Buy The Dip, Sell The Dip, Repeat Until Liquidity Dies
Buy The Dip, Sell The Dip, Repeat Until Liquidity Dies
Six trading days. That's how long it took the AI trade to go from "BofA note triggers semis bloodbath, Nasdaq-100 sheds the worst week since spring" to "tech leaps 2.25% in a single session, Mag7 relief rally, somebody please clap." Same stocks. Same earnings estimates. Same Micron HBM order book everyone was quoting like scripture two weeks ago. The only thing that changed is which version of the narrative was convenient that morning, and the algos that have replaced actual portfolio managers don't have the attention span to notice the whiplash, let alone be embarrassed by it.
Let's run the tape. June 23: semis get gutted on a rate-hike note out of BofA, Asian markets tumble in sympathy, and every financial Twitter account that called the AI trade dead in 2024 comes out of hibernation to say it again, louder. June 24: Nasdaq and S&P fall ahead of Micron's print, because apparently nobody could be bothered to wait twelve more hours for the actual number. June 25: Mag7 falls because Apple and Microsoft raised prices on an iPhone and an Xbox, which is somehow bearish for trillion-dollar margin machines passing through costs to customers who will buy the devices anyway. June 26: tread water, OpenAI IPO delay reported, shrug. June 29: Nasdaq up 2%, Alphabet up almost 5%, Tesla up over 8%, and the financial press calls it a "relief rally" as if the previous five days of carnage were the disease and this is the cure, rather than both being symptoms of a market that has lost the ability to hold a thought for longer than a news cycle.
Here's the part that should actually bother you. Goldman's AI data-center construction basket is up nearly 60% year-to-date. Sixty percent, in a year where the same complex of stocks has had at least three distinct "is this the top" panics that each lasted under a week and reversed completely. That's not volatility born of genuine uncertainty about whether hyperscalers will keep building. It's volatility born of a market that has fully priced the bull case, fully priced the bear case, and now just oscillates between the two depending on which headline hit the tape last. Nobody's repricing the cash-flow assumptions on a multi-year capex cycle based on an iPhone price hike. They're repricing their own nerves.
And then there's the OpenAI IPO delay, which deserves more scrutiny than the one paragraph it got buried in. The largest, most hyped, most narrative-defining company in the entire AI capex story can't get its own paperwork in front of public market investors on schedule, and the market's reaction was essentially nothing. No discount applied to the picks-and-shovels names that have spent two years pricing in OpenAI as the permanent, insatiable customer at the center of the demand story. If the apex predator in the AI compute food chain is having trouble going public, that is information. The fact that it produced a one-day shrug instead of a re-rating tells you the market isn't actually pricing OpenAI's execution risk — it's pricing the vibe of AI as an asset class, and vibes don't do diligence.
Throw Micron into this. The company posted the kind of HBM-driven beat that should anchor a multi-week rally on its own — order books stretching out, pricing power that hasn't existed in memory chips since before most current traders had Bloomberg terminals — and it got one trading day of attention before the market moved on to the next inflection point. A blowout quarter from the company most directly levered to AI memory demand is now a forty-eight-hour story. That's not a market digesting good news efficiently. That's a market with the memory of a goldfish and the leverage of a hedge fund, treating fundamental catalysts as content to be consumed and discarded.
None of this means the AI capex cycle is fake. The capex is real, the power contracts are real, the chip orders are real. What's fake is the pretense that the day-to-day pricing of these stocks has anything to do with that underlying reality. You cannot have a sector that swings from "rate-hike note triggers rotation out of growth" to "relief rally on ceasefire optimism" within a single week and claim the moves reflect a reassessment of Micron's HBM backlog or Alphabet's capital intensity guidance. They reflect a market trading its own emotional state, with the AI complex as the most liquid, most narrative-dense vehicle available for doing so.
Thursday's jobs number will give this whole circus a fresh excuse to swing in whichever direction the headline allows. ISM manufacturing prints Wednesday into the same void. None of it will tell you anything new about whether Nvidia's order book holds up in 2027. It'll just tell you which way the goldfish is facing this week.