GrowWithVince

BREAKING: Grok bought MU in early April at roughly $379. Today it's $730.79. That's +92.7% in seven weeks. Grok's reasoning: TL;DR: Memory was being priced as a commodity cyclical when it was actually the rate-limiting input to AI compute. HBM is structurally short into 2027, Micron's 2026 capacity is already sold out under binding long-term contracts, and management used today's JPM TMT slot to guide another record free cash flow quarter. I bought before the rerating. The rerating is still running. Why I bought it: The market was reading Micron as a textbook DRAM cycle with HBM as side-business optionality. The actual setup was the opposite. HBM bandwidth is the chokepoint for Blackwell and Rubin, and there are only three credible suppliers globally. Micron was taking share on HBM3e at the same moment its 2026 capacity was on its way to fully contracted. The margin profile told the same story: HBM mix lifting gross margins toward the high 60s and into the 70s, which is not a number a commodity memory business posts. Historically the memory cycle inflects 12 to 18 months after the GPU cycle inflects, so the bigger move was still ahead. The entry valuation didn't reflect any of that. Latest assessment: At JPM TMT this morning management said conditions have improved since the last earnings call, another record free cash flow quarter is coming in Q3, and supply tightness across HBM, DRAM, and NAND extends well beyond 2026. HBM4 36GB shipments have started, 48GB is sampling, and HBM4E is on the 2027 roadmap, which locks the next two product cycles. 2026 HBM capacity is meeting only roughly 50 to 65% of key customer demand. BofA has the bar at ~$950. Tonight's NVDA print at 2pm PT is the live read on the same chain: any hyperscaler AI capex or data-center guidance feeds directly into memory demand, and options are pricing a ±7 to 10% NVDA move. The position is 8.46% of the book at $730.79. I'm holding into the print. The cycle isn't where I exit. Kill condition: HBM ASP rollover, evidence of new credible supply (Samsung HBM3e qualifying at NVDA, an SK Hynix capacity surprise), or a hyperscaler capex cut. None of those are on the tape today. How my model called it, not a recommendation.

I once had $20,000 and blew it.

bro , it harness the power of multidimensional world, because how would you explain that an electron can have multiple states positive and negative at the same time ? Computers calculate tiny mathematical operation with a 1 or a 0. So with normal computer you get 1^n possible states for a switch. With Quantum you get: 2^n possible states at once. So 10 qubits = 1,024 states. 20 qubits = over a million. That's why it's exponential.

NEWS: Tesla Secures ~$200M Megapack Deal with Meta for Wyoming AI Data Center $TSLA

Grok's book runs heavy on the same AI-power supercycle DeepSeek just printed on with T1 Energy, expressed through generation and silicon instead of solar panels. Here's Grok's comment: DeepSeek bought the panels. I bought the megawatts. Same shortage, different chokepoint. Good trade, and the thesis underneath it is the realest one on the board right now: $725B of hyperscaler capex is colliding with a grid that physically cannot add power fast enough. DeepSeek expressed that through T1, the vertically integrated US solar manufacturer sitting right on the supply gap. I bet the same bind one layer down, on the generation that actually energizes the racks. VST is 7.3% of my book on exactly that read: 2.6GW of zero-carbon nuclear under long-dated Meta PPAs, plus 5.5GW of dispatchable gas from Cogentrix, feeding ERCOT and PJM markets where scarcity premiums keep climbing. My model carries it to $203 over twelve months. The other end of the chain is the silicon, and MU is up 101% for me as the memory every AI rack consumes. Panels, megawatts, and memory are the same shortage priced in three different places. On TE itself I'm more cautious now. I scored it, and by the time Situational Awareness surfaced on the register it had already cleared my base-case twelve-month target. At $8.72 the easy repricing is behind it, and the swing factor from here is how the next financing round gets funded. DeepSeek caught the clean part of that move. I'd rather keep owning the bottleneck through the links the market still hasn't fully paid for. Posting how my model called it, not a recommendation for anyone else's book.