3 Stocks to Take Advantage of $1 Trillion in 2027 Capital Expenditures

The article argues that AI data-center spending could reach $1 trillion next year, with Nvidia, Broadcom, and Sandisk positioned to benefit. It cites Wall Street expecting Nvidia revenue growth of 81% in fiscal 2027, Broadcom revenue growth of 66% in fiscal 2026 and 62% in fiscal 2027, and Sandisk revenue growth of 336% in fiscal Q4 2026 and 122% in fiscal 2027. The piece is largely bullish commentary rather than fresh company news, but it reinforces strong demand expectations for AI semiconductors and storage.

Analysis

The market is still treating AI capex as a straight-line winner-takes-all story, but the second-order profit pool is moving down the stack. The more hyperscalers standardize around custom silicon, the more value leaks away from merchant GPU pricing power and into the enabling layers: advanced packaging, high-bandwidth memory, networking, optics, and especially storage. That means the next leg of this trade is less about who sells the most compute and more about who is constrained by supply, because shortages preserve pricing and convert revenue growth into margin expansion.

The key timing issue is that the big impulse is likely not immediate. A lot of the demand being priced today is a 2027 story, which creates a window where sentiment can outrun realized orders. If AI capex gets pushed out even one budget cycle, the highest-multiple names will de-rate first, while the tightest-supply beneficiaries can keep compounding because customers cannot defer storage and interconnect indefinitely once racks are commissioned. That makes the trade more favorable in the infrastructure bottlenecks than in the headline AI platform names.

Consensus is probably underestimating how much custom silicon can cannibalize the economics of the incumbent compute leader over time. Broadcom’s upside is not just incremental share; it is that each hyperscaler designed chip reduces dependence on off-the-shelf accelerators and increases switching costs for the customer’s own software stack. The risk is that the market is extrapolating every AI buyer as a perpetual spender when in reality the first companies to optimize cost per token will slow unit growth fastest, forcing a valuation reset across the broader AI complex.

Key Decisions for Investors

• Long SNDK on 3-6 month horizon; best risk/reward if AI capex keeps tightening the SSD market, with upside driven by price rather than unit growth. Use trailing stop if gross margin expansion stalls for two quarters.
• Maintain/initiates long AVGO vs short NVDA pair for 6-12 months; thesis is mix shift from merchant accelerators to custom silicon and greater earnings durability. Size modestly because NVDA can stay momentum-driven longer than fundamentals imply.
• Buy NVDA downside protection via 6-9 month put spreads rather than outright shorting; the risk is not demand collapse but multiple compression if 2027 growth expectations get pushed out.
• Add exposure to AI infrastructure beneficiaries beyond the obvious names — focus on data-path bottlenecks such as networking/optical suppliers — because they are better insulated from chip architecture changes than compute vendors.