The high-beta momentum basket fell 12% in a week. The AI hedge portfolio dropped 10% in five days. Leverage is unwinding from extreme highs. This is not the sound of a bubble bursting. It's the sound of a market recalibrating its assumptions.
Goldman Sachs' latest note cuts through the noise with a clear thesis: the AI trade is not over, but the era of indiscriminate beta is. The days of buying any ticker with 'AI' in the name and watching it appreciate are gone. What remains is a more demanding, more selective market that wants to see revenue, not just roadmaps.
This is a structural shift, not a cyclical dip. The first phase of the AI trade, driven by liquidity and narrative, has concluded. We are now in the second phase, defined by fundamental differentiation. The market is no longer paying a uniform premium for 'AI vision.' It is demanding 'AI income.'
The most telling signal is the rotation within the AI complex itself. Semiconductors and AI conglomerates have been added to Goldman's short portfolio. Meanwhile, software has overtaken semiconductors as the largest weight in the three-month momentum long portfolio. This is a quant-level confirmation of a narrative shift: the market's attention is moving from the 'picks and shovels' of AI hardware to the 'gold miners' of AI applications.
This isn't just a stylistic preference. It's a judgment on where value is being captured. The semiconductor trade was built on scarcity and hype. The software trade is built on the promise of monetization. The market is betting that the latter is now more tangible.
Goldman identifies storage and data centers as the 'most tactically attractive' sectors, citing a 'profit recovery' that is not yet fully reflected in stock prices. This is a direct call on the AI infrastructure layer. The logic is simple: as AI models move from training to inference, the demand for data storage and compute capacity explodes. The companies providing this infrastructure are seeing real revenue growth, but the market hasn't fully priced it in.
I've seen this pattern before. In 2020, I tracked Uniswap V2 liquidity pools and found that large swap orders caused slippage exceeding 5%, leading to significant MEV extraction. The market was inefficient because it was focused on the narrative of 'DeFi' rather than the mechanics of the protocols. The same thing is happening here. The market is focused on the narrative of 'AI' rather than the economics of the infrastructure.
The contrarian angle here is that correlation is not causation. Goldman's recommendation of storage and data centers is based on a 'profit recovery.' But is this recovery driven by AI, or by other factors? Traditional enterprise IT spending is recovering. Cloud service providers are in a capex cycle. The AI-specific contribution to this 'profit recovery' is unclear. If the recovery is driven by non-AI factors, the investment thesis is weaker than it appears.
This is a critical blind spot. The market is treating 'storage' and 'data centers' as a monolith, but the drivers of their profitability are diverse. HBM demand is a different beast than enterprise SSD demand. Data center utilization is different from data center construction. The 'profit recovery' might be real, but it might not be an AI story.
The capital rotation out of AI into European and Japanese banks, gold miners, and copper stocks is another signal. This suggests that the marginal capital inflow into AI is slowing. The high-quality AI names are getting crowded, and money is seeking value elsewhere. This is not a sign of AI's demise, but it is a sign of maturity. The easy money has been made.
Nvidia's Q2 earnings, due at the end of August, are the next catalyst. The market will be looking for guidance on data center revenue and the sustainability of AI compute demand. A strong report could reignite the trade. A weak one could trigger another round of deleveraging. The market is on edge, and Nvidia holds the key.
Data doesn't lie. The momentum factor has shifted. The short portfolio has changed. The capital is rotating. These are all verifiable facts. The interpretation is where the risk lies. Goldman's call is a sophisticated one, but it's not without its own biases. As a sell-side firm, its recommendations can be influenced by its banking relationships. Nvidia is a major client. The 'AI trade is not over' narrative is a comfortable one for its partners.
I don't buy the doom-and-gloom narrative, but I also don't buy the 'buy the dip' reflex. The market is entering a phase where the data matters more than the story. The next few weeks will be defined by earnings, not tweets. The AI trade isn't dead. It's just getting harder. The crash wasn't the end. It was the beginning of a more rational market.
The takeaway is not to abandon AI, but to be more surgical. The broad-based rally is over. The era of stock-picking has begun. Watch the storage and data center names. Watch the software names. Watch the momentum factor. The signals are there. The question is whether you're reading them.
This is the immutable ledger of the market. Every trade is a record. Every rotation is a data point. The story of AI is still being written, but the market is now demanding proof, not promises. The next signal will come from Nvidia's earnings. The data will tell us where we go from here.
