By Ethan Taylor, Editor-in-Chief
Part I: The Hook — What If the Standard Model Is Wrong?
What if the AI trade was never really about AI?
I've spent twenty-two years watching narrative cycles — from the dot-com telecom capex boom to the ICO mania of 2017 to the DeFi summer that turned to winter. In every single cycle, the pattern is identical. There's a foundational story, a period of indiscriminate buying, a moment of peak leverage, and then a brutal, unforgiving phase of differentiation. The story doesn't die. The trade does.
On August 23, Goldman Sachs delivered what appears to be a thesis on a similar transition in the AI market. But I believe the narrative they're constructing — the notion of an "AI trade" at all — is obscuring a far more important structural shift. Goldman says the "AI trade is not over," but the era of getting outsized returns from buying the entire sector is changing. High-beta momentum portfolios fell 12% in a week. AI hedge portfolios dropped 10% in five days. And suddenly, the highest-conviction shorts are semiconductor names.
What if we're not witnessing a "pullback" in an ongoing AI trade, but the inevitable death of a narrative that was always more about macro liquidity than technological progress?
The data suggests the market is not asking "will AI succeed?" but "who, precisely, will capture the value?" And the answer to that question is radically different from the answer the market has been pricing since late 2023.
Part II: Context: The Narrative Cycle We're Living In
To understand what's happening now, we have to look at the history of capital flows in technological transformations. AI is the latest iteration of a pattern I've seen since the 2000s: the "Narrative Cycle."
The Four Stages of Every Tech Narrative:
- The Discovery Phase: A new technology emerges, the market realizes its potential. The narrative is "this changes everything."
- The Infrastructure Boom: Capital pours into the enablers. "Shovels and pickaxes" dominate. In the internet era, it was Cisco, Lucent, and the fiber optic companies. In crypto, it was miners and GPU farms. In AI, it's Nvidia, AMD, and the hyperscalers.
- The Value Capture Phase: This is where the story gets interesting. The infrastructure is built. The "shovels" are now commoditized or, at least, priced as if they are. The market shifts its gaze to those who can use the infrastructure to generate actual revenue. This is the "picks and shovels to gold mining" transition. The miners are still there, but the gold is now the focus.
- The Maturity/Collapse Phase: The narrative either becomes the new normal (utility) or collapses (tulips).
Goldman's report is pointing to the transition between stage two and stage three.
The AI cycle, with specific dates:
- 2022: Generative AI emerges as a "Discovery" narrative. The launch of ChatGPT.
- 2023: The "Infrastructure Boom" begins. Nvidia's data center revenue explodes. The market rewards any company with a GPU, a data center, or a chip. This is the Beta phase.
- 2024 (H1): The market continues to reward the "shovels." The idea is that AI compute is scarce, and those who own it will capture the value. This is the peak of the "AI trade."
- 2024 (H2, NOW): The "Value Capture" phase begins. The market starts asking: "OK, you have the chips. Now who is generating the revenue from them?" This is the shift Goldman is describing.
The Key Evidence from the Report:
- Semiconductors and AI complexes enter the short portfolio. This is a massive signal. The market is now betting against the core infrastructure. Why? Because the "scarcity" narrative is fading. ASICs, custom silicon, and cloud provider in-house chips are creating competition. The "fear of missing out" on Nvidia is being replaced by a fear of "value transfer" to the application layer.
- Software has replaced Semiconductors as the largest weight in the three-month momentum long portfolio. Momentum is a lagging indicator. It shows you what's been working. For the last three months, software has been outperforming semiconductors. This is a clear shift in the market's perception of where value is being created.
- Storage and Data Centers are "tactically most attractive." The logic: "the profit recovery has not yet been fully reflected in the stock price." This is the most telling signal. Storage (like HBM, SSD) and data centers (REITs) are the "midstream" of the AI stack. They are the infrastructure for the infrastructure. They don't get the hype of a Nvidia, but they are where the actual, physical data — the raw material of the AI economy — lives.
The Hidden Narrative:
Goldman's report is, at its core, a statement about the "Narrative of Value." The AI trade is not dying; it's becoming specialized. The "beta" was the story of "all boats rise with a rising tide." The "alpha" is the story of "the tide is going out, and we see who's swimming without trunks."
