The AI Trade Is Deleveraging: What Goldman's Pivot Reveals About the Next Liquidity Cycle

The ledger does not lie, only the noise obscures. On August 23, Goldman Sachs published a tactical note that, stripped of its institutional polish, reads as a confession: the AI trade, the most crowded and leveraged bet of the post-2023 era, is now in active deleveraging. The high-beta momentum basket fell 12% in a single week. The AI hedge fund basket dropped 10% in five days. These are not corrections. These are forced liquidations, the mechanical unwinding of positions built on borrowed conviction. I have spent the better part of three decades auditing balance sheets, first in traditional finance, then in the crypto derivatives market. I have seen this pattern before. In 2020, I watched DeFi protocols with triple-digit APYs collapse under the weight of their own emissions schedules. In 2022, I modeled the solvency of Terra's anchor protocol and concluded that its yield was a phantom. The same mathematical inevitability is now playing out in the AI equity complex. The only difference is the asset class. Goldman's core message is not that AI is a bubble. It is that the beta phase is over. The period where buying any AI-adjacent stock guaranteed outsized returns has ended. What remains is a market of structural alpha, where the ability to distinguish between real earnings and narrative-driven multiples will determine survival. This is not a bearish thesis. It is a maturity signal. And for those of us who have spent years analyzing liquidity cycles, it is a familiar one. Let me be precise about what the data shows. The AI hedge fund basket, which Goldman tracks as a proxy for institutional positioning, has shed 10% of its value in five sessions. The high-beta momentum basket, a collection of the most volatile and trend-following stocks, is down 12% weekly. These are not random fluctuations. They are the signature of deleveraging, the process by which margin calls force the sale of assets regardless of their fundamental quality. When leverage is unwound, price discovery is suspended. The market does not care about the quality of your thesis. It only cares about your ability to meet the margin requirement. This is where my experience in crypto becomes directly relevant. In the summer of 2020, I modeled the liquidity decay of Curve Finance's initial token emissions. The protocol was offering yields that were mathematically unsustainable, and I shorted the governance token accordingly. The subsequent Harvest Finance collapse validated that thesis. The lesson I learned was simple: when a market is driven by leverage and narrative, the first sign of trouble is not a price drop. It is a change in the structure of flows. Goldman's note is precisely that signal. The firm identifies storage and data centers as the most tactically attractive sectors, arguing that their profit recovery has not yet been fully reflected in stock prices. This is a critical insight. It suggests that the AI value chain is shifting from the compute layer, where margins are being compressed by competition and export controls, to the infrastructure layer, where pricing power is returning. Storage companies, particularly those with exposure to high-bandwidth memory (HBM) and enterprise SSDs, are seeing demand that is not yet priced in. Data center operators, especially those with scale and efficient power management, are experiencing rising utilization rates and rental yields. But here is the contrarian angle that most market participants will miss. Goldman's recommendation to rotate into storage and data centers is not a bullish signal for AI. It is a defensive move. The firm is acknowledging that the semiconductor trade, which has been the primary vehicle for AI exposure, is now overcrowded and overvalued. By moving down the capital stack, Goldman is effectively hedging against the possibility that AI training demand, the primary driver of GPU sales, is peaking. The shift from training to inference is real, but it does not require the same density of compute. It requires bandwidth, memory, and physical infrastructure. These are lower-margin, higher-volume businesses. The inclusion of semiconductors in the short basket is the most telling signal. Goldman is not just reducing exposure. It is actively betting against the sector. This is a strategic judgment, not a tactical one. It reflects a belief that the competitive dynamics of the AI chip market have changed. Nvidia's dominance is no longer uncontested. Custom ASICs, cloud provider in-house silicon, and AMD's MI series are eroding the moat. Add the overhang of US export controls, which limit the addressable market, and the risk-reward profile of semiconductor equities becomes asymmetric to the downside. Software, by contrast, has become the largest weight in the three-month momentum long basket. This is a quantifiable shift in factor allocation. Momentum strategies are trend-following by nature, and the fact that software has overtaken semiconductors suggests that the market is pricing in a new phase of AI commercialization. The narrative has moved from 'we are building the infrastructure' to 'we are deploying the applications.' This is where the real revenue growth will occur, but it is also where the failure rate is highest. Most AI software companies will not survive. The ones that do will have proprietary data, distribution channels, and a clear path to monetization. I have seen this movie before. In the crypto market, the equivalent moment