Liquidity rotated. Truth is in the tape. The broad AI rally is not dead, but the uniform beta trade that minted money for two years just got a hard reset. Goldman Sachs' latest note, which crossed my desk early this morning, breaks the AI trade into two distinct phases: the easy money phase, where you bought the index and slept well, is over. The new phase demands you check the earnings reports of storage vendors and data center operators, because that's where the smart money is now pointing its guns. Data checked. Community warned.
I've spent a decade translating institutional talking points for retail communities, and let me tell you, this is the first time in this cycle a major bank has explicitly told its clients to stop buying the whole basket and start picking a specific, unpopular corner of the hardware stack. It's a signal, and we need to decode it before the next market open. The floor price of the 'AI everything' narrative has dropped, but it hasn't broken. The truth, verified by the flow, is that we're entering a far more delicate, and potentially more rewarding, environment.
Let's cut through the noise and get to the core of the Goldman note. The headline numbers are stark. The high-beta momentum basket, a favorite haunt for leveraged bulls, has fallen 12% in a single week. The bank's own AI hedge fund basket has shed 10% in five days. This is not a healthy pullback; it's a structural deleveraging. The crowded trade is unwinding, and margin calls are likely forcing these moves. Trust bridge crossed. This is the crash of leverage, not the crash of AI itself. The bank's key takeaway is that the trade is not dead, but the excess returns from simply owning any AI-related ticker are gone. This is where the game changes from betting on the tide to picking the right ships.
What's driving this specific narrative shift? The bank's core logic is centered on a profit and loss statement. They are pointing to 'profit recovery that hasn't yet been reflected in stock prices,' specifically in memory and data centers. This is a hard pivot away from the 'narrative' stocks to the 'earning' stocks. The market is finally asking a question that my community has been screaming about for months: Where are the actual dollars coming from? In the past, you had a vision. Now, you need a balance sheet. The bank's argument is that the 'Tactically most attractive sectors' are those where the operational money has already started flowing, but the stock price hasn't caught up yet.
Here is where the institutional trade is heading. The financial flow signals are stark. The bank's three-month momentum portfolio has made a dramatic switch: software has replaced semiconductor and AI complex as the largest weight in the long basket. Simultaneously, the semiconductor and AI complex has been moved into the short basket. This is a massive re-rating of the hardware trade. It tells me that the easy money in selling picks and shovels for the training phase is done. The market is now looking for the picks and shovels for the deployment and inference phase. The 'compute' bottleneck is evolving. We are transitioning from training models to running them at scale, which is a different infrastructure beast entirely. This is a competitive landscape shift from 'who has the best chip' to 'who can store the data, power the data, and run the inference.'
I have been writing about the perils of oracle feed latency and the DeFi Achilles' heel for years. The same principle applies here, but the time horizon is longer. The storage sector isn't just about the AI model weights; it's about the enormous 'KV Cache' needed for inference, the massive data lakes for training, and the high-bandwidth memory that makes these models fast. The market is realizing that you can't run the world's AI on last year's storage infrastructure. The profit recovery in this sector isn't about the 'AI narrative' anymore. It's about actual shipping of HBM (High Bandwidth Memory) and enterprise SSDs to data centers. This is a real-world supply chain story, not a speculative dream.
The contrarian angle here is interesting, and it's the one most retail traders will miss. The public market is interpreting the 12% drop in the momentum basket as the end of AI. It is not. The deep read is that the cost of capital for AI projects is rising. This is a price discovery mechanism, not a repudiation of the technology. For crypto, this is a familiar pattern. We saw the same in 2018 when the narrative of 'blockchain' failed, but the underlying technology of digital scarcity began to build its real infrastructure. Here, the AI trade is not dead; the arbitrage of buying a 'vision' and selling it to the next guy for 50% more is dead. The new alpha is in the 'boring' corners of the market where the accountants are doing the math.
Furthermore, the bank's suggestion to look at 'previously overlooked areas' like European and Japanese banks, gold miners, and copper stocks is a tell. This is a rotation of liquidity. When an asset class that has been the 'hot money' destination for 18 months starts to leak liquidity into 'boring' value sectors, it means the risk appetite is shrinking. It does not mean the AI story is broken. It means the market's risk appetite is shrinking. The AI trade isn't over; the market's risk appetite for a concentrated, high-leverage AI trade is over. That's the distinction that will define the next quarter.
The immediate catalysts are clear. All eyes are on the tech earnings cycle, particularly the upcoming Q2 earnings from the major AI chip vendor. This report, due at the end of August, will be the first real test of the 'profit recovery' thesis. The next data point will be the September industry conferences, where new product roadmaps and supply chain commitments are announced. If the chip vendor's guidance is weak, the momentum short will press even harder. If the guidance is strong, we might see a short squeeze in the 'hardware' names. The market is at a critical inflection point where the fundamentals have to back up the price action.
From my perspective, having navigated the 2021 floor price verification sprint in NFTs, I see a familiar pattern in the quant flows. In that sprint, we built a Python script to flag wash trading bots to protect the community from fake floor prices. The same logic applies here. The 'floor price' of the AI trade is being verified by the banks' quant baskets. They are not wash trading, but they are cleaning out the speculative excess. We are at the 'Bored Ape' moment of the AI cycle, where the bots are dominating the volume, and the real value is obscured by the noise. The bank's note is the signal to start looking at the actual earnings reports, not the price charts.
