Zuckerberg Just Gave Away Why You Should Buy Micron, Not Meta

Meta said it is increasing 2026 infrastructure capex, with most of the increase tied to higher memory pricing, underscoring rising AI data center costs across hyperscalers. The article argues that this spend is a tailwind for Micron, which it says trades at 9x forward earnings despite 118% annual EPS growth and 68% annual revenue growth. It also cites a recent UBS target increase to $1,625 from $535, while flagging that Wall Street’s average target of $633 implies about 30% downside versus current levels.

Analysis

The key second-order trade is not simply “AI capex up,” but that the bottleneck is shifting from compute silicon to memory and the broader component stack. That changes who captures incremental margin: memory vendors with tight supply discipline benefit more than the hyperscalers do, because pricing can re-rate faster than end-demand and with less customer concentration risk. If buildouts stay front-loaded into 2H26, the earnings revision cycle for memory can keep extending even if hyperscaler capex growth slows on optics.

META is the cleaner short only on a relative basis, not as a standalone outright short. The market may be underappreciating how much incremental capex inflation leaks into free cash flow and balance-sheet flexibility, but the shares can still work if ad pricing and AI product monetization offset the spend. The more attractive expression is to own the beneficiaries of the spending wave rather than fade the spenders, because the latter have multiple offsetting levers while the former have direct operating leverage.

The contrarian risk on the long memory trade is that the market is already paying for a “supercycle” and any inventory digestion or pricing moderation will hit high-beta names hard. The timing matters: over the next 1-3 quarters, a single hyperscaler spending pause or lead-time normalization could compress multiples even if absolute earnings remain strong. That said, the asymmetry is still favorable while management teams are openly talking about component inflation, because that usually extends procurement urgency and supports spot pricing.

UBS’s target increase is a useful sentiment tell, but consensus is still likely lagging the fundamental curve by one to two quarters. The more important signal is that AI capex is now self-reinforcing: higher component costs push spend higher, which tightens supply further, which keeps pricing elevated. In that regime, valuation matters less than duration of scarcity, and the right question is how long hyperscalers can absorb inflation before slowing the buildout.

Key Decisions for Investors

• Long MU vs short META in a 1-3 month pair trade; thesis is direct capture of capex inflation by memory vs margin drag on the spender. Use a 1:1 dollar-neutral structure and take profits on any 15-20% relative move.
• Add to NVDA on pullbacks over the next 2-6 weeks; it remains the cleanest beneficiary of AI infrastructure spend with better balance-sheet optionality than the hyperscalers. Risk/reward improves if the market rotates from software monetization back to hardware scarcity.
• Initiate a small AMD/MU basket long against META if you want diversified AI-hardware exposure; this reduces single-name memory-cycle risk while keeping exposure to hyperscaler spend. Best entry is on post-earnings weakness or broad tech selloffs.
• Avoid chasing META strength until capex guidance stabilizes for at least one quarter; if you must own it, prefer call spreads rather than stock to cap downside from rising depreciation and component inflation.
• Set a tactical exit on MU if forward pricing or consensus EPS revisions start flattening; that would be the first sign the supercycle is maturing and the multiple could compress even with good absolute earnings.