Agentic AI will drive Micron to new heights, says Baird

Baird raised Micron's price target to $1,520 from $1,280, implying 40% upside from Friday's close, citing accelerating agentic-AI demand, slower 2027 industry DRAM supply-bit growth and higher-margin HBM sales. Micron shares have already gained 279% in 2026 amid a memory-chip shortage, while consensus expects fiscal Q4 earnings to rise 939% year over year. Of 49 analysts tracked by LSEG, 46 rate Micron buy or strong buy, with the average target implying 34% upside.

Analysis

MU’s upside now depends less on aggregate DRAM pricing and more on whether HBM mix can lift through-cycle gross margin without sacrificing conventional-server DRAM supply. The key earnings sensitivity is HBM revenue conversion: every 500bp improvement in mix and yield can matter more to EPS than a modest spot-price move, while constrained supply also gives MU leverage to contract repricing. Samsung Electronics and SK Hynix remain the relevant competitive risks; a faster-than-expected qualification ramp at either could narrow HBM premiums before the broader DRAM cycle rolls over.

The immediate catalyst is guidance quality rather than the reported quarter. With positioning and sell-side expectations already unusually one-sided, a beat that merely validates current demand could produce a sell-the-news reaction over days; sustained upside requires raised calendar-2027 supply/demand assumptions, evidence of multi-quarter HBM allocation visibility, and gross-margin guidance above consensus. Conversely, any indication that customers are double-ordering memory ahead of platform transitions, or that CPU/AI-server demand is converting more slowly into memory content, would expose MU to rapid multiple compression even if near-term earnings remain strong.

Over 6-18 months, the more important contrarian issue is supply discipline. Memory upcycles historically self-correct when elevated returns induce capex, but advanced-node DRAM and HBM capacity are harder to add than commodity bits; this can extend the cycle if technology transitions constrain effective output. The consensus may still underappreciate that risk-adjusted returns improve if industry supply remains rational, but it also appears to underprice a qualification or yield setback: HBM is not a fungible commodity, and a delayed product ramp would have disproportionate margin consequences.

Key Decisions for Investors

• Maintain a tactical long MU into earnings only via defined-risk exposure: buy a 3-6 month call spread, using approximately 105%/125% strikes versus spot. This captures a guidance-driven re-rating while limiting downside from a crowded-expectations reset; target 2:1 or better payoff-to-premium.
• For existing cash longs, trim 20-30% before the release and re-add only if management raises forward HBM margin or supply guidance rather than merely reporting a beat. Falsifier: forward gross-margin guidance below consensus or evidence that HBM qualification/yields are slipping.
• Express the structural memory view as long MU / short WDC on a 3-6 month horizon. MU has greater exposure to high-value DRAM/HBM margin expansion, whereas WDC is more exposed to NAND’s historically less-disciplined supply cycle; exit if NAND pricing materially outperforms DRAM or MU’s HBM mix misses expectations.
• Set a post-earnings alert for a 10-15% MU decline without a reduction in forward revenue, HBM allocation, or gross-margin guidance. That would be a higher-quality entry than chasing a headline beat, given the elevated consensus and valuation sensitivity.