Let's talk about the specific recommendation. The 'memory and data center' call is a bet on the 'pick-and-shovel' providers of the inference age. The logic is that even if the AI model developers fail or merge, the data still needs to be stored. The models still need to be served. The 'profit recovery' is already visible in the memory makers' earnings, which have been rebounding from the cyclical low. The market is treating this as a cyclical recovery, not a structural AI play. That's the opportunity. The market is giving a discount because it believes the 'recovery' is just the next peak in the semiconductor cycle. The contrarian angle is that the AI inference wave is a structural shift that will sustain this recovery longer than the market expects.
My advice to the community is to watch the 'earnings dispersion.' The stock and EPS divergence is the technical metric to follow. The bank's point is that this divergence is the highest in memory and data centers. That means the earnings are going up, but the stock price hasn't followed. In my experience in the 2022 Terra Luna collapse, I learned that when the fundamentals are absent, the price will eventually catch down to the 'truth.' Here, the fundamental is improving, but the price hasn't caught up yet. That's the 'true' alpha. This is the same as finding the 'floor price' of an NFT in 2021, but instead of verifying wash trading, we are verifying the balance sheet.
Let's address the 'elephant in the room': the fear of a full 'tech wreck.' The bank's note does not call for a complete exit. It calls for a change in the method of return. The 12% drop in the momentum basket is painful, but it's a necessary correction. The 'AI trade' is not over; it's just transitioning from a 'liquidity-driven' phase to a 'earnings-driven' phase. This is a positive sign for the long-term health of the industry. It means the market is finally starting to distinguish between the 'visionaries' and the 'integrators.' The visionaries will see their stocks fall to earth, while the integrators will see their stocks fly to the moon.
I'm not saying you should run out and buy the storage stocks right now. The deleveraging process can go on for a few more weeks. But the 'observation' of the market is that the flow is moving. The 'AI' trade is entering a new, more mature phase. The 'Goldman' note is a map for the next phase of the trade. The warning is clear: do not buy the whole sector. Do your homework. Look for the divergence between price and earnings. Look for the 'real' supply chain of AI, not the story. The 'trust' bridge has been crossed. The 'crash' is not in the technology, but in the leverage.
The 'bottom line' is this: the era of the 'blanket buy' is over. The era of the 'selective audit' has begun. The momentum has shifted from the semiconductor to the software, from the compute to the storage. This is a 'contrarian' move because the market is still looking at the past. The market is still buying the 'GPU' because it's the hero of the last cycle. The smart money is buying the 'memory' because it's the hero of the next cycle. The 'memory' of the AI world is the new 'gold mine.'
The next few months will be a period of divergence. You'll see some high-profile AI stories crash, and you'll see some boring 'storage' stocks quietly rally. The key is to check the fundamentals. Let's watch the next earnings report. Let's see if the data matches the flow. The 'AI' is not dying; it's growing up. And like any teenager, it's getting messy. The 'bull' market is not ending; the 'valuation' is resetting. The floor price of the 'AI' theme has been broken. The 'truth' is in the balance sheet. Run to the data.
As I wrote in the 2021 NFT sprint, the tools to protect the community are only as good as the data they use. The data from the institutional flows is now pointing to a 'storage' trade. This is not a 'get-rich-quick' scheme. It's a 'get-rich-slowly' scheme. The 'speed' of the information is high, but the 'speed' of the value creation is low. The market is a voter, but the 'long-term' is a weighing machine. The current 'votes' are moving to the storage and data centers. The 'weighing' will happen in the next quarterly earnings. My job is to make sure you're watching the right weights.
The trick is to be patient. The big money is not trying to buy a rally; it is trying to build a position. The momentum will continue to be volatile. The 'catalyst' for the next leg is the earnings report. If the report is good, the 'high-flyers' will get a temporary reprieve. But the 'value' is in the 'storage' and 'memory' trade. The 'growth' is there, but the 'price' has not caught up. This is the 'alpha' the bank is talking about. We are entering a period of 'stock-picking' in a 'beta-less' environment. The 'cheetah' is fast, but it has to be accurate. The 'speed' of the market is the 'data'.
I have seen this pattern in the crypto markets. The 'DeFi' summer of 2020 was the 'momentum' trade. The 'Blue chip' NFT of 2021 was the 'momentum' trade. The 'Layer-2' of 2023 was the 'momentum' trade. In every cycle, the moment the 'momentum' breaks, the market looks for 'quality'. The 'quality' is in the profit. The 'Goldman' note is the first major 'sell' signal for the 'pure momentum' AI trade. It is the first 'buy' signal for the 'real' AI infrastructure. This is a good thing for the market's health. It means the 'big money' is coming back to the fundamentals.
Don't be the last one to leave the 'momo' trade. Be the first one to enter the 'value' trade. The 'AI' is a huge secular story. It's not going away. But the 'story' needs to be supported by the 'numbers.' The 'numbers' are in the storage. The 'numbers' are in the data centers. The 'numbers' are in the software. The 'numbers' are not in the 'AI-hedge-fund' basket. The 'floor' is gone. The 'alpha' is the 'truth'. The 'run' is on to the next 'sector.'