Plurimi CIO Sees Value in AI Supply Chain

Plurimi Wealth CIO Patrick Armstrong said he sees value in the AI supply chain, favoring stocks such as Micron, Samsung and Nvidia. He said the firm is underweight the Magnificent Seven hyperscalers and prefers companies receiving the capital being spent on AI datacenters. The comments are constructive for AI infrastructure and component suppliers, but they are commentary rather than a new company-specific catalyst.

Analysis

The market is still pricing AI as a single trade, but the better risk/reward is shifting one layer down the stack. If hyperscalers keep front-loading capex, the economics accrue to the component vendors and picks-and-shovels suppliers first; the buyers are converting balance sheet cash into an installed base that may take multiple years to monetize, while the sellers of memory, accelerators, networking, and advanced packaging capture revenue now. That makes the supply chain more insulated from any near-term disappointment in AI monetization than the software/platform cohort, which needs continued valuation support from future usage growth.

The second-order dynamic is that capex intensity itself can become a constraint on the leaders. The hyperscalers can still fund spending, but the market will eventually force a higher bar for free-cash-flow conversion and ROIC, especially if AI demand remains concentrated in training rather than broad inference adoption. If that happens, the market will likely rotate from “who can spend the most” to “who can supply the bottlenecks,” which is constructive for semis with tight capacity and pricing power, and negative for the mega-cap platforms if capital intensity keeps rising faster than incremental revenue.

The contrarian risk is that this is already crowded positioning in the obvious beneficiaries, particularly Nvidia. Near term, any digestion in AI order growth, lead-time normalization, or evidence that customers are over-ordering to secure allocation could compress multiples faster than fundamentals deteriorate. Over a 3-12 month horizon, the key reversal catalyst would be a capex pause from one or more hyperscalers, which would hit the supply chain first on sentiment but also remove the strongest argument for staying underweight the platform names.

Bottom line: the trade is not just long AI, it is long AI scarcity. The more constrained the ecosystem remains, the more pricing power migrates to memory, compute, interconnect, and packaging — and the more vulnerable the hyperscalers become to a future capex hangover if payback periods lengthen.

Key Decisions for Investors

• Overweight NVDA on a 3-6 month horizon, but prefer entries on pullbacks; the stock still benefits from supply bottlenecks and pricing power, though upside is more asymmetrical if hyperscaler capex stays elevated. Use a staggered entry rather than chasing strength after each capex headline.
• Pair long NVDA vs short a basket of megacap hyperscalers over 6-12 months; the thesis is that suppliers capture the near-term economics while platform ROIC becomes the eventual debate. Keep the short leg small initially because the trade can work against you if AI revenue inflects faster than expected.
• Add exposure to the broader AI supply chain through memory/packaging/networking names on any weakness; these tend to have more direct leverage to capex intensity and less valuation risk than the platform layer. Risk/reward is strongest where capacity is still tight and order visibility extends 2-3 quarters.
• Consider call spreads on NVDA rather than outright calls if implied volatility is elevated; this preserves upside to continued capex momentum while reducing premium decay if the market digests crowded positioning. Best used on post-rally consolidation.
• Set a tactical risk trigger: if hyperscaler capex commentary turns cautious for two consecutive earnings cycles, rotate out of the supply-chain beta and into the platform leaders. That would likely mark the first real challenge to the “selling shovels” trade.