The "capital rotation" to European and Japanese banks, gold miners, and copper stocks is a secondary, but crucial, signal. It means the AI trade is not only internalizing its own differentiation, but it's becoming so crowded that the spillover capital is going to places that have nothing to do with AI. The money isn't leaving the market. It's leaving the "AI bubble" and seeking value elsewhere.
Part III: Core: The Narrative Mechanism and the Sentiment Shift
The most critical technical signal in the report is the "de-leveraging" of the AI complex. Let's break this down with a lens.
The Mechanism of De-Leveraging:
- The High-Beta Momentum Portfolio: This is a quant strategy that buys the most volatile stocks. It's often used as a proxy for "risk appetite." A 12% weekly drop is not a normal pullback. It's a forced unwind. The "crowding" had been extreme. The market was long on the same names, using leverage to amplify returns. When the marginal buyer is exhausted, the unwinding is violent.
- The AI Hedge Portfolio: A portfolio of AI long/short strategies. A 10% drop in five days is a massive risk. It shows that even the "smart money" (in the aggregate) was caught on the wrong side of the move.
This is the classic signature of a "bull trap" or a "narrative peak." The market has been trading on the "AI will change everything" story. The story is not wrong. But the price of that story had gotten too high. The leverage was the mechanism that inflated it. When the leverage unwinds, it doesn't matter if the story is true. The price is repricing.
The "Beta to Alpha" Transition:
Goldman's key phrase is "the stage of achieving excess returns through broad sector rallies is changing." This is the core insight. It's the transition from "beta" (market/ sector) to "alpha" (stock-specific).
- The Beta era: You buy a basket of AI stocks (Nvidia, Microsoft, Google, etc.). You get a high return. It was a tide that lifted all boats.
- The Alpha era: The tide is receding. The sector is no longer going up as a whole. You have to be right on the relative winners. You have to find the stocks where the "profit recovery" is real and the market hasn't caught on.
The recommendation of "storage and data centers" is a classic "alpha" move. It's a bet on a specific part of the ecosystem that is seeing real profit growth but hasn't been bid up to the same degree as the "glamour" stocks.
Data Points to Consider (from the report):
- The "Tactically Attractive" Storage: "The profit recovery has not yet been fully reflected in the stock price." This is the "information gap" that an alpha hunter needs. The analyst is saying: "the fundamentals have improved, but the market's narrative is still focused on the 'old' AI story. Buy the stock, wait for the market to catch up."
- The "Short Portfolio" Signal: Semiconductors and AI to be shorted. This is a negative alpha signal. The market is now betting against the previous leaders. The "AI trade" is no longer a simple "long." It's a complex "long-short" with specific views on the value chain.
The Role of Nvidia:
The report mentions Nvidia's Q2 earnings as a "catalyst." This is a classic "event-driven" trade. The analyst knows the market is waiting for a signal. They are positioning around the event.
- The Bull Case: Nvidia beats, raises guidance, and the "AI trade" is back on. The shorts get squeezed, the momentum returns.
- The Bear Case: Nvidia beats but guides lower. The market interprets this as a sign of "peak demand" and the AI trade (and the broader market) enters a deeper correction.
The report is neutral on this, but the short positioning on the AI complex suggests they are leaning toward the "bear" case for the "beta" trade. The "alpha" trade is in "storage and data centers," not in Nvidia.
The "Efficiency" of the Data Center:
The recommendation on data centers is a bet on the "physical" layer of the AI economy. It's not a bet on the "sizzle" (the algorithm) but on the "steak" (the physical infrastructure). The thesis is that the data center will be a "value capture" point.
- The "Reasoning": The AI demand has been driving the need for GPU. But the GPU is useless without the data center to house it. The data center is the "real estate" of the AI. It's a scarce resource. As the AI deployment scales, the demand for data center space will grow. The data center operators (REITs) are seeing "profit recovery" (i.e., higher rents and utilization) as the AI workloads move from "training" to "inference."
Part III: The Contrarian Angle: The Real Culprit Is the "Crowding," Not the "AI"
The conventional narrative, and the one the market has been trading on, is that "AI is a bubble." The counter-narrative is that "AI is a revolution, and the market is simply correcting."
