Everyone's been staring at the same chart: AI stocks ripping higher, portfolio managers patting themselves on the back, and the narrative machine running at full capacity. Then the last five trading days happened, and suddenly the AI hedge basket is down 10%, the high-beta momentum complex is off 12%, and the same people who were screaming about a structural paradigm shift are now quietly checking their stop-losses.

Goldman Sachs just published a note telling you not to panic. The AI trade, they insist, is not over. But here's what caught my attention: the same note quietly admits that the era of "buy the whole sector and collect alpha" is dead. The momentum factor is rebalancing. Software has replaced semiconductors as the largest weight in the three-month momentum long basket. Semiconductors and the AI complex have moved into the short basket.

Let me translate that from sell-side speak into something operational: the smart money is rotating out of the picks-and-shovels narrative and into something with actual earnings visibility. And Goldman is telling you to follow them into storage and data centers — the boring stuff that makes AI actually work — because the profit recovery there hasn't been priced in yet.

Code is law, but bugs are justice. And the bug in the current AI trade is that everyone priced in the dream without checking whether the underlying infrastructure companies were actually making money.

The Context: What Goldman Is Actually Looking At

Let me set the scene properly. This isn't a random Tuesday. We're in a specific window — late August, with Nvidia's Q2 earnings looming and a cluster of industry conferences scheduled for September. The market is in a bull phase, but the internal dynamics have shifted from "everything goes up" to "everything is a trade."

Goldman's note, which I've been dissecting since it crossed my desk, is built on three pillars:

First, the momentum factor is rebalancing. This is the quantitative backbone of their argument. The three-month momentum long basket — the thing that's been printing money for anyone who just bought whatever was going up — has undergone a structural shift. Software is now the largest weight. Semiconductors and the AI complex have been pushed into the short basket. This isn't a minor adjustment; it's a regime change in how systematic money is positioned.

Second, there's been a violent deleveraging event. The AI hedge basket dropped 10% in five days. The high-beta momentum complex fell 12%. These aren't normal pullbacks — they're forced selling events. When you see moves like that, you're watching leveraged positions get unwound, margin calls getting made, and risk parity or vol-targeting strategies mechanically reducing exposure. The kind of move that happens when everyone is on the same side of the boat and someone yells "shark."

Third, and this is the part most people will skim past: Goldman is recommending storage and data centers. Their logic is straightforward — the valuation gap is the most significant, and the profit recovery hasn't been fully reflected in stock prices. In plain English: these companies are about to start making real money from AI infrastructure spending, and the market hasn't caught up to that reality yet.

Now, I've been in this game long enough to know that when a major bank publishes a note like this, there are layers beneath the surface. Let me peel them back.

The Core: Order Flow, Momentum, and the Mechanics of the Rotation

Let's get into the weeds, because that's where the actual information lives.

The Momentum Factor Rebalancing

The momentum factor is one of the most reliable signals in quantitative finance. It's not predictive — it's descriptive. It tells you where money has been flowing, not where it's going. But when the factor itself undergoes a structural rebalancing, that's a signal worth respecting.

Here's what the data shows: software has replaced semiconductors as the largest weight in the three-month momentum long basket. Semiconductors and the AI complex have moved into the short basket. This is a massive shift in systematic positioning.

Think about what this means mechanically. There are billions of dollars in momentum strategies — both retail and institutional — that mechanically buy whatever's in the long basket and short whatever's in the short basket. When the factor rebalances, those strategies are forced to execute trades. They're selling semiconductors and buying software. They're adding shorts to the AI complex and covering their longs.

This isn't a discretionary call. It's a mechanical response to price action over the trailing three months. And it creates a self-reinforcing loop: the selling pushes prices down, which extends the momentum signal, which triggers more selling.

The Greeks don't capture this kind of risk. You can model delta and gamma all day, but the systemic risk here is the crowding — the fact that everyone's in the same trade, and the exit door is narrower than the entry.

The Deleveraging Event

The numbers are stark: AI hedge basket down 10% in five days, high-beta momentum complex down 12%. These are not organic pullbacks. They're forced unwinds.

When you see moves of this magnitude in a short window, you're looking at a few things happening simultaneously:

Margin calls. Leveraged positions get liquidated when they breach maintenance requirements. The selling feeds on itself — liquidation pushes price down, which triggers more liquidations.

Volatility targeting. Risk-parity and vol-targeting strategies mechanically reduce exposure when realized volatility spikes. They don't care about fundamentals. They just see risk and cut.

Dealer hedging. When the market drops, options dealers who are short gamma need to sell more to hedge their positions. This amplifies the downside.

