Analysts Expect AI Capex to Reach $800 Billion in 2026. That Number Could Actually Surpass $1 Trillion. Here Are 2 Stocks That Will Benefit.

AI capex expectations are rising: consensus calls for about $800B in 2026 AI infrastructure spending, but Goldman Sachs estimates the true figure is closer to $1T, citing undercounted investment outside the U.S. and by private AI firms (global/private +$200B vs expectations). McKinsey projects AI capex to reach ~$7T by 2030, supporting the broader build-out narrative. The article highlights Nvidia’s international expansion and suggests a “cheaper” AI exposure via Qualcomm, though it flags geopolitical and core-business stagnation risks.

Analysis

The market is still treating AI buildout like a single-stock story, but the better lens is a multi-year industrial bottleneck trade. The first-order winner remains NVDA, yet the second-order winners are the companies solving adjacent constraints — networking, power, cooling, and rack density — because those are the parts of the stack where pricing power can persist even if chip ASPs eventually normalize. That argues for staying long the infrastructure layer rather than chasing application-layer AI names that still need proof of monetization.

The key near-term catalyst is not the capex headline itself but the next round of hyperscaler and enterprise guidance over the next 1-3 months. If reported spend continues to outrun consensus, the biggest upside surprise should be in suppliers with long backlogs and short product cycles; if spend is only re-phased rather than expanded, the trade will rotate from semis into power and data-center infrastructure. QCOM is a more speculative, lower-quality AI proxy: the edge-AI optionality is real over 6-18 months, but the stock still has to overcome handset cyclicality and China exposure before that narrative matters.

The contrarian miss is that capex can stay elevated while equity returns disappoint if utilization and revenue conversion lag. In other words, the bullish case for the picks-and-shovels group is stronger than the bullish case for the broad AI basket. What would falsify the thesis is a clear slowdown in customer capex guides, a material gross-margin reset at NVDA, or signs that international demand is being constrained by export controls, power availability, or procurement delays rather than expanding organically.

Key Decisions for Investors

• Go long NVDA on pullbacks and use 3-6 month call spreads rather than outright common; the setup is strongest into the next earnings/capex guide cycle, but reduce exposure if hyperscaler spend decelerates sequentially.
• Express the relative view with a long NVDA / short IGV or XLK pair for 1-3 months; AI capex should benefit infrastructure suppliers faster than software names that still lack monetization proof.
• Take a smaller tactical long in QCOM as a 6-18 month edge-AI optionality play, but size it below NVDA because China/handset risk can swamp the AI premium; cut if core handset guidance weakens further.
• Add a watchlist for power/cooling/networking beneficiaries such as VRT, ETN, and ANET; if backlog commentary improves again, these may offer better risk/reward than the headline semis with less valuation crowding.