5 big analyst AI moves: Micron price targets hiked, cautious on SpaceX valuation

Micron’s fiscal Q3 results beat expectations decisively, with revenue of $41.46B vs $35.91B consensus and adjusted EPS of $25.11 vs $20.86, while Q4 guidance of about $50B revenue and $31 EPS also topped Street forecasts. Susquehanna raised its target to $2,000 and now sees free cash flow above $110B in FY27, reflecting major demand from HBM and strategic customer agreements; Bernstein also lifted memory targets on expected HBM price increases. Separately, Argus initiated SpaceX at Hold after its $1.75T-plus IPO valuation, and Morgan Stanley upgraded Qualcomm on a $5B data center revenue target, reinforcing positive AI-related sentiment despite valuation and execution concerns.

Analysis

The key market implication is that memory has shifted from a cyclical commodity trade to a quasi-contracted utility stream, and that matters more than headline EPS beats. If long-duration supply agreements with floors become the norm, the multiple should migrate upward for the most constrained HBM and advanced packaging nodes while lower-quality memory peers lose pricing leverage; the economic rent moves from buyers to the few suppliers with scale, yield, and qualification depth. That argues for relative outperformance in TSM and AMAT as the toll collectors of the AI buildout, not just the memory names themselves.

The second-order inflation effect is underappreciated: if HBM and adjacent inputs reprice into 2027, hyperscaler capex may not slow in nominal terms, but the same dollar of AI spend will buy less compute. That creates a subtle loser/beneficiary split — GPU leaders can preserve unit demand if they absorb some input inflation, while second-tier accelerator vendors and cloud custom-silicon efforts face a harsher hurdle rate because they lack pricing power and ecosystem lock-in. Qualcomm’s data-center opportunity is therefore more credible than before, but still vulnerable if the market starts valuing AI participation by margin durability rather than revenue aspirations.

The contrarian read is that the market may be too early in extrapolating a straight-line AI supercycle. The defensiveness rotation UBS is making suggests the easiest money in semiconductor beta may already have been captured, and the next leg likely comes from stock selection rather than index exposure. Over the next 1-3 months, any pause in hyperscaler capex commentary or a setback in HBM pricing negotiations would hit the more speculative AI names first, while foundry, equipment, and contracted memory should hold up better.

Key Decisions for Investors

• Go long TSM vs short a basket of lower-quality AI hardware names for 1-3 months; thesis is that contracted, high-yielding supply-chain toll collectors will outperform as memory pricing stays elevated and capex remains intact.
• Initiate a tactical long AMAT position into any 2-4 week pullback; risk/reward favors the equipment names with pricing power and backlog visibility if AI capex gets re-ranked rather than cut.
• Hold a small long QCOM / short INTC pair for 3-6 months; Qualcomm has a credible AI monetization path, while Intel remains more exposed to execution risk and a slower re-rating if AI spend broadens unevenly.
• Fade aggressive long-beta exposure in NVDA on strength via call overwrites or capped upside structures for the next earnings cycle; the risk is not demand collapse but margin compression if input costs rise faster than end-customer willingness to pay.
• Use any drawdown in SK Hynix or MU to add only on weakness, not momentum, because the market is likely to overpay for the first innings of contractual earnings visibility but underprice the eventual supply response over 6-12 months.