The $600 Billion Question Hanging Over Big Tech This Earnings Season

Big Tech’s 2026 capex is tracking to roughly $600B across Amazon, Microsoft, Meta, and Alphabet, with investors focused on whether AI spend is translating into monetization. Amazon has guided to about $200B in capex, Alphabet to $175B-$185B, and Meta to $115B-$135B, while Microsoft spent $29.88B in Q2 FY26 alone. The article highlights strong backlog metrics—Microsoft commercial RPO up 110% to $625B and Google Cloud backlog at $155B—but also warns that weak proof of return could pressure margins and stocks.

Analysis

The market is no longer debating whether AI capex exists; it is debating who can convert it into operating leverage fast enough to justify the depreciation drag. The near-term winners are the vendors closest to deployed workloads and recurring inference demand, while the losers are the platforms that are still converting backlog into visible revenue too slowly relative to spend. That creates a second-order risk to the entire semiconductor and networking stack: if one hyperscaler blinks on growth, procurement can shift from broad-based expansion to utilization optimization, which typically hits the highest-beta hardware names first.

The key catalyst is not the earnings print itself but the commentary on payback period. If backlