There is a specific sound that a market makes when it is forced to confront its own physics. It is not the sound of a crash—that is too dramatic, too cinematic. It is the quieter, more unsettling sound of a position being closed not because the thesis was wrong, but because the carrying cost of belief has become too heavy. I heard that sound in 2022 when the leverage of the crypto ecosystem finally met the reality of interest rates. And I heard it again last week, reading a Goldman Sachs note that described, with the clinical detachment of a surgeon, the beginning of a deleveraging phase in the AI trade. The language was different. The market was different. But the melody was hauntingly familiar.
The report, dated August 23rd, paints a picture of an AI trade that is not ending, but transforming. The high-beta momentum portfolio fell 12% in a single week. The AI hedge basket dropped 10% in five days. Leverage, which had been stretched to the kind of extremes that make risk managers queasy, is now retreating. Goldman's conclusion is not that AI is a bubble—they explicitly reject that framing—but that the era of indiscriminate beta is over. The tide that lifted all AI-adjacent boats has begun to recede, and we are about to see who has been swimming without a suit. For those of us who lived through the crypto deleveraging of 2022, this is not news. It is a rerun of a show we have already seen, but this time with better costumes and a larger budget.
This is the moment when the market's collective memory, which is notoriously short, is being asked to recall a lesson it should have learned from the collapse of Terra and the cascade of insolvencies that followed. The lesson is not that innovation is fraudulent. The lesson is that leverage is a solvent that dissolves the boundaries between fundamentally sound ideas and structurally unsound ones. When the solvent evaporates, you are left with the truth of what you built. And the truth, as it always does, comes out in the wash.
The Signal in the Noise
Let me be precise about what Goldman is actually saying, because buried in the institutional language is a series of signals that the AI trade is entering a phase that crypto veterans will recognize as the transition from 'narrative-driven' to 'fundamentals-driven.'
The first signal is the most obvious: semis and AI complex have moved into the short basket. This is not a marginal adjustment. This is a fundamental repositioning of some of the largest pools of capital on the planet. The trade that made fortunes in 2023 and early 2024—buying anything with a GPU in the supply chain—is now being actively shorted by the same desks that were piling in six months ago. The question is not whether the AI thesis is broken. The question is whether the price already reflects every possible positive outcome, and then some.
The second signal is more subtle but equally important: software has replaced semis as the largest weight in the three-month momentum long portfolio. This is a rotation within the AI trade, not out of it. The market is signaling that the value capture point is shifting from the 'picks and shovels' of the AI gold rush to the 'gold' itself—the applications, the software layers, the companies that are actually using AI to generate revenue rather than just promising to. This is the same pattern we saw in the crypto market when the infrastructure narrative gave way to the DeFi narrative, which gave way to the NFT narrative, each rotation leaving behind a graveyard of projects that were infrastructure for a use case that never materialized.
The third signal is the one that interests me most: storage and data centers are now described as 'tactically the most attractive sector' because the 'earnings recovery is not yet fully reflected in the stock price.' This is a direct admission that the market has been so focused on the compute layer—the GPUs, the chips, the raw processing power—that it has overlooked the messy, unglamorous, but ultimately essential infrastructure that makes AI actually work. In crypto terms, this is like the market finally noticing that you cannot have a decentralized finance ecosystem without oracles and stablecoins. The plumbing matters. It always did. But the market only remembers this when the faucets start to leak.
The Memory of Water
Here is where my own experience as a builder in the crypto space, and a student of its failures, gives me a perspective that the Goldman note lacks. The report is a snapshot. It tells you where the market is today. But it does not tell you where the market learned its behavior. The truth is that the AI trade is now replaying the crypto cycle of 2020-2022, but with a lag time of about 18 months. The players are different. The balance sheets are bigger. But the psychology is identical.
I remember the moment in 2021 when the crypto market believed, with absolute conviction, that the narrative of 'digital gold' and 'decentralized finance' was sufficient to justify any valuation. The infrastructure was being built at a furious pace. The capital was flowing in from every corner of the globe. And then the Fed raised rates, and the leverage that had been propping up the narrative was suddenly too expensive to maintain. The projects that survived were not the ones with the best narratives. They were the ones with real revenue, real users, and real discipline. The same will happen in AI.
