwind for the storage and data center segment, not a headwind.
The second interesting movement is the flow of funds into non-AI sectors like European and Japanese banks, gold miners, and copper stocks. On the surface, this looks like a panic retreat from technology. But the silence is the only honest metadata here. This is not a retreat from AI — it's a search for cross-correlation. Copper is an AI trade dressed in mining clothing. Every GPU server rack, every data center power infrastructure, every advanced chip packaging facility requires a massive amount of copper. The fact that copper miners are being bought while semiconductors are being sold tells me that investors are not abandoning the AI thesis; they are playing the AI thesis through upstream resources rather than downstream assembly. It's a smarter, more resilient playbook.
Contrarian: The Blind Spot in the "Fundamental" Pivot
Here's the counter-intuitive angle that most retail investors will miss: the pivot to "fundamentals" is itself a momentum signal. When Goldman says the "EPS and share price divergence" is the key metric, they are essentially admitting that the market has moved from a narrative-driven phase to a cash-flow-driven phase. This sounds healthy, but it is not necessarily a safe signal.
Because here's the problem — EPS in the storage and data center sector is not yet realized. It is projected. And projections are nothing more than a market consensus about future AI spending. If the next round of earnings from the major chip players misses expectations, the entire supply chain re-prices. Storage is not as resilient as the recommendation suggests; it is a leveraged bet on the same AI capital expenditure cycle, just one layer removed. The market believes it has found "diversification" by moving from semiconductors to storage, but in reality, both are the same trade: a long-term bet on the same AI CapEx cycle, with different beta. When the cycle turns, they turn together. The only question is who falls faster.
My own forensic analysis of the Q1 AI infrastructure orders reveals a disturbing pattern: the order pipeline is concentrated in three or four hyperscalers. If any one of them were to pull back their capex guidance, the entire sector's "projected EPS" collapses. The market is currently paying a premium for a piece of paper that says "profits will recover" — but there is no assurance on that piece of paper. It's just a promise. The ledge remembers every trembling hand, and this is a ledger built on a promise.
Conclusion: The Pivot Point Is on the Calendar
Forget the daily noise. The next real catalyst is the NVIDIA Q2 earnings report on August 28th, followed by the September industry conference. These two events are the fulcrum. They will not just determine NVIDIA's stock price — they will determine whether the storage/data center thesis has legs or is merely a one-quarter flash in the pan.
My positioning thesis is simple: the broad AI trade is done; the infrastructure trade is beginning. But be careful not to confuse "infrastructure" with "safe." The market is in a de-leveraging phase, and those who are over-leveraged on the old narrative will be shaken out. The traders who survive this chop will be those who recognize that the momentum shift is not a signal to exit, but a signal to re-enter with a more surgical weapon.
Speed wins the trade, clarity wins the war. The next move is to watch the earnings, ignore the daily noise, and position for the sub-sector rotation. The trade is not over — it has just learned to walk again. The question is, will you be ready to follow it?