The key insight here is that this deleveraging is happening within the AI complex, not across the entire market. Money isn't leaving equities — it's leaving AI and going into other sectors. That's a rotation, not a risk-off event.

The Storage and Data Center Recommendation

This is where Goldman's analysis gets interesting. They're recommending storage and data centers because the "valuation gap is the most significant" and "profit recovery hasn't been fully reflected in stock prices."

Let me translate: these companies are about to see their earnings inflect upward, and the market hasn't priced it in yet.

The logic chain is straightforward. AI training and inference require massive compute infrastructure. That infrastructure requires storage — both high-bandwidth memory (HBM) for the GPUs themselves and traditional storage for the data lakes that feed the models. It also requires data centers — physical facilities with power, cooling, and networking.

The capital expenditure cycle for AI infrastructure is still in its early innings. Hyperscalers are committing hundreds of billions to data center buildouts. But the revenue recognition for storage and data center companies lags the capex commitments. That lag creates the opportunity.

Here's the thing about this recommendation that most retail investors will miss: Goldman isn't telling you to buy the AI leaders. They're telling you to buy the picks and shovels — the companies that benefit from AI infrastructure spending regardless of which AI model wins.

This is a classic barbell strategy: you're not betting on a specific AI winner, you're betting on the aggregate capital expenditure that's already been committed.

The Capital Flow Diversion

The note also mentions that money is flowing into "previously overlooked areas" — European and Japanese banks, gold miners, and copper miners. This is worth unpacking because it reveals something about the market's internal logic.

European and Japanese banks: These are value plays. They're trading at reasonable valuations, they're benefiting from rising interest rates, and they're not exposed to the AI narrative. Money flowing here suggests investors are looking for cheap, stable returns while the AI complex digests its gains.

Gold miners: This is a hedge. When investors are uncertain about the sustainability of a rally, they buy gold. The fact that gold miners are attracting flows suggests some investors are positioning for potential market stress.

Copper miners: This is the most interesting one. Copper is the metal of electrification — it's used in everything from data center power infrastructure to EV charging networks. The fact that copper miners are attracting flows suggests investors are thinking about the physical infrastructure requirements of AI, not just the digital layer.

This is the kind of cross-sector deduction that I find most valuable. The AI trade isn't just about semiconductors and software. It's about power generation, cooling systems, networking equipment, and the raw materials that make all of it possible.

The Contrarian Angle: What Goldman Isn't Telling You

Now let me put on my skeptical hat, because there are several things about this analysis that don't sit right with me.

The Sell-Side Incentive Problem

Goldman Sachs is a sell-side institution. Their research is distributed to clients, and their trading desk executes on that research. There's an inherent conflict of interest: if Goldman's clients are positioned in storage and data centers, the bank has an incentive to talk up those sectors.

I'm not saying the analysis is wrong — the momentum data is real, the valuation gap is measurable, and the profit recovery thesis is plausible. But I am saying that you should understand the incentive structure before you act on any sell-side recommendation.

Based on my audit experience, I've learned to check the code before I trust the output. The same principle applies to sell-side research: verify the data, understand the incentives, and make your own judgment.

The "Profit Recovery" Assumption

Goldman's recommendation for storage and data centers rests on the assumption that profit recovery is coming. But what if it doesn't materialize?

The AI infrastructure buildout is real, but the revenue recognition is uncertain. Data center companies are signing long-term leases, but those leases are contingent on the lessees actually deploying the capacity. If AI adoption slows — if the models don't find enough real-world use cases, if the regulatory environment tightens, if the energy costs become prohibitive — the lease commitments could be renegotiated or abandoned.

Storage is similarly exposed. The demand for HBM and SSD is driven by AI training and inference, but that demand is concentrated in a handful of hyperscalers and AI labs. If those customers pull back, the storage companies have limited alternative markets.

The point is that "profit recovery" is a thesis, not a fact. It's a reasonable thesis, but it's not guaranteed.

The Momentum Factor Lag

Momentum is a lagging indicator. It tells you what has been happening, not what will happen. The fact that software has replaced semiconductors in the momentum long basket tells you that software has been outperforming — it doesn't tell you that software will continue to outperform.

This is a critical distinction. The momentum rebalancing could be the beginning of a sustained rotation, or it could be a mean-reversion setup where the laggards bounce back.

The key variable is Nvidia's earnings. If Nvidia delivers a blowout quarter and raises guidance, the semiconductor trade could reassert itself. The momentum factor would then be wrong — it would have rotated out of semiconductors just before they resumed their uptrend.

