SpaceX Lost $2 Billion in the First Half of 2026. Micron Made $28 Billion in Profits in Its Latest Quarter. But the Better Stock Pick Isn't as Obvious as Those Numbers Suggest.

SpaceX reported a $541M net loss in Q2 and more than $2B loss in H1 2026, alongside 92% YoY revenue growth to $7.8B, while Micron posted $28.2B profit in fiscal 2026 Q3 on $41.5B revenue. Despite SpaceX’s steep valuation (73.4x trailing sales) versus Micron’s low forward multiple (5.3x), Wall Street targets show higher near-term upside for SpaceX (>$90%) than Micron (~65%), though the article favors Micron for the next 2–3 years due to more favorable risk-reward in the current memory upcycle and concerns that AI data center buildout could slow.

Analysis

The market is still overpaying for optionality and underweighting cash conversion. For MU, the key mechanism is that memory is one of the few AI-adjacent businesses where incremental pricing can still flow almost directly into FCF and buybacks; that makes the current multiple vulnerable to upside revisions for as long as hyperscaler capex stays intact. For SPCX, the valuation is much more rate- and supply-sensitive than the narrative suggests: once private-market supply hits via lockups/secondary sales, the “story premium” can compress quickly even if revenue keeps growing.

Second-order effects matter more than the headline comparison. If satellite broadband scales, the pressure on T/VZ is not an overnight revenue hit but a slow ARPU and rural-share bleed over 6-18 months, which keeps telecom multiples capped and leverage more painful. For semis, the real risk to MU is not broad AI enthusiasm fading; it is a digestion phase in data-center spending that first shows up as memory order slippage, then inventory build, then abrupt multiple compression across the whole complex.

Contrarianly, the consensus may be too eager to treat SPCX as a better growth asset simply because the end market is exciting. In public markets, what matters is how much of that growth can be monetized without repeated dilution, secondary supply, or balance-sheet stress. MU looks better on a 1-3 year horizon because it already converts demand into profits, but that thesis breaks if DRAM/NAND pricing flattens for two quarters or hyperscaler guidance cuts capex; that is the main falsifier to watch.

Key Decisions for Investors

• Long MU on any 5-7% pullback over the next 2-4 weeks; target 3-6 months as estimates likely keep drifting up while the multiple stays compressed. Falsifier: two straight quarters of guide-downs or visible inventory days rising.
• Pair trade: long MU / short NVDA in equal dollars for 1-2 quarters to isolate AI memory scarcity versus high-multiple AI capex exposure. Risk/reward: best if AI spending remains solid but the market de-rates duration assets.
• Avoid chasing SPCX into lockup/supply events; if access exists via private secondary, wait for post-lockup discounts rather than paying headline marks. This is a supply-overhang trade, not a fundamentals-only buy.
• Use any telecom strength to add a small structural short or put-spread hedge in VZ or T over 6-18 months. Thesis: satellite substitution is slow but persistent, and leverage limits multiple expansion.