Jim Cramer says two stocks will decide stock market’s next move

SanDisk reported fiscal Q3 2026 revenue of $5.95 billion, up 97% sequentially, with datacenter revenue surging 233% and gross margins moving toward 80%; it also authorized a new buyback. Oracle posted fiscal Q3 2026 revenue of $17.2 billion, cloud infrastructure revenue up 84% to $4.9 billion, and a record $553 billion in remaining performance obligations, while raising FY27 revenue guidance to $90 billion. The article frames both companies as AI infrastructure bottlenecks, signaling sustained demand and supply constraints across the tech stack.

Analysis

The key read-through is that AI capex is no longer a single-threaded Nvidia story; it is becoming a bottlenecked systems trade where memory and cloud buildout are the gating variables. That creates a more durable earnings backdrop for the infrastructure layer than for application-layer AI names, because the scarcity is showing up in supply contracts, pricing power, and backlog conversion rather than in speculative usage metrics.

Second-order winners sit upstream of the bottlenecks. DRAM/NAND peers and equipment vendors should see a favorable mix of tight supply, better ASP discipline, and multi-quarter visibility; the risk is that the market overpays for the near-term scarcity and misses how fast memory cycles can normalize once capacity comes back. On the cloud side, Oracle’s backlog implies long-duration demand, but the real margin lever is execution speed on data center power, networking, and GPU procurement — any delay pushes revenue recognition out, not necessarily away.

The main contrarian setup is that both names may already be discounting a lot of perfection. SanDisk’s move looks closer to a cyclical peak in pricing power than a clean straight-line comp story, while Oracle’s backlog is impressive but partly a timing asset: backlog quality, customer concentration, and deployment cadence will matter more than headline size over the next 2-4 quarters. If supply starts to loosen in memory or if AI spend shifts from buildout to optimization, the multiple expansion could compress quickly even if fundamentals stay strong.

From a market-structure perspective, this is a relative-value signal more than an outright beta call. The cleanest expression is long infrastructure scarcity winners versus short the downstream margin takers who lack supply contracts and cannot reprice quickly; that trade should work over the next 3-6 months if pricing remains tight. But if investors chase the theme indiscriminately, the better risk/reward may come from call spreads or pair structures rather than naked longs.

Key Decisions for Investors

• Long ORCL into the next 1-2 quarterly prints; use a 3-6 month horizon and size for backlog conversion rather than headline backlog growth. Risk/reward is favorable if cloud growth keeps compounding above the current guide, but trim if delivery cadence slips.
• Long SNDK on pullbacks, but treat it as a 3-6 month tactical trade, not a core compounder at current momentum. Best entry is after any post-earnings digestion; downside risk rises sharply if NAND ASPs or contract terms soften.
• Pair trade: long ORCL / short a basket of cloud and enterprise buyers exposed to memory and AI infrastructure inflation. The thesis is that suppliers with contracted demand outperform margin-sensitive demand users over the next 1-2 quarters.
• Pair trade: long SNDK / short a downstream hardware assembler or PC/smartphone exposure with weak pricing power. This captures the spread between constrained input sellers and buyers forced to absorb higher component costs over the next 6 months.
• Use call spreads rather than outright longs if chasing the AI infrastructure theme; implied upside is already elevated, so defined-risk structures offer better reward-to-premium if the trade stalls after the next earnings cycle.