📌 Original Viewpoint Summary

🧠 AI Semis & Memory Cycle

• $NVDA still does not look expensive on the stated setup. The core pushback: it is the world’s largest company by market cap, yet products remain supply-constrained, revenue is growing over 70%, gross margin is 75%, and PE is only 22. Hard to frame that as obvious overvaluation. 1

• $NVDA Vera Rubin NVL72 could become a major HBM revenue unlock. Bernstein estimates each rack at ~$9.1M, with HBM4 pricing rising to ~$53/GB by 2027. Memory/storage would be about ~$3.2M per rack, implying a big step-up for $MU, Samsung, and SK Hynix. 2

• AI memory demand still looks structurally under-supplied. Jensen Huang said SK Hynix doubling wafer capacity by 2030 still “is not enough,” while partnerships with SK Hynix and SK Telecom could bring hundreds of billions of dollars of business to South Korea. That reads like demand is still outrunning planned capacity. 3

📈 Market Tape & Risk Appetite

• $QQQ quickly clawing back almost half of Friday’s losses suggests dip buyers are still active. The tape is not behaving like a clean risk-off unwind. 4

🌍 Geopolitics & Middle East Risk

• The Israel-Iran tension may be easing at the margin, but the bigger watchpoint is whether Netanyahu shifts from Trump’s close ally into a political “liability.” The logic: both the U.S. and Iran appear unwilling to actively escalate, with small frictions staying limited. But Israel’s interests—especially Netanyahu’s personal political incentives—may increasingly diverge from Trump’s preferred de-escalation path. 5

⚛️ Nuclear & Infrastructure Supply Chain

• $OKLO bringing ARMEC in-house is more than a simple acquisition. ARMEC has 20+ years supporting nuclear and energy projects, and the deal internalizes high-precision machining, prototyping, and fabrication. That tightens execution control as Oklo moves toward deployment. 6