was the transition from the 2020 DeFi summer to the 2021 NFT boom. The infrastructure was built, the protocols were audited, and then the market shifted to applications. Most of those applications failed. But the ones that succeeded, the ones that solved a real problem and had a sustainable business model, generated outsized returns. The same will happen in AI. The key is to identify which software companies have a defensible position and which are riding the narrative. Goldman's note also mentions capital rotating into European and Japanese banks, gold miners, and copper stocks. This is a classic late-cycle signal. When the most crowded trade starts to unwind, the capital does not leave the market. It rotates into undervalued, unloved sectors. The fact that Goldman is highlighting these areas suggests that the AI trade is not just deleveraging. It is being replaced. The marginal buyer of AI stocks is becoming the marginal buyer of value stocks. This is a significant shift in market structure. For crypto investors, the implications are profound. The AI trade and the crypto trade are not separate markets. They are both expressions of the same macro liquidity cycle. When the Federal Reserve expands its balance sheet, risk assets rise. When it contracts, they fall. The deleveraging of the AI trade is a leading indicator for the broader risk complex, including digital assets. If the AI trade is unwinding because of margin calls and forced selling, the crypto market will feel the same pressure. The correlation between Bitcoin and the Nasdaq is not a coincidence. It is a reflection of the same underlying liquidity dynamics. My framework for analyzing this is based on the concept of liquidity decay. Every asset class has a carrying cost, and when the cost of leverage exceeds the return on the underlying asset, the trade becomes unsustainable. The AI trade reached that point in August. The high-beta momentum basket's 12% weekly decline is the market's way of repricing the cost of leverage. The question is whether this is a short-term correction or the beginning of a longer-term trend. Based on my analysis of historical cycles, I believe this is the beginning of a structural shift. The AI trade has been the primary driver of equity market returns since 2023. It has absorbed an enormous amount of capital and leverage. The unwinding of that position will not happen in a week. It will take months, and it will be characterized by high volatility and sharp reversals. The key is to avoid catching the falling knife and to position for the next phase of the cycle. Storage and data centers are the obvious beneficiaries. The profit recovery in these sectors is real, and it is not yet priced in. But the window of opportunity is narrow. Once the market recognizes the shift, the valuation gap will close quickly. The same applies to software companies with genuine AI revenue. The momentum factor has already identified them, but the fundamental investors have not yet arrived. This is the classic arbitrage between quant and fundamental strategies. I will offer a specific example from my own experience. In 2024, I conducted a comparative risk assessment of the custody structures of BlackRock's IBIT and Fidelity's FBTC. The analysis focused on insurance coverage and cold-storage key management. The conclusion was that IBIT had superior institutional safeguards. This was not a popular opinion at the time, but it was based on a code-first verification bias. I audited the structures, not the narratives. The same approach applies to the current AI market. Do not listen to the story. Audit the balance sheet. The Goldman note is a valuable piece of analysis, but it is not a complete picture. It is a single source, and it has inherent biases. Goldman is a sell-side institution with market-making and underwriting relationships. Its recommendations are not purely altruistic. However, the data it provides is verifiable. The 12% weekly decline in the high-beta momentum basket is a fact. The 10% drop in the AI hedge fund basket is a fact. The rotation from semiconductors to software in the momentum factor is a fact. These are the signals that matter. My takeaway is this: the AI trade is not over, but the easy money has been made. The next phase will reward those who can identify structural alpha, not those who chase beta. For crypto investors, the lesson is to monitor the macro liquidity cycle and to understand that the AI trade and the crypto trade are intertwined. When the AI trade deleverages, crypto will feel the pressure. But when the cycle turns, both will recover. The key is to survive the transition. Liquidity is a phantom; solvency is the skeleton. The AI trade is not insolvent. The underlying companies are generating real revenue and real profits. But the leverage that has been built on top of those profits is unsustainable. The deleveraging process will be painful, but it will also create opportunities. The investors who can identify the companies with real earnings, real cash flow, and real competitive advantages will emerge from this cycle stronger. The ones who are trading on narrative and leverage will be eliminated. Inversion is the only constant in chaos. The current market chaos is an opportunity to reassess, to rebalance, and to reposition. The AI trade is not dead. It is evolving. And the investors who evolve with it will be the ones who profit from the next phase of the cycle. The ledger does not lie. The data is clear. The question is whether you are willing to read it.