You've heard it before. Michael Burry drops the predictable line: for any stocks going parabolic, reduce positions almost entirely. Reject the greed. AI enthusiasm has valuations stretched to the breaking point. The crowd nods along, then piles into more semis anyway. That's the consensus view right now—retail FOMO driving the melt-up while dismissing Burry as the guy who always cries bubble and misses the move. Fair enough on the surface. But zoom out and the data tells a sharper story. Parabolic moves do deserve risk management. Selling broad tech and AI exposure 'almost entirely' is exactly the wrong play when real demand is still slamming into tight supply.
Start with the tape. SOXX closed recently near $520. Its 200-day moving average sits around $316. That's roughly 65% above the trend line, velocity that echoes late 1999 before things got interesting. Momentum like this always invites the Burry warning—and history shows those warnings can be right on timing pain. Yet the setup underneath differs. This isn't vaporware speculation. It's infrastructure buildout with hard dollars attached.
NVIDIA reported Q4 FY2026 revenue of $68.13 billion, up 73% year-over-year. The company guided Q1 to roughly $78 billion. Market cap sits at about $5.2 trillion with a forward P/E around 26x on expectations of continued strong growth. Analysts model mid-40%+ revenue increases into next year. You can call that extended—and it is—but compare it to earnings power that keeps surprising higher. Burry owns SOXX January 2027 puts struck at $330, betting on a roughly 30-35% drop from here. His disclosed longs tilt toward healthcare, financials, and consumer names. That's his edge. Yours is recognizing the distinction between stretched multiples and actual backlog.
Look at the memory market for the clearest signal nobody is talking about. Korea export prices show DRAM up 18%, DRAM excluding modules up 35%, and flash memory up 47% in just one recent month. TrendForce data points to conventional DRAM contract prices rising 58-63% quarter-over-quarter in Q2 2026, with NAND Flash up 70-75%. Supply is fully booked through the year. Why? AI server demand is reallocating capacity away from consumer electronics toward high-bandwidth memory. This isn't hype. It's physics—hyperscalers committing nearly $700 billion in combined 2026 capex, up massively from prior years, to build the data centers that need these chips. When supply can't flex fast enough, pricing tightness reinforces the revenue flywheel for the leaders.
Burry is right that greed drives parabolic moves and valuations can compress violently when sentiment flips. He's wrong to paint the entire AI infrastructure theme with the same brush. The collision here is what matters: stretched multiples meeting verifiable earnings power that extends visibility through 2027. Companies aren't guiding down. Customers aren't pausing. They're signing long-term agreements to lock in supply because the alternative is falling behind on agentic AI and inference workloads that are scaling faster than expected. You don't sell the picks and shovels in a gold rush just because the prospectors are getting too excited. You trim size, manage risk, and stay positioned for the dig that keeps paying.
Connect the macro to the specific. That $700 billion capex isn't abstract—it's translating into data center orders, GPU racks, and memory modules that show up in quarterly beats. SOXX 65% above its 200-day MA looks insane until you map it against the supply chain reality where memory prices are exploding higher on allocation. NVDA's forward multiple looks reasonable when the numerator keeps compounding at 50%+. The market is pricing one path: endless growth without hiccups. Reality offers three: acceleration, digestion, or disruption. The first two favor disciplined holders. Only the third hands Burry the win.
This doesn't mean blind buy-and-hold. Parabolic stocks warrant trimming for risk management. Position sizing matters when velocity matches dot-com extremes. But broad reduction ignores the earnings bridge already visible. The real opportunity sits in that gap between consensus fear of the top and the operational data confirming demand. Ignore the noise, watch the numbers, and let reality deliver the punchline.
Deadpan fact bomb: Burry telling you to sell the parabolic move is as predictable as the parabolic move itself. Both have happened before. One resolves faster than most position for.