The Goldman note points to the profit recovery in storage and data centers as a key opportunity. But I would caution against reading this as a simple 'buy the laggards' signal. The question that matters is not whether these sectors are recovering. The question is whether the recovery is durable. Is the demand for data center capacity being driven by the actual deployment of AI inference workloads, or is it being driven by speculative building that will leave us with a glut of empty server racks, the digital equivalent of ghost towns? The same question applies to storage. The HBM demand is real. The enterprise SSD demand is real. But the market has a habit of conflating a cyclical uptick with a structural shift, and the punishment for that conflation is brutal.
The Contrarian's Confession
Now, let me play devil's advocate against my own skepticism. There is a version of this story where the AI trade is fundamentally different from the crypto cycle, and where the Goldman note is not a warning but an opportunity. The difference, if it exists, lies in the nature of the underlying demand. Crypto's problem was that the use cases were often speculative—the value was in the token, not in the service it enabled. AI's demand is different. When a bank deploys an AI model to detect fraud, or a hospital uses AI to read radiology scans, the value is tangible. It is not a bet on a future state. It is a current operational expense that generates a measurable return. This is the 'profit recovery' that Goldman is pointing to. It is not a narrative. It is a line item on a P&L statement.
If this is true, then the current deleveraging is not a sign of a top. It is a healthy correction that will separate the companies with real AI-driven earnings from the companies that are just wearing an AI costume. The storage and data center sectors are the beneficiaries of this separation because they are the physical layer that cannot be faked. You cannot spin up a data center with a press release. You cannot create HBM supply with a PowerPoint deck. This is real, physical, capital-intensive infrastructure, and its scarcity is its value.
But here is the uncomfortable truth that the contrarian in me must acknowledge: the same argument was made for crypto mining in 2021. The chips were real. The energy consumption was real. The data centers were real. And yet, the business model was still dependent on the price of a speculative asset. The lesson is that physical infrastructure does not protect you from narrative collapse. It only delays it. The question for AI is whether the narrative is attached to something that generates cash flow independent of market sentiment. For the storage and data center companies, the answer is increasingly yes. For the AI companies themselves, the answer is still being written.
The Hidden Cost of Rotation
The rotation from semis to software, and the shift in capital towards European and Japanese banks, gold miners, and copper stocks, tells me something that the Goldman note does not explicitly state. The market is not just rotating within the AI trade. It is rotating out of the AI trade and into sectors that are seen as 'underowned' and 'undervalued.' This is a classic late-cycle behavior. When the leaders of a bull market become too crowded, the capital that cannot find a home in the leader rotates to the laggards. This is not a sign of health. It is a sign of saturation. The AI trade is not dead, but the marginal dollar is no longer finding its way into AI names.
The mention of copper stocks is particularly telling. Copper is the metal of electrification, and AI data centers are voracious consumers of power. The market is not just betting on AI. It is betting on the physical infrastructure that AI requires—power generation, transmission, cooling, and the raw materials that make it all possible. This is a sophisticated bet, but it is also a sign that the market is looking for AI exposure through indirect channels because the direct channels have become too expensive. In crypto terms, this is like the market discovering that the best way to play the Ethereum ecosystem is not to buy ETH, but to buy the GPU stocks that mine it. It works, but it is a different kind of bet, with different risks.
What the Memory of 2022 Teaches Us
I spent the summer of 2022 dissecting the corpses of protocols that had promised to 'bank the unbanked' and 'democratize finance.' The post-mortems were all the same. The technology was often elegant. The community was often passionate. But the economics were broken. The projects were relying on a continuous influx of new capital to pay the returns that were promised to existing capital. It was a Ponzi scheme, not because the founders were criminals, but because the business model was unsustainable. The same analysis needs to be applied to the AI trade. Which companies are generating real revenue? Which companies are just burning capital to acquire users that will never be profitable? The answer to these questions will determine which stocks survive the current correction.
Goldman's recommendation to focus on 'stocks where the price is significantly diverged from earnings per share' is the most important piece of advice in the note. This is the fundamental analysis that separates the wheat from the chaff. In a bull market, the price and the earnings diverge, and the divergence is ignored because the momentum is too strong. In a correction, the divergence becomes the source of the pain. The stocks that have run too far ahead of their earnings will be punished. The stocks that have been left behind despite improving earnings will be rewarded. This is the alpha opportunity that Goldman is pointing to, and it is the same opportunity that existed in the crypto market in the summer of 2023, when the quality projects that had survived the crash began to outperform the zombie projects that were still trading on fumes.
