Goldman's AI Trade: The Momentum Signal Flashing in the Data Center

The AI hedge basket dropped 10% in five days. The high-beta momentum composite fell 12%. These are not crash numbers from a black swan event. They are the fingerprints of a coordinated de-leveraging event, written in the language of margin calls and forced selling. Yet Goldman Sachs says the AI trade is not over. That's a statement that demands forensic verification, not blind acceptance. I've spent the last decade dissecting protocol failures and market dislocations. From the Parity wallet vulnerability in 2017 to the Terra-Luna oracle race condition in 2022, the pattern is always the same. When the crowd is positioned uniformly in one direction, the unwind is violent. The question is whether this is a healthy reset or the first block in a cascading failure. The Goldman note suggests the former. But their own data reveals a more nuanced story. The momentum factor is rebalancing. Software has replaced semiconductors as the largest weight in the three-month momentum long portfolio. Semiconductors and the AI complex have moved into the short portfolio. This is not a rotation. This is a structural shift in how the market is pricing AI exposure. Let's break down the mechanics. The de-leveraging event was sharp but contained. A 10% drawdown in a concentrated basket over five days is significant, but it's not the 50% collapse we saw in crypto lending protocols during the 2022 contagion. The question is what happens next. Goldman's answer is that the AI trade transitions from a beta story to an alpha story. The era of buying the whole sector for outsized returns is over. The market is now demanding earnings visibility. This is where the analysis gets interesting from a technical perspective. Goldman recommends storage and data center stocks. The rationale is straightforward: the valuation gap is the most significant, and profit recovery has not yet been fully reflected in share prices. This is classic value investing applied to AI infrastructure. But the deeper signal is about where the actual economic value is accruing in the AI stack. For the past two years, the narrative has been dominated by GPU scarcity. Nvidia was the pick-and-shovel play. But the market is now recognizing that the real bottleneck is shifting. AI training demands massive compute, but AI inference — the ongoing operational cost of running models — is where the recurring revenue lies. This requires storage, networking, and data center capacity. The chips are the ignition; the infrastructure is the engine that keeps running. I've been tracking this shift through on-chain data and infrastructure metrics. The correlation between GPU orders and data center construction starts is breaking down. Cloud providers are signaling that their capital expenditure is moving from pure compute procurement to a more balanced allocation across storage, networking, and power infrastructure. This is not a short-term trend. This is a fundamental re-architecture of how AI systems are deployed. The momentum data from Goldman confirms this. The three-month momentum long portfolio now weights software highest. This is the market saying that the application layer has more near-term earnings certainty than the hardware layer. Software companies have recurring revenue models. They have existing customer bases. They can monetize AI features through price increases and new product tiers. Hardware companies face cyclical demand, inventory risk, and geopolitical headwinds from export controls. But here's the contrarian angle that most analysts are missing. The flow of funds into European and Japanese banks, gold miners, and copper stocks is not just a defensive rotation. It's a signal about AI's physical footprint. AI data centers consume enormous amounts of electricity. They require cooling infrastructure. They need copper for wiring and power distribution. The gold and copper miners are not just a safe haven play; they are a bet on the physical build-out of AI infrastructure. I've seen this pattern before in the crypto mining industry. During the 2020-2021 bull run, the focus was on ASIC chips. But the real bottleneck became energy infrastructure. Mining operations that secured cheap power contracts outperformed those with the latest hardware. The same logic applies to AI. The constraint is not just compute; it's the physical resources required to run and cool that compute. This brings us to the critical catalyst: Nvidia's Q2 earnings and the September industry conferences. Goldman identifies these as the key events that will determine the direction of the AI trade. The market is positioned for a potential disappointment. The de-leveraging event suggests that hedge funds are reducing risk ahead of the announcement. This is rational behavior. The risk-reward is asymmetric. If earnings beat, the rally resumes. If they miss, the de-leveraging accelerates. But here's what the sell-side reports won't tell you. The earnings guidance is more important than the headline numbers. The market has already priced in a strong quarter. What it hasn't priced in is the forward guidance on capital expenditure. If Nvidia signals that its customers are pausing or delaying orders, the entire AI complex reprices. If they confirm accelerating demand, the current valuation gap in storage and data centers closes quickly. The storage thesis is particularly compelling from a technical analysis perspective. The market is treating storage as a commodity business, but AI workloads have fundamentally changed the storage requirements. High-bandwidth memory (HBM) is a critical component for AI accelerators. Solid-state drives (SSDs) are replacing HDDs in data centers for AI training datasets. This is not the same storage cycle we've seen before. The demand profile is structurally different. Let me give you a concrete example from my own experience. When I was auditing smart contract protocols, I noticed that the most common failure point was not the core logic, but