The Crowding Paradox

Here's the thing that keeps me up at night: the AI trade is crowded. Everyone's in it. The positioning is extreme. And when a trade is this crowded, the risk is asymmetric — the downside is much larger than the upside.

Goldman is essentially saying, "The AI trade is crowded, so let's find less crowded ways to play it." That's smart. But it also implies that the AI trade itself is vulnerable. If the AI leaders correct sharply, the storage and data center names will likely correct with them — they're correlated through the AI infrastructure theme.

So the recommendation to buy storage and data centers isn't a hedge against AI downside. It's a more selective way to express the same thesis. If the thesis fails, both positions lose.

The Technical Analysis: Where the Real Information Lives

Let me get into the specific technical details that most commentary on this note will miss.

The Momentum Factor Construction

Goldman's momentum factor is constructed using trailing returns over a specific window — typically 3-12 months, with the most recent month excluded to avoid short-term reversal effects. The fact that software has become the largest weight in the three-month momentum long basket tells us that software stocks have been the strongest performers over the trailing three months, excluding the most recent month.

This is significant because it means the rotation isn't just a recent phenomenon — it's been building for months. The market has been quietly rewarding software companies while punishing semiconductor names.

The question is whether this reflects a fundamental shift in the AI value chain or just a mean-reversion trade. My read is that it's a bit of both. Software companies are starting to monetize AI — they're embedding AI features into their products and seeing real revenue growth. Semiconductors, meanwhile, have been priced for perfection, and any disappointment triggers outsized selling.

The Deleveraging Mechanics

The 10% drop in the AI hedge basket and the 12% drop in the high-beta momentum complex over five days are textbook deleveraging events. Let me walk through the mechanics:

Step 1: Initial shock. Something triggers selling — maybe a disappointing earnings report, maybe a macro headline, maybe just profit-taking after a strong run.

Step 2: Margin calls. Leveraged positions breach maintenance requirements. Brokers issue margin calls. Investors either add capital or get liquidated.

Step 3: Forced selling. Liquidations push prices down further. The selling attracts more selling as other leveraged positions approach their own margin thresholds.

Step 4: Volatility spike. Realized volatility increases. Vol-targeting strategies mechanically reduce exposure. This amplifies the selling.

Step 5: Dealer hedging. Options dealers who are short gamma need to sell more to hedge their positions. This adds to the downward pressure.

The result is a cascade that has nothing to do with fundamentals. It's a mechanical unwind of crowded positioning.

The important question is whether the deleveraging is complete. My assessment is that it's probably not. The AI trade was extremely crowded, and the positioning is still elevated. If Nvidia's earnings disappoint, we could see another leg down.

The Storage and Data Center Valuation Gap

Goldman says the valuation gap for storage and data centers is "the most significant." Let me quantify what that means.

The storage and data center complex includes names like Dell, Super Micro Computer, and Micron. These companies have seen their stock prices rise with the AI narrative, but their earnings haven't caught up yet. The P/E ratios are elevated, but the earnings growth is expected to inflect upward as AI infrastructure spending translates into revenue.

The "valuation gap" is the difference between the current P/E and the forward P/E. If the forward P/E is significantly lower than the current P/E, it means the market is expecting earnings growth. The gap is "significant" when the expected earnings growth isn't fully reflected in the current price.

This is a classic value-plus-growth setup: you're buying companies with reasonable valuations that are about to see earnings acceleration. The risk is that the earnings acceleration doesn't materialize, and the valuation gap closes through price decline rather than earnings growth.

The Nvidia Catalyst

Goldman identifies Nvidia's Q2 earnings and the September industry conferences as the key catalysts. This is the right call — Nvidia is the bellwether for the entire AI complex.

Here's what I'll be watching in the earnings report:

Revenue guidance: The most important number. If Nvidia guides above consensus, the AI trade gets a new lease on life. If they guide in line or below, the selling could resume.

Data center revenue: This is the core of Nvidia's AI business. I want to see not just the number, but the commentary on demand drivers and supply constraints.

Customer concentration: How much of the revenue is coming from a handful of hyperscalers? If it's too concentrated, there's a risk of demand pull-forward.

Gross margins: Nvidia's gross margins are extraordinary — above 70%. Any compression would be a red flag, suggesting competitive pressure or pricing power erosion.

Inventory and supply chain: I want to see commentary on whether Nvidia can meet demand. Supply constraints are bullish in the short term but could create a demand vacuum in the future.

The September conferences are also important. These are where companies announce new products and roadmaps. If there's a major AI infrastructure announcement — a new data center design, a new storage architecture, a new chip — it could reignite the AI trade.

