Abstract
The AI trade is not over. It is restructuring. Goldman Sachs' latest tactical recommendations reveal a market that has moved from beta to alpha, from narrative to fundamentals. But for those of us who follow the liquidity rather than the narrative, the real signal is not in the strategy note. It is in the arc of the trade itself.
Hook
The high-beta momentum portfolio fell 12% in a week. The AI hedge basket dropped 10% in five days. Leverage in the AI complex has retreated from extremes.
Goldman Sachs calls this a deleveraging phase. They say the AI trade is not finished but the alpha phase is over.
The data points are stark. Storage and data centers are now the tactical favorites because profit recovery has not yet been priced in. The software sector has overtaken semiconductors as the biggest weight in the three-month momentum portfolio. Capital is rotating out of AI into European and Japanese banks, gold miners, and copper stocks.
Follow the liquidity, not the narrative. The liquidity is shifting.
Context: The Trade That Defined the Cycle
The AI trade, in its purest form, was a bet on the entire sector. From late 2023 through the first half of 2024, the trade was simple: buy the infrastructure, buy the whole ecosystem, and the tide lifted all boats.
The market bought the narrative wholesale. Nvidia's AI revenue guidance, the data center buildout, the storage demand, the software potential — every component of the AI stack was purchased as a single unit.
Goldman's latest note, dated August 23, 2024, marks a formal shift in the investment thesis.
"AI trade is not over, but the phase of getting excess returns through overall sector rise is changing," the bank wrote.
This is not a bearish call. This is not a bullish call. This is a market structure call.
The sector is moving from beta to alpha. From buy-the-complex to pick-your-winner.
That is not a semantic distinction. It is the difference between a rising tide and a stock picker's market.
Core: The De-risking Is Real, the Alpha Is Real, the Value Is Real
Let's start with the numbers that matter.
The high-beta momentum portfolio fell 8% in a week. The AI hedge portfolio fell 10% in five days. The AI sector's leverage has fallen from extreme highs.
These are classic deleveraging signals. The AI trade was crowded. The trade is now being unwound. It's not a crash — it's a process.
Goldman's core tactical recommendation is: focus on stocks where price is significantly divorced from earnings per share. And it says storage and data centers are the biggest value gap in the market.
"Storage and data center sector's valuation gap is most pronounced, profit recovery has not been fully reflected in the stock price."
This is the thesis. The market has not yet priced in the recovery. That's the alpha.
The data supports this. AI infrastructure is entering a new phase. The first phase was training-centric — massive GPU clusters, massive compute. The second phase is inference-centric — the deployment of AI applications at scale. The latter has different demand profiles: less compute, more memory bandwidth, more storage for model weights and inference caches, and more data center capacity for the AI inference clusters.
HBM is a critical bottleneck. Storage is an unsung hero of AI.
And yet the market continues to price the AI complex as if it were a monolithic, undifferentiated mass.
Hashes don't lie. Wallets do.
Now let's get into the macro-level flow.
Capital is also turning to previously overlooked areas, such as European and Japanese banks, gold miners and copper stocks.
This is the capital rotation. AI is not getting new money. The money is shifting from AI to other sectors. The AI trade is a shrinking pool, and the smart money is looking for the next trade.
This is not a lack of AI conviction. This is a crowding out. The easy trades are done.
The market is moving to phase two: the fundamental differentiation.
The theme is clear: from "sell shovels" (hardware) to "pan for gold" (applications/software).
Contrarian: The De-leveraging Is Not a Bearish, It's a Correction
The consensus view is that the AI trade is unwinding and the sector is heading for a correction.
Goldman Sachs disagrees. They say the trade is not over. It's the alpha-beta that is changing. The leverage is coming down, but that's not a signal of a bearish trend. It's the market normalizing after a period of extreme positioning.
The contrarian view is to look at what's actually happening underneath.
The semiconductors and AI complex have been added to the short portfolio. The software has become the biggest weight in the three-month momentum long portfolio.
This is a subtle but important signal. It says the market is not dumping AI. It's rotating within the AI complex. The market is saying: "We've had enough of the hardware hype, now show me the software that can actually make money."
This is a maturation signal, not a reversal signal.
The AI narrative is shifting. It's no longer about the "potential" of AI. It's about the "profit" of AI.
The real signal here is the "profit recovery" in storage and data centers. This is not a tech story. This is a financial story.
The storage and data center markets are mature, consolidated industries. The AI demand is providing a new growth curve. The HBM (High Bandwidth Memory) market is tight, and the prices are going up. The enterprise SSD demand is growing.
The value is real. It's just not priced yet.
Takeaway: The Alpha is in the Data Center, Not the Narrative
The Goldman report is a tactical call. It's not a long-term bearish.
The market is moving from the "narrative phase" to the "earnings phase" of AI. The market is no longer willing to pay a premium for "AI vision". It is starting to ask for "AI revenue".
This is a positive sign for the long-term health of the AI industry. But in the short term, it may bring valuation corrections.
The key to this market is not to be a bull or a bear. It is to be an alpha trader.
Follow the liquidity, not the narrative.
On-chain truth > Twitter narrative.
The "on-chain" here is not the blockchain. It's the balance sheet. It's the P&L. It's the revenue. The narrative is what people say. The data is what the earnings say.
In this environment, the old playbook is simple: do not buy the whole sector. Buy the parts that are producing earnings. And the parts that are producing earnings are the storage and data center companies that are now being picked up.
Fragmented yields, fragmented trust. The AI trade is entering a fragmented phase. The trust is no longer in the sector. It is in the individual company. It's in the individual earnings report.
The report is the new signal. The next few weeks will be the test. Nvidia's Q2 earnings report is the catalyst. The AI chip giant's guidance will either confirm the AI thesis or undermine it.
If the data center revenue is strong, the storage and data center thesis will be validated. If the guidance is weak, the AI trade will face another round of sell-off.
The AI trade is not over. The leverage is coming down. The value is shifting.
The question is not whether AI will survive. The question is whether you are positioned for the next phase of the trade.
Follow the liquidity, not the narrative.
Hashes don't lie. Wallets do.
The next signal is on the earnings release.
Data center's profit is the next alpha.
Don't be the last one holding the beta.
Be the one holding the alpha.
Risk Matrix
| Risk | Probability | Impact | Mitigation | |:---|:---|:---|:---| | AI Trade Further Deleveraging | Medium | High | Monitor Nvidia's earnings and September industry conference. | | Storage/Data Center "Profit Recovery" Disappoints | Medium | Medium | Track storage manufacturers (Micron, SK Hynix, Samsung) and data center REITs. | | Momentum Reversal (Software → Semiconductors) | Medium | Medium | Monitor the weekly momentum factor and relative strength of software. |
Opportunity Matrix
| Opportunity | Difficulty | Time Horizon | Action | |:---|:---|:---|:---| | Storage Sector Valuation Recovery | Medium | Short-term (1-3 months) | Track HBM demand and pricing from major storage players. | | Data Center Profit Improvement | Medium | Mid-term (3-6 months) | Monitor data center utilization and rental rates. | | AI Application/Software Momentum | Medium | Mid-term (3-6 months) | Focus on AI application revenue growth and customer acquisition. |