the oracle mechanisms. The price feeds were the weakest link. The same principle applies to AI infrastructure. The compute is the core logic, but the data pipeline is the oracle. If the storage and data transfer layers fail, the entire system underperforms. This is why the storage and data center thesis has fundamental backing, not just momentum. Goldman's analysis is solid, but it carries the typical sell-side biases. The report is designed to generate trading ideas, not to provide a complete risk assessment. The information selectivity bias is moderate. They highlight the favorable factors — valuation gap, profit recovery — while deemphasizing the risks, such as a potential slowdown in AI capital expenditure. The interest alignment bias is high. Goldman's clients include companies in the storage and data center sectors. Their recommendations may not be entirely disinterested. But the data itself is reliable. The momentum factor analysis is based on observable market behavior. The fund flow data is verifiable. The valuation gaps are quantifiable. The logic is sound. The execution is where the risk lies. My assessment is that the AI trade is transitioning from a phase of collective beta to a phase of selective alpha. The de-leveraging event was a necessary correction. The market was too crowded. The positioning was too uniform. The unwind was healthy. But the next phase requires more sophisticated analysis. You cannot just buy the sector. You need to identify the specific companies where profit recovery is not yet priced in. The storage and data center thesis is the clearest expression of this new phase. The profit recovery is not a hope; it's a function of AI workload growth. The data is there. The demand is there. The question is whether the market has fully priced it in. Goldman says no. My analysis suggests they're right, but with a caveat. Here's the caveat. The storage and data center trade is not immune to the Nvidia earnings risk. If Nvidia guides down, the entire AI complex sells off, including storage and data centers. The correlation during de-leveraging events approaches one. You cannot hide from the beta when the market is in risk-off mode. The opportunity is to position before the catalyst, not after. Building on chaos, then locking the door. This is the approach that has served me well through multiple market cycles. You identify the dislocation, you analyze the fundamentals, and you position before the crowd arrives. The crowd is currently focused on Nvidia's earnings. The smart money is looking at what happens after. The storage and data center thesis is the post-earnings trade. The other signal worth monitoring is the flow into copper and mining stocks. This is not just a defensive rotation. This is a bet on the physical infrastructure build-out. AI data centers consume massive amounts of copper for power distribution and cooling. The market is starting to price this in. The copper trade is a longer-duration play on AI infrastructure, with a 6-12 month time horizon. Silicon ghosts in the machine, verified. The AI trade is not over, but it has changed. The market is no longer paying for potential. It's paying for proof. The proof is in the earnings. The proof is in the profit recovery. The proof is in the data center utilization rates. The companies that can demonstrate real revenue from AI infrastructure will outperform. The companies that are still selling potential will underperform. Let me be clear about the risks. The de-leveraging event could resume. If Nvidia's earnings disappoint, the AI complex faces a second wave of selling. The storage and data center thesis would be temporarily disrupted. The profit recovery could be delayed. The valuation gap could widen before it narrows. These are real risks. They are not hypothetical. But the fundamental thesis remains intact. AI is not a bubble. It's a technology cycle. The build-out is real. The demand is real. The revenue is coming. The question is timing. The market is in a consolidation phase. The chop is for positioning. The data points to storage and data centers as the next leg of the trade. Static analysis reveals what intuition ignores. The momentum data, the fund flows, and the valuation gaps all point in the same direction. The market is repricing AI from a growth story to a value story. The companies with real earnings visibility will be rewarded. The companies with only promises will be punished. This is the new reality of the AI trade. The takeaway is forward-looking, not a summary. The Nvidia earnings report on August 28 is the immediate catalyst. The September industry conferences will provide additional color. But the real signal is the structural shift in momentum. Software over semiconductors. Storage over compute. Infrastructure over hype. This is the trade for the next 3-6 months. Logic is the only law that doesn't lie. The data is clear. The market is rotating. The AI trade is not over, but it has evolved. The winners will be those who understand the new mechanics. The losers will be those still trading the old narrative. Build your positions accordingly. Verify everything. Trust nothing. The market will tell you when you're right. Proving existence without revealing the source. The source is the data. The data is in the momentum factors, the fund flows, and the earnings reports. The proof is in the profit recovery. The market is moving. The question is whether you're moving with it or against it. The data says storage and data centers. The data says software over semiconductors. The data says the AI trade is entering a new phase. The choice is yours. I've made mine. Composability is just controlled anarchy. The AI market is no different. The components are recombining. The capital is flowing to new destinations. The anarchy is controlled by earnings visibility. The companies with real revenue will thrive. The companies with only narratives will fade. This is the natural order of markets. This is the new phase of the AI trade. Position accordingly.