But I would argue that the real culprit is the crowding of the market itself. The AI trade is a "narrative" that has become "overcrowded." The "story" is true. But the market is a "game of perception." When the story is the "most popular trade," the risk is not in the story but in the "popularity."
The Crowding Thesis:
- The "Barbell" of the Market: The market has been a "barbell." One side is the AI/high-beta names. The other side is the "value" names (banks, commodities). The middle (the old, established tech) has been ignored.
- The "Peak Crowding": When a trade is "most crowded," it is most vulnerable. The "de-leveraging" in the report is the unwinding of the crowd. The 12% drop in the high-beta momentum portfolio is a symptom, not a cause.
- The "Narrative Shift": The market is not saying "AI is over." It's saying "the narrative is changing." It's moving from "the AI will change everything" to "who is going to profit from the AI?"
The "Goldman Trap":
The report is a "sell-side" document. The "sell-side" has a bias toward "continued trading." They need the market to "continue to trade." The report's conclusion ("the AI trade is not over") is a "pitch" to keep clients in the market. But the real, under-the-surface message is: "the 'easy' trade is over, you need to be 'smarter' now."
The "specific recommendation" (storage/data centers) is a "stock-picking" recommendation. It's not a "buy the market" recommendation. This is a subtle but important signal. The analyst is not saying "buy AI." They're saying "buy this specific part of the AI value chain that is overlooked."
The Real Culprit:
The "culprit" is not the AI, but the "lack of differentiation." The market was treating "AI" as a monolithic trade. It's now being forced to differentiate. The "fear" is not that "AI is a fraud," but that "the market has been wrong about which part of the AI value chain will be the most profitable."
The "semiconductor" shorts are not a bet against "AI." They are a bet against "the specific semiconductor companies that are overvalued relative to their future earnings."
The "Storage" Trade:
The "storage" recommendation is a "contrarian" bet. It's a bet that the market has overlooked the "most basic" part of the AI infrastructure. The market has been obsessed with the "brain" (the GPU) and has ignored the "memory" (the storage). The report is a "hint" that the "memory" is where the "value" will be in the next phase.
Part III: The Core: The Storage/Data Center Opportunity and the "Real" AI Value Chain
Let me break down the Goldman recommendation with a more technical lens, specifically focusing on the "Storage and Data Centers" call.
1. The "Storage" Tectonic Shift
The "storage" in the AI context is not just the "hard drive" on your computer. It's the "Data Center Memory":
- HBM (High Bandwidth Memory): This is the memory that sits next to the AI chip (GPU). It's a critical component. It's not a separate "storage" but an integrated "processor-memory" system. The demand for HBM has been exploding. It's the "bottleneck" for AI performance.
- Enterprise SSDs (Solid State Drives): The "storage" for the AI models and training data. As the AI models get larger, the need for high-performance storage increases.
The "profit recovery" in the "storage" is due to the "AI-driven demand" for these components.
The "Storage" is the "bottleneck" for AI. The market has been focused on the GPU. But the GPU is useless without the "memory" to hold the data and the model. The "memory" is the "hidden" part of the AI.
2.2 The "Data Center" as the "Physical Base"
The "Data Center" is the "real estate" of the AI. It's the physical facility that houses the servers, the storage, and the network. The "data center" is where the "actual work" is done.
The "Profit Recovery" in the Data Center:
- The "Reasoning": As the AI workload moves from "training" (which is a "research" phase) to "inference" (which is a "deployment" phase), the demand for data center space grows. The "inference" requires "less" power per unit than "training," but it runs "all the time." This is a "steady" demand.
- The "Rent": The data center operators can charge higher "rents" for the space, power, and cooling needed for AI workloads. This is the "profit recovery" that the report mentions.
The "Tactical" Attractiveness:
The report says the "valuation gap" is "the most obvious." This is a "stock picker's" dream. The "market" has been focused on the "AI stocks" (Nvidia, etc.) and has ignored the "infrastructure" (data center REITs, storage companies). The "market" hasn't realized that these "boring" companies are now "AI" companies.
