The high-beta momentum portfolio just lost 12% in a single week. The AI hedge portfolio dropped 10% in five days. Leverage is coming out of the AI trade at a speed that should concern anyone still holding a passive index fund and calling it an AI strategy.
Goldman Sachs is telling you something important: the AI trade is not over, but the way you make money from it has fundamentally changed. The era of buying the whole sector and watching it rise is finished. What replaces it is a market that demands verification, differentiation, and a willingness to look where the crowd isn't.
This is not a call to abandon AI. It is a call to abandon laziness.
The Context: A Market Structure Shift
For the past eighteen months, the AI trade has been a beta trade. Buy the semiconductor names, buy the cloud providers, buy anything with an AI narrative, and the rising tide lifted all boats. The logic was simple: AI infrastructure buildout was in its early innings, and every dollar of capex from the hyperscalers was a tailwind for the entire chain.
That phase is over. The deleveraging we are witnessing is not a crash. It is a rebalancing. The froth is being skimmed off, and what remains is a market that is starting to differentiate between companies that generate AI revenue and companies that merely talk about AI revenue.
Goldman's specific recommendations reveal the new playbook. Storage and data centers are now listed as the most tactically attractive sectors. The rationale is direct: profit recovery has not yet been fully reflected in stock prices. This is a classic value signal hiding inside a growth narrative. The market has been so focused on the glamour of chip design that it has ignored the unglamorous but essential infrastructure that makes AI actually work.
The Core: Reading the Order Flow
The most telling signal in Goldman's analysis is not what they recommend buying. It is what they recommend selling. Semiconductors and the broader AI complex have moved into the short portfolio. Meanwhile, software has replaced semiconductors as the largest weight in the three-month momentum long portfolio.
Let that sink in. The hardware that powered the first wave of AI is now a crowded short. The software that will monetize that hardware is becoming the momentum leader. This is a rotation, and it is happening in real-time in the quant factors.
From my experience building arbitrage systems in the 2020 DeFi summer, I learned that momentum shifts like this are rarely noise. When the factor models start rotating, they tend to overshoot. The semiconductor short could get more crowded before it reverses. The software momentum could extend further than fundamentals justify. The key is to understand what is driving the shift, not just to follow it blindly.
The storage and data center call is the most interesting. Goldman identifies a clear valuation gap: profits are recovering, but prices have not caught up. This is the kind of structural inefficiency that I built my 2020 arbitrage bot to exploit. The market is slow to reprice companies that are not in the spotlight. Storage is not glamorous. Data centers are not exciting. But they are where the AI workload actually lives.
Consider the technical reality. AI inference requires massive amounts of memory bandwidth. Model weights, KV caches, and training data all need to be stored and accessed at high speed. The demand for HBM and enterprise SSDs is not a narrative. It is a physical requirement of the technology. The companies that supply these components are seeing real revenue growth, and Goldman is correct that the market has not fully priced this in.
The Contrarian Angle: The Crowd Is Wrong About the Bubble
The mainstream narrative is that AI is a bubble and that the selloff is the beginning of the end. This is lazy thinking. What we are seeing is not a bubble bursting. It is a bubble deflating in specific areas while other areas are just beginning to inflate.
The semiconductor trade was crowded. The valuation of some chip names had detached from any reasonable near-term earnings scenario. The correction in those names is healthy. It is the market doing its job.
But the AI trade as a whole is not over. The shift from training to inference is just beginning. Training was a one-time cost. Inference is a recurring cost. Every time a user interacts with an AI application, it requires compute, storage, and bandwidth. This is a recurring revenue stream for the infrastructure layer, and it is only going to grow.
The real risk is not that AI is a bubble. The real risk is that you are positioned in the wrong part of the trade. If you are long the semiconductor names that are now in the short portfolio, you are fighting the trend. If you are ignoring the storage and data center names that Goldman is highlighting, you are missing the rotation.
There is also a hidden signal in the capital rotation to European and Japanese banks, gold miners, and copper stocks. This is not a rejection of AI. It is a search for value. The AI trade has become crowded, and smart money is looking for opportunities where the risk-reward is better. Copper is particularly interesting because it is a direct play on AI infrastructure. Data centers require massive amounts of copper for power distribution and cooling. The demand for copper from AI is a real, quantifiable factor that is not fully priced in.
The Takeaway: Position for the Rotation, Not the Narrative
The AI trade is entering a new phase. The beta era is over. The alpha era has begun. This means you cannot just buy the index and hope. You need to be selective, and you need to be willing to look where the crowd is not.
Storage and data centers are the clear opportunity. The profit recovery is real, and the market has not fully priced it in. The upcoming Nvidia earnings and the September industry conferences will be catalysts, but you should not wait for them. The time to position is now, before the crowd rotates.
Watch the momentum factors. If software continues to outperform semiconductors, the rotation is confirmed. If the semiconductor short gets more crowded, the deleveraging is not over. Use these signals to adjust your positioning.
Discipline turns noise into a tradable signal. The noise is the daily price action. The signal is the structural rotation that Goldman has identified. Follow the signal, not the noise.
Volatility exposes the weak foundations first. The weak foundation was the crowded semiconductor trade. The strong foundation is the infrastructure that supports AI inference. That is where the alpha is hiding.
Alpha hides in the friction between chains. In this case, the friction is between the AI narrative and the physical reality of data storage and processing. The market is starting to recognize this friction, and the re-pricing has begun. Get positioned before it completes.
Conviction without verification is just gambling. Verify the profit recovery in storage and data centers. Check the earnings reports. Confirm the demand for HBM and enterprise SSDs. Then act with conviction.
Ledgers don't lie. The profit recovery in storage and data centers is on the ledger. The market just hasn't fully read it yet.