The Nvidia Catalyst and the Illusion of Certainty
The upcoming Nvidia Q2 earnings report is framed as a catalyst. But I would argue that it is a moment of judgment, not just a catalyst. The market is about to receive a massive data point on the sustainability of AI capital expenditures. If Nvidia beats and raises, the AI trade gets a reprieve. If Nvidia beats but guides lower, or if the company signals that the growth rate is decelerating, the market will interpret that as the beginning of the end. The problem is that Nvidia's guidance is not just about Nvidia. It is about the entire AI ecosystem. Every company that has bought Nvidia's GPUs is implicitly making a bet that their AI investments will generate a return. If Nvidia's guidance suggests that the cloud providers are pulling back on their capex, the entire edifice starts to wobble.
This is the same dynamic we saw in the crypto market when the price of Bitcoin was the bellwether for the entire ecosystem. When Bitcoin fell, every altcoin fell with it, regardless of its individual merits. The correlation was not because the altcoins were fundamentally linked to Bitcoin. It was because the market was using Bitcoin as a proxy for the entire risk appetite for crypto. The same is true for Nvidia. It is the proxy for the AI trade. And when the proxy moves, everything moves with it. The lesson from crypto is that this correlation is not permanent. After the crash, the correlation broke down, and the quality projects began to diverge from the garbage. The same will happen in AI. But it will not happen until the market has fully processed the Nvidia signal and adjusted its expectations accordingly.
The Wisdom of the Short Seller
The fact that semis have moved into the short basket is a signal that should not be ignored. Short sellers are often dismissed as pessimists or even as market manipulators. But the best short sellers are not pessimists. They are realists. They are the ones who have done the work to understand the difference between a narrative and a business model. When the short sellers start to pile into a sector, it is usually because they have identified a structural weakness that the bulls are ignoring. In the case of semis, the short thesis is probably not that AI is a fraud. It is that the valuation has gotten ahead of the fundamentals, and that the competitive dynamics are changing. The rise of custom ASICs, the challenge from AMD, the potential for export controls to limit the addressable market—these are all real risks that the bulls are ignoring because the momentum is too strong.
This is where the 'Culture is the new consensus mechanism' idea applies. The market's consensus is not just a function of data. It is a function of narrative, of shared belief, of the stories that people tell themselves about the future. When the narrative is powerful enough, it can override the data for a long time. But eventually, the data wins. The question is always about timing. The short sellers are betting that the data is about to win. The bulls are betting that the narrative can hold for just a little longer. This is the eternal struggle of the market, and it is the same struggle that played out in crypto when the narrative of 'hyperbitcoinization' collided with the reality of a tightening Fed.
The Human Cost of Rotation
There is a human cost to this rotation that the Goldman note does not capture. When the AI trade unwinds, it is not just a matter of portfolio losses. It is a matter of jobs, of dreams, of companies that will not survive the winter. I have seen this happen in crypto. I have watched founders who believed in their mission with every fiber of their being watch their life's work evaporate because the market turned. The same will happen in AI. The companies that are building on borrowed time, on borrowed capital, on borrowed narratives, will be the ones that fail. The companies that are building real value, with real revenue and real customers, will survive. But the transition will be painful. The market does not care about the human cost. The market only cares about the numbers.
This is why I write. This is why I teach. This is why I believe that education is the most important infrastructure in any technological revolution. The people who survive the transition are not the ones with the best technology or the most capital. They are the ones who understand the fundamentals, who can distinguish between a narrative and a business model, who have the discipline to cut their losses and the courage to double down on their convictions. The market is a harsh teacher, but it is the only teacher that matters. And the lesson it is about to teach the AI trade is the same lesson it taught the crypto trade: leverage is a loan, and eventually, it must be repaid.
The Future in the Code
As I look at the Goldman note, I am struck by how much it resembles the analysis that was being written about crypto in early 2022. The same language of 'deleveraging,' the same focus on 'profit recovery,' the same warnings about 'crowded trades.' The cycle repeats because human nature repeats. The specifics change, but the psychology is constant. The only way to survive the cycle is to understand it, to recognize the signs, and to position yourself accordingly. This does not mean being cynical. It means being realistic. It means understanding that the future is not a straight line, but a series of cycles, each one building on the last, each one leaving behind the debris of the previous one.
