The AI Trade Is Deleveraging: Goldman Sachs Points to Storage and Data Centers as the Next Alpha Play
The numbers hit the tape on August 23. Goldman Sachs' high-beta momentum basket fell 12% in a single week. The bank's AI hedge portfolio dropped 10% in five days. Leverage in the AI complex is coming off extreme highs. The ledger does not lie, only the auditors do. And the auditor here is Goldman Sachs, telling clients that the AI trade is not over, but the era of buying the whole sector is.
This is not a call to abandon the narrative. It is a call to change the method. The bank's core message: the AI trade is entering a deleveraging and rebalancing phase. The beta phase is dead. The alpha phase has begun. For on-chain analysts and infrastructure investors, this shift is not noise. It is a structural signal about where value accrues in the AI stack.
Let me establish the context. Since late 2022, the AI trade has been a liquidity-driven, narrative-led rally. Semiconductors, cloud providers, and AI-focused ETFs moved in near-lockstep. The market paid a premium for exposure to the AI vision, regardless of underlying earnings. That phase is ending. Goldman's positioning data shows a clear rotation: semiconductors and AI complexes have moved into short portfolios, while software has become the largest weight in the three-month momentum long portfolio. Storage and data centers are now tagged as the most tactically attractive sectors, with the bank noting that their profit recovery is not yet fully reflected in stock prices.
This is where the data detective work begins. The bank's logic is straightforward: the AI value chain is shifting from training compute to inference and data infrastructure. Training required GPUs. Inference requires storage, memory bandwidth, and data center capacity. The profit recovery in storage and data centers is the first hard evidence that AI is moving from capability demonstration to revenue contribution. Based on my experience auditing ICO contracts in 2017, I learned to trust backend logic over front-end promises. The same principle applies here. The backend of the AI trade is storage and data centers. The front-end is the narrative. Goldman is telling you to follow the backend.
Let me trace the evidence chain. First, the deleveraging signal. The 12% weekly drop in the high-beta momentum basket and the 10% five-day decline in the AI hedge portfolio are classic deleveraging markers. This is not a fundamental collapse. It is a reduction in crowded positioning. The AI trade was over-leveraged. The air is coming out of the balloon, but the structure underneath remains intact. Second, the rotation signal. Software replacing semiconductors as the largest momentum weight is a quant-factor level confirmation of a narrative shift. The market is now rewarding companies that can convert AI into revenue, not just those that sell the picks and shovels. Third, the valuation gap signal. Goldman explicitly states that storage and data centers have the most obvious valuation gaps, with profit recovery not yet priced in. This is a direct call to look at companies like Micron, SK Hynix, and data center REITs. The profit recovery is real, but the market has not caught up.
Now, the contrarian angle. Correlation is not causation. Goldman's recommendation is based on the assumption that the profit recovery in storage and data centers is AI-driven. But is it? Storage has a cyclical component. Data center demand is partly driven by traditional enterprise IT spending and cloud service provider capex cycles. The bank does not break down the AI contribution to this profit recovery. If the recovery is primarily cyclical, the AI investment thesis is weaker than it appears. Furthermore, the bank's positioning is a single-source signal. Goldman is a sell-side institution with potential conflicts of interest. It underwrites securities for many of these companies. Its call for continued AI investment may be influenced by its own business relationships. The data is clean, but the source is not neutral.
Another blind spot: the semiconductor short. Goldman has placed semiconductors in short portfolios. This could be a tactical call or a strategic one. If it is strategic, it implies a belief that Nvidia's dominance is eroding. The rise of custom ASICs, AMD's MI series, and cloud providers' in-house chips are all threats to the GPU monopoly. But if the short is tactical, it is a bet on near-term volatility, not a long-term bearish view. The distinction matters. The bank's report does not clarify this. The market is left to guess.
Let me also address the capital rotation signal. Goldman notes that capital is moving to previously overlooked areas: European and Japanese banks, gold miners, and copper stocks. This is a classic late-cycle rotation. When the leading sector becomes crowded, capital spills into value plays. The mention of copper is particularly telling. Copper is a key material for power transmission, and AI data centers are power-hungry. This is an indirect bet on AI infrastructure, but through a traditional commodity lens. The AI trade is not ending. It is spreading.
What does this mean for the next week? The catalyst is Nvidia's Q2 earnings, due at the end of August. The bank lists this as a near-term catalyst, not a risk event. That is a subtle but important distinction. Goldman expects positive signals. If Nvidia delivers strong data center revenue and a solid guide, the AI narrative gets a fresh injection. If it disappoints, the deleveraging could accelerate. The September industry conferences will provide additional direction. The market is waiting for a signal. The data suggests the signal will be positive, but the positioning is fragile.
My takeaway is this: the AI trade is not dead, but it is no longer a tide that lifts all boats. The next phase rewards precision. Storage and data centers are the most obvious candidates for repricing. The profit recovery is real, but the market has not fully priced it. The semiconductor short is a warning, not a death sentence. The software momentum shift is a confirmation that AI value is moving down the stack. The ledger does not lie. The profit recovery in storage and data centers is on the books. The question is whether the market will read it in time. Follow the data, not the narrative. The chain remembers what the market forgets.