Micron Slides as Citi Cuts Target: 6% DDR5 Softness vs. Sold-Out AI Memory

Citi's target cut turned a near-term shock into a broader debate

Citi lowered its Micron target to $425 from $510 after mainstream DDR5 16GB DRAM prices fell about 6% since earnings. Even so, the stock's steeper post-note decline showed that investors were using the hit to revisit a bigger question: is this a temporary spot-price wobble, or the start of a wider re-rating?

Pre-holiday position rebalancing ahead of Good Friday likely intensified the first leg of the selloff. That does not make the concern less real, but it does explain why a move that began as a pointed warning quickly turned into a more emotional trade.

The core debate is straightforward. Bears can point to softer mainstream DRAM pricing and ask how durable Micron's margin strength really is. Bulls can point to AI-memory tightness and note that Citi did not cut its earnings outlook, emphasizing that contract negotiations could help cushion spot weakness.

AI memory scarcity vs. inference efficiency

Why efficiency matters even if supply stays tight

The key repricing question is whether inference efficiency can reduce memory demand per task fast enough to weaken the scarcity premium. Citi flagged that efficiency techniques can reduce compute and memory cost per query, which helps explain why the stock can stay under pressure even with HBM capacity sold out through 2026.

That tension is not new. Micron said it had sold out its HBM chips for this year and the next, yet the shares still sold off because its fourth-quarter revenue forecast of $7.6 billion only matched consensus. The takeaway was less about current HBM demand and more about investor expectations: when a stock is priced for outsized AI growth, matching estimates can still disappoint.

Why the scarcity case still has support

The bullish case is not that efficiency is impossible. It is that lower cost per query does not automatically mean less total memory demand. Citi argued that those efficiency gains could further unlock usage, which ultimately increases compute/memory demand, with the net effect on demand depending on how adoption scales.

Citi also highlighted the role of KV cache in AI workloads, noting that compute intensity rises as each new token attends to previous tokens. Combined with ongoing AI-grade server DRAM tightness, that supports the view that even cheaper inference can coexist with stronger overall memory demand if usage, context length, and deployments keep expanding.

Bull case vs. bear case

• Bull case: investors keep treating Micron primarily as an AI-memory supplier, in which case current DDR5 softness looks like a near-term overreaction.
• Bear case: investors focus more on mainstream DRAM pricing and on the possibility that AI efficiency lowers memory intensity per unit of work, which could pressure the multiple before it supports volume.

What to watch after a sharp post-earnings drop

After roughly 30% since the March 18 report, Micron sits in a tricky trading zone: weak enough to test conviction, but not fully washed out. The Wall Street consensus target cited in that coverage was about $527.60, which still leaves the stock under scrutiny rather than obvious capitulation.

Signals that could support a rebound

• Signs that negotiations between Micron and hyperscalers could help stabilize contract pricing.
• Evidence that mainstream DDR5 pricing pressure is stabilizing rather than broadening.
• A return to AI-memory tightness as the dominant narrative, especially if HBM and server-DRAM supply constraints remain in force.

Signals that could reinforce a lower-multiple setup

• A market that starts pricing Citi's view that DRAM and NAND prices peaking in 2Q of next year, reflecting softer pricing momentum into next year.
• A bearish resolution to the efficiency debate, where investors focus on lower cost per query before they are convinced usage growth will fully offset it.
• More widespread spot weakness, which would remind investors that long-term agreements can cushion Micron but not fully isolate it.

For now, the next few sessions look less like a philosophy debate and more like a positioning test: investors have to decide whether Micron is still an AI scarcity story or whether memory pricing and efficiency concerns deserve more weight in the valuation.