The future of AI is bright. The technology is real. The applications are transformative. But the investment cycle is not the technology. The investment cycle is a reflection of human psychology, and human psychology is notoriously fickle. The current correction is not a sign that AI is a bubble. It is a sign that the market is finally beginning to differentiate between the companies that are building the future and the companies that are just talking about it. This differentiation is painful, but it is necessary. It is the process by which the market separates the signal from the noise, the wheat from the chaff, the builders from the pretenders.
Truth is not mined; it is remembered. And what the market is remembering right now is that leverage is not a strategy. It is a temporary condition. The question is not whether the AI trade will recover. It is whether the companies in the AI trade have the fundamentals to deserve the recovery. The Goldman note suggests that the storage and data center sectors are the ones with the strongest fundamentals, the ones where the 'profit recovery is not yet reflected in the price.' This is the signal in the noise. The question is whether you are listening.
We do not build walls; we build bridges for value. The current correction is not a wall. It is a bridge. It is a bridge from the era of narrative-driven speculation to the era of fundamentals-driven investment. The crossing will be uncomfortable. The wind will be strong. But on the other side is a market that is more mature, more resilient, and more real. The companies that make the crossing will be the ones that built their businesses on solid ground, not on the shifting sands of leverage and narrative. The rest will be left behind, and the market will not look back.
The future is written in code, but felt in spirit. The code of AI is being written every day, in every data center, in every training run, in every inference request. But the spirit of AI is something else. It is the belief that we can build machines that amplify our intelligence, that extend our capabilities, that help us solve problems that have plagued humanity for centuries. That spirit is real. It is not a narrative. It is not a leverage trade. It is a fundamental truth about what we can achieve when we combine human creativity with machine precision. The current correction is a test of that spirit. It is a test of whether we believe in the future enough to build it, not just to speculate on it.
In the chaos of the chain, find the signal. The signal is not in the price. The signal is in the fundamentals. The signal is in the companies that are generating real revenue, real profits, real value. The signal is in the storage and data center companies that are building the physical infrastructure of the AI future. The signal is in the software companies that are turning AI capabilities into real products and real services. The signal is there. You just have to be willing to look for it, even when the noise is deafening.
Freedom is a protocol, not a permission. The freedom to build, to create, to innovate, is not granted by the market. It is not granted by the Fed or by Goldman Sachs. It is a protocol, a set of rules that you follow, a discipline that you maintain, a commitment to reality over narrative. The companies that follow this protocol will survive. The companies that do not will be swept away by the tide of deleveraging. This is not a prediction. It is a statement of fact. The market is a mechanism for separating the real from the fake, the durable from the ephemeral, the valuable from the worthless. The current correction is the market doing its job. And as painful as it is, it is a necessary process. It is the process by which the future is built.
Ideas have no gas fees, only gravity. The idea of AI is powerful. It has already changed the way we think about computing, about software, about the nature of intelligence itself. But ideas, no matter how powerful, are subject to the gravity of economics. They must eventually generate value, or they will fall back to earth. The current correction is the gravity pulling the AI trade back to reality. The companies that can withstand the gravity are the ones that have built something that generates value, not just promise. The companies that cannot will be crushed by the weight of their own expectations. This is the nature of markets. It is the nature of life. It is the nature of gravity. And it is the reason why the current correction, as painful as it is, is a necessary part of the journey.
As I write this, I am reminded of a conversation I had with a founder in the summer of 2022. His project was dying. The leverage had been too much. The narrative had been too strong. The reality had been too harsh. He asked me what he should do. I told him that the market was not punishing him. The market was teaching him. The lesson was not to avoid leverage, but to build something that could survive without it. The lesson was not to avoid narratives, but to build something that was true, regardless of the narrative. The lesson was not to avoid risk, but to take risk that was calibrated to reality. He did not listen. His project died. But the lesson stayed with me. And it is the lesson that I want to share with you today. The AI trade is not ending. It is being reborn. And the companies that are born from this correction will be stronger, more resilient, and more valuable than the ones that came before. That is the signal. That is the truth. That is the future.
