Alex Karp Grouped Palantir With These 3 Unstoppable Stocks as the Only True Artificial Intelligence (AI) Infrastructure Winners

Palantir CEO Alex Karp argues only a few AI-infrastructure names matter, highlighting Palantir alongside Nvidia, Micron, and SK Hynix. Palantir’s Rule of 40 (revenue growth + operating margin) reportedly rose from 83% in Q1 2025 to 145% one year later, driven by accelerating commercial adoption of its AIP and expanding operating margin via operating leverage. The article is constructive on the AI stack overall, but notes Palantir’s valuation leaves limited margin of safety if growth moderates.

Analysis

This cluster is not a single AI trade; it is a competition for where incremental AI spend converts into cash most efficiently. NVDA is the cleanest monetization of hyperscaler capex because every added rack still pulls through compute, networking, and software attach, while MU/SKHY benefit from the more durable bottleneck: memory content per accelerator. PLTR’s leverage is real, but it is further downstream and therefore more exposed to enterprise proof-of-value scrutiny; that matters because downstream software tends to lose multiple support faster when growth normalizes.

The near-term catalyst is the next round of hyperscaler capex commentary and memory pricing updates over the next 1-2 earnings cycles. If capex stays firm, NVDA and memory should continue to outperform because the market still underestimates how much of AI demand is being forced into scarce components rather than absorbed by broad software adoption. Over 6-18 months, the bigger structural winner may be memory: as inference scales, memory bandwidth/content often rises faster than raw GPU unit growth, which keeps pricing power and utilization high.

The contrarian risk is that the market is treating Rule-of-40 strength as a permanent moat for PLTR when it is really a lagging signal of current demand intensity. If growth decelerates even modestly, the stock can re-rate sharply because its valuation leaves little room for multiple compression. The main falsifier for the bullish AI infra view is a visible digestion phase: lower hyperscaler capex guides, HBM inventory buildup, or export restrictions that blunt GPU demand within 1-2 quarters.

Key Decisions for Investors

• Long NVDA / short PLTR as a 3-6 month valuation-dispersion pair trade; NVDA is the more direct capex beneficiary while PLTR carries higher multiple compression risk if enterprise adoption slows. Size dollar-neutral and use any PLTR strength to initiate.
• Add MU on pullbacks for a 6-12 month trade, or use call spreads if you want defined risk; thesis is continued HBM tightness and utilization leverage. Falsify on gross margin compression or management signaling memory ASP normalization.
• If SKHY is accessible in size/liquidity, prefer it as a higher-beta expression of the same memory thesis versus PLTR; otherwise MU is the cleaner U.S. vehicle. Use as a tactical long into the next AI capex read-through.
• Watch for a reversal trigger: any hyperscaler capex guide cut or weaker-than-expected HBM commentary. If that appears, trim PLTR first, then reassess NVDA/MU exposure.