The Cross-Sector Deduction: AI and the Physical World

Here's where I'm going to go beyond what Goldman is saying and connect some dots that most people are missing.

The note mentions that money is flowing into copper miners. This is a signal that the market is starting to price in the physical requirements of AI infrastructure.

AI data centers are massive consumers of electricity and cooling. They require copper for power distribution, for networking, for the servers themselves. The buildout of AI infrastructure is going to require enormous quantities of copper, and the supply is constrained.

This is a classic cross-sector trade: AI isn't just a software story, it's a physical infrastructure story. The companies that provide the raw materials and components for AI infrastructure are going to benefit regardless of which AI models win.

The same logic applies to power generation. AI data centers need electricity — lots of it. The companies that provide power generation, transmission, and distribution are going to benefit from the AI buildout. This includes natural gas producers, nuclear power operators, and renewable energy companies.

The market is starting to price this in, but I think it's still early. The copper trade, in particular, has room to run as the AI infrastructure buildout accelerates.

The Institutional Volatility Synthesis

Let me bring this back to my area of expertise: options and volatility.

The current market structure is creating some interesting opportunities for options traders. The AI complex is experiencing elevated volatility, and the options market is pricing in significant moves around Nvidia's earnings.

Here's what I'm seeing:

Implied volatility is elevated. The options market is pricing in a large move for Nvidia and other AI names around earnings. This creates opportunities for volatility sellers — you can collect premium by selling options that are pricing in more movement than is likely to occur.

The skew is steep. Put options are more expensive than call options, reflecting the market's fear of downside. This creates opportunities for put spreads — you can buy downside protection at a reasonable cost by selling further-out-of-the-money puts.

The term structure is in contango. Longer-dated options are more expensive than shorter-dated options, reflecting uncertainty about the medium-term outlook. This creates opportunities for calendar spreads — you can sell short-dated options and buy longer-dated options to profit from the time decay differential.

The key insight is that the AI trade is now a volatility trade as much as a directional trade. The fundamentals matter, but the positioning and the flows matter just as much.

Greeks don't tell you the whole story. You need to understand the market structure — who's long, who's short, who's forced to buy or sell — to make informed decisions.

The Takeaway: What I'm Actually Doing

Let me cut through the analysis and give you something actionable.

The AI trade isn't over, but it's changed. The era of buying the whole sector and collecting alpha is dead. You need to be selective, and you need to be aware of the positioning and the flows.

Storage and data centers are the most interesting opportunity. The valuation gap is real, and the profit recovery thesis is reasonable. But you need to be selective — not all storage and data center companies are created equal.

Nvidia's earnings are the key catalyst. The market is going to react violently to the report, regardless of the actual numbers. If you're positioned in the AI complex, you need to be prepared for that volatility.

The physical infrastructure trade is underappreciated. Copper, power, cooling — these are the picks and shovels of the AI era, and the market is only starting to price them in.

The deleveraging isn't complete. The AI trade is still crowded, and the positioning is still elevated. If Nvidia disappoints, we could see another leg down.

Here's my specific playbook:

For the next two weeks: I'm positioning for volatility around Nvidia's earnings. I'm selling premium in names where the implied volatility is elevated relative to my assessment of the actual risk. I'm using put spreads to protect against downside in my core AI positions.

For the next three to six months: I'm building positions in storage and data center names that have reasonable valuations and clear earnings catalysts. I'm looking for companies with strong balance sheets, diversified customer bases, and exposure to the AI infrastructure buildout.

For the next six to twelve months: I'm adding exposure to the physical infrastructure trade — copper, power, and cooling. These are the companies that benefit from AI infrastructure spending regardless of which AI models win.

The market is always trying to tell you something. The question is whether you're listening. The momentum factor is telling you that the AI trade is rotating. The flow data is telling you that money is moving into overlooked sectors. The valuation gaps are telling you where the opportunities are.

The question isn't whether the AI trade is over. It's whether you're positioned for the next phase.

This analysis is based on my review of Goldman Sachs' recent note on AI trading markets, combined with my own experience auditing smart contracts during the 2017 ICO boom, running delta-neutral strategies through DeFi Summer, detecting NFT wash-trading patterns in 2021, hedging through the Terra/Luna collapse, and trading volatility around the 2024 ETF approvals. The market is a complex system, and the only way to navigate it is to understand the mechanics beneath the surface.

Code is law, but bugs are justice. The bug in the current AI trade is that everyone priced in the dream without checking whether the underlying infrastructure companies were actually making money. The fix is to find the companies where the profit recovery is real and the valuation hasn't caught up yet.

That's where the opportunity is. That's where the edge is. That's where the next phase of the AI trade is going to be won.