Strategic-customer long-term agreements deepen (16 to 26 QoQ, about $32 billion of cash commitments, over 35% of pre-2030 revenue)

Customer financial commitments, predominantly in the form of cash prepayments, have accumulated to roughly $32 billion. Economically, this is customers pushing cash forward to lock in capacity, not just purchase orders. Combined with the multi-year duration, it extends demand visibility for the current up-cycle to 2030 and directly hedges the market's biggest concern, a rollover in the memory cycle. The moat extends from process technology to commercial lock-in. (2) HBM capacity is Micron's biggest soft spot. Monthly HBM capacity is about 60K wafer starts, only about one-third of that at SK Hynix and Samsung each. On NVIDIA's next-generation Vera Rubin orders, SK Hynix captured 60% to 70% share. Demand is sold out (2027 capacity is already locked), but Micron's supply footprint is clearly smaller than that of the two Korean rivals. Capacity is the ceiling on share, which is why management is spending aggressively (see the next point). Near-term HBM share is unlikely to catch up with SK Hynix quickly. (3) DRAM pricing is decelerating but still elevated. Fiscal Q4 DRAM pricing is up in the high teens QoQ, against about 60% in fiscal Q3. Third-party projections for calendar Q4 are DRAM up 15% to 20% QoQ and NAND up about 20%, with HBM strongest and Q1 2027 potentially up 80% to 100% QoQ. The deceleration is the expected high-base maths, not a cycle peak: contract prices are still rising, just moving from explosive to steady growth. The next real pricing engine shifts to the HBM4/4E repricing in 2027. (4) The capacity expansion roadmap is explicit, and total investment is extended to $250 billion through 2035. Singapore HBM packaging is due in early 2027, the Idaho ID1 fab in mid-2027, the ID2 fab at the end of 2028 and the New York fab in 2030. On Taiwan labour, the union has set a Taipei rally for 19 Oct (explicitly non-strike), with mediation on 22 Oct. That is slightly cooler than last week, but the 22 Oct outcome is still a key supply-side watch point. The long-dated, staged $250 billion of capex carries the current cash-flow surge forward into a long-term capacity moat, but most new capacity does not ramp until after 2027, so the near-term HBM shortfall gets little relief. Net: the demand and visibility side continues to strengthen (customer prepayments, multi-year long-term agreements, HBM sold out through 2027), while the supply side is both the biggest 2026 constraint and the biggest long-term moat lever.