Hook: Over the past seven days, the high-beta AI momentum basket lost 12% of its value. The Goldman Sachs AI hedge fund basket dropped 10% in five sessions. This is not a crash. This is a levered position being unwound. And if you’ve traded crypto cycles, you know exactly what happens next: the smart money doesn’t exit — it rotates.
Context: Goldman’s latest note on the AI trade is the most honest thing I’ve read from a sell-side firm in years. They don’t call it a bubble. They call it a "de-leveraging and rebalancing" phase. The AI trade isn’t dead — the days of buying any ticker with "AI" in the name and getting a free beta ride are over. The narrative-driven phase is kaput. What’s replacing it is a fundamentals-driven phase where you need to look at P&L, not pitch decks.
I’ve seen this pattern before. In 2020, during DeFi Summer, the entire sector pumped on liquidity injections. Then, in mid-2021, the broad gains stopped. Only the protocols with real revenue — Uniswap, Aave, GMX — continued to compound. The rest bled out. Goldman is saying the same thing about AI: the overall sector beta is done, but the alpha opportunities are more concentrated than ever.
Core: The most actionable signal from Goldman’s report is the rotation of the momentum factor. Software has replaced semiconductors as the largest weight in the three-month momentum long portfolio. Semiconductors have moved into the short portfolio. That is a quant-level signal that the market is repricing where value is captured in the AI stack.
Let me break this down with a crypto analogy. In 2023, the biggest winners were the "shovel sellers" — GPU miners, cloud GPU rental platforms, and the L1s that hosted the most compute. That was the semiconductor phase. Then, in late 2023, the market started rewarding the "application layers" — AI agents, decentralized inference networks, and data storage protocols. That’s the software phase. Goldman is saying the same rotation is happening in TradFi. The first wave of AI investment was about building the infrastructure. The second wave is about what you can build on top of it.
Goldman explicitly calls out storage and data centers as the most tactically attractive sectors — because their profit recovery is not yet priced into the stock. This is the equivalent of finding a DeFi protocol that’s generating $10M in monthly fees but trading at a 5x P/E. The institutional market hasn’t caught up yet. The data center REITs and storage manufacturers (think HBM, enterprise SSDs) are currently trading at a discount to their earnings improvement. That’s the gap.
Let me ground this in my own experience. In 2020, I built a yield farming bot that arbitraged fee discrepancies between Compound and Uniswap. The strategy worked because the market was inefficient — it hadn’t yet priced in the fee differentials. The same principle applies here. The market is inefficient in pricing the profit recovery of AI infrastructure plays. The thesis is simple: as AI inference scales, the demand for storage (model weights, KV cache, training data) and data center capacity (inference clusters) grows faster than the supply. The operators have pricing power. The margins are improving. The stock prices haven’t caught up.
The contrarian angle: retail investors are panicking about an AI bubble, but the smart money is positioning for the next leg. The narrative that "AI is a bubble" is itself a lagging indicator. When the Goldman hedge fund basket drops 10%, the media screams "AI crash." But what’s actually happening is a rotation from over-owned semiconductors to under-owned storage and data centers. The capital is leaving the hype and entering the infrastructure that actually generates cash flow.
I saw this exact dynamic in 2022 during the Terra/Luna collapse. I shorted Luna weeks before the crash because I audited the peg mechanism and found no cryptographic reserves. The market was still pricing Luna at $80. The smart money had already moved to stables and Bitcoin. The same thing is happening now in AI. The market is still pricing NVIDIA at $1,000B, but the real value is accruing to the companies that operate the infrastructure, not just the ones that sell the picks and shovels.
Goldman also notes that the de-leveraging isn’t over. The high-beta momentum basket is still elevated above historical norms. But the catalyst is clear: the next inflection point is NVIDIA’s Q2 earnings (late August) and the September industry conferences. If NVIDIA’s guidance is strong, it will confirm the infrastructure demand narrative. If it’s weak, the de-leveraging will accelerate. Either way, the rotation is already happening.
Takeaway: The AI trade is not over. It’s rotating from hype to fundamentals. The most actionable play right now is to look at storage and data center plays where the profit recovery is real but the stock price hasn’t caught up. Monitor NVIDIA’s earnings as a directional signal. If the earnings confirm the trend, the rotation will accelerate. If not, the de-leveraging will deepen. Either way, the days of buying the whole sector are done. It’s time to pick individual winners.
— Root: Auditing the DAO and Ethereum
— Root: Auditing the DAO and Ethereum
We farmed the yields until the protocol farmed us.
— Root: Auditing the DAO and Ethereum
