TD Cowen Marks Up Micron as the 2D NAND Spiral Hits the Tape

What Triggered Friday's Bid

The proximate catalyst was a TD Cowen note flagging that Micron's 2026 HBM capacity is entirely pre-sold, with the firm reiterating that pricing and volume agreements now cover the full calendar-2026 HBM book — HBM4 included. The target moved to $660 from $550, a 20% raise. That target is already through the spot price, but the message the tape is trading on is supply visibility, not the number itself: a memory company has effectively booked next year's revenue at the highest-margin product line in the stack, which is not how this industry normally behaves.

This sits on top of DA Davidson's Street-high $1,000 initiation earlier in the week and the hyperscaler chorus confirming memory cost pressure running into the June quarter. The bid on Friday was less about a new datapoint and more about the cumulative weight of three consecutive sessions of the same story.

The 2D NAND Spiral

The newer piece is what showed up in Digitimes on Friday morning: the 2D NAND shortage is now spiraling as the major manufacturers walk away from mature-process MLC supply. Samsung is ending its own MLC shipments in June, with Micron and SK Hynix also dialing back, and MLC NAND capacity is projected to drop 41.7% in 2026. The reporting describes panic buying at the spot level and industry sources who do not see how the shortage gets resolved on a normal cycle.

The strategic logic is the same one driving HBM. The three big memory makers have repointed cleanroom space and capex toward enterprise-grade DRAM and NAND for AI workloads, abandoning low-margin legacy nodes. That is what makes 2026 different from prior memory crunches: the shortage is being engineered, not absorbed.

Reading the Tape

$62.4M of 24h perp volume on the HIP-3 market is real, and the 19-hour move to $688.30 puts MU through the prior intraday high near $683 from May 7. The setup that's pricing in is bilateral: high-margin HBM revenue locked through 2027 on one side, panic-bid legacy NAND on the other, with consumer-facing memory prices expected to stay elevated well into the June quarter.

The risk that's actually worth tracking is not whether the cycle exists — it does — but whether the market starts pulling 2027 and 2028 capex risk forward. Memory has historically punished extrapolation. For now the tape is willing to pay for visibility it has rarely been offered.