Nvidia Is Cooling, Microsoft Is Stumbling, and the Fed Won't Cut. Here's Why I'm Still Buying Artificial Intelligence (AI) Stocks.

The article argues that recent pullbacks in leading AI stocks such as Nvidia and Microsoft create a buying opportunity for long-term investors. It highlights the shift from AI training to inference, benefiting companies like Broadcom and Intel, and points to a projected 30.6% CAGR for the global AI market from 2026 to 2033. The piece is broadly bullish on the AI sector, but it is opinion-driven rather than a new company-specific catalyst.

Analysis

The market is likely mispricing the AI stack as a single trade when the next leg is probably a dispersion trade. Training spend is still the headline, but inference is the more durable revenue pool: it creates a recurring demand curve for networking, custom silicon, edge CPUs, optical interconnect, racks, and power, which should compress the valuation premium of pure GPU exposure over the next 6-18 months as customers optimize total cost per token.

That has second-order winners beyond the obvious beneficiaries. Broadcom and Intel are better positioned than sentiment suggests because inference economics favor custom silicon and general-purpose CPUs, while the physical bottlenecks increasingly migrate to optics, datacenter real estate, and electrons. The underappreciated risk is that the next AI capex cycle shifts from compute scarcity to infrastructure scarcity, which can cap hyperscaler ROI even if model demand keeps rising.

The contrarian view is that the recent weakness in megacap AI is less a secular warning than a valuation reset driven by rates and positioning. If macro stays sticky, higher discount rates hurt long-duration software and mega-cap platform names more than picks-and-shovels suppliers with shorter payback and clearer unit economics. The key catalyst is not a macro pivot; it is evidence that inference workloads are scaling faster than training budgets, which would re-rate the supply chain over the next 1-2 reporting cycles.

Tail risk is that enterprise AI monetization remains slower than infrastructure buildout, creating a temporary glut in capacity and a shakeout in cloud-infrastructure leasing names. In that scenario, higher-beta entrants and less-capitalized providers are most vulnerable. The best trade setup is to own the constrained inputs and avoid the assets whose margins depend on flawless utilization assumptions.

Key Decisions for Investors

• Long AVGO vs short NVDA for 3-6 months: express the view that inference/custom silicon monetization will outgrow pure training GPU spend; target ~15-20% relative outperformance if capex mix continues to shift.
• Buy INTC on weakness for a 6-12 month horizon: this is a mean-reversion trade on inference-driven CPU demand and share gains in non-GPU workloads; keep size modest because execution risk remains high.
• Long LITE/GLW basket vs short CRWV/NBIS for 3-9 months: favor infrastructure toll collectors over capital-intensive AI cloud capacity providers if hyperscaler demand stays strong but pricing power remains uneven.
• Initiate a small long BEPC as a 12-month hedge against AI power scarcity: renewable power contracting is a cleaner way to express datacenter load growth than chasing crowded software names.
• Trim or avoid incremental MSFT exposure until the next two earnings prints: the setup is still vulnerable to multiple compression if AI monetization lags the pace of product rollout; use any rebound to reduce exposure.