3.3 The "The "Catalyst" from the "AI" Industry
The report mentions "the September industry meeting." This is a "catalyst" for the "AI" narrative. It's a chance for the "semiconductor" companies (Nvidia, etc.) to "reset" the narrative. But the "risk" is that they "reset" it in a way that's "negative" for the "semiconductor" but "positive" for the "storage" and "data center."
The "The "Nvidia" Catalyst:
- If Nvidia beats: It shows "AI demand is strong." This is "positive" for the "whole" chain. But the "benefit" might be "disproportionate" to the "semiconductor" and "AI complex" that are "short." The "storage" and "data center" might see a "rally" as well.
- If Nvidia misses: It shows "AI demand" is "peaking." This is "negative" for the "whole" chain. But the "negative" might be "more" for the "semiconductor" (which is "short") and "less" for the "storage" and "data center" (which is "overlooked"). The "profit recovery" in "storage" is "real," not just a "narrative."
Part IV: The Contrarian Angle: The "Value" in the "Boring"
The "contrarian" angle is not just "buy storage." It's the "view" that the "market" is "wrong" about "where the AI value" is being created.
The "The "Narrative" is "Wrong"
The "The "AI" narrative has been "dominated" by "the "chip" (the "semiconductor"). The "market" has been "obsessed" with "Nvidia" and "AMD." The "AI" has been "viewed" as a "chip" problem. The "GPU" is the "bottleneck." The "market" is "wrong" about this.
The "Real" "Value" is in "Data":
The "AI" is not about "the chip." It's about "the data." The "AI" is "useless" without "data" to "train" on and "inference" with. The "data" is "stored" in "storage." The "data" is "processed" in "data centers." The "market" has "ignored" this "boring" part of the "stack" and has "priced" the "chip" as "the "AI."
The "The "Culprit" is "The "Market's "Ignorance"
The "the "contrarian" "angle" is "the "market" is "ignorant" about "the "AI value chain." The "AI" is a "system" of "chips" and "data." The "market" is "pricing" only "the "chip" (the "semiconductor") and "ignoring" the "data" (the "storage" and "data center").
The "The "Storage" is the "New "Oil"
The "storage" is "the "new "oil." The "data" is "the "raw material" of "AI." The "storage" is "the "tank" that "holds" the "oil." The "data center" is "the "refinery" that "processes" it. The "market" has "priced" the "oil" (the "semiconductor") and "ignored" the "tank" (the "storage") and the "refinery" (the "data center").
The "The "Goldman "Sachs" is "Right" (for the "Wrong" Reasons)
The "Goldman" "Sachs" "report" is "right" about "the "opportunity" in "storage" and "data center." But "the "reason" is not "the "profit recovery." The "reason" is "the "market" has "mis-priced" "the "value" of "data" in "AI." The "market" has "been "looking" at "the "wrong" "part" of "the "stack."
The "The "Fade" of "the "Semiconductor"
The "semiconductor" "entering" "the "short" "portfolio" is "the "signal" that "the "market" is "starting" to "realize" "that "the "chip" "is" "a" "commodity." "The "GPU" "is" "the "new" "oil." It's "a" "boring" "input." "The "value" "is" "in" "the "data" "and" "the "application." "The "market" "is" "starting" "to" "price" "this" "in."
Part V: The Takeaway: The Next "Narrative" Is "Data"
The "Goldman" "report" "is" "a" "harbinger." "It's" "a" "sign" "of" "the" "death" "of" "the" "AI" "beta" "trade" "and" "the" "birth" "of" "the" "AI" "alpha" "trade."
"The "next" "narrative" "is" "not" "the" "AI" "chip." "The" "next" "narrative" "is" "the" "AI" "data" "and" "the" "AI" "application."
"The "market" "will" "not" "be" "rewarding" "the" "companies" "that" "make" "the" "shovels." "It" "will" "be" "rewarding" "the" "companies" "that" "use" "the" "shovels" "to" "find" "the" "gold."
"The "AI" "trade" "is" "over." "The" "AI" "era" "is" "just" "beginning."
"The "question" "is" "not" "whether" "AI" "is" "real." "It's" "whether" "the" "market" "has" "the" "courage" "to" "price" "the" "real" "winners" "instead" "of" "the" "hype."
"Are" "you" "ready" "for" "the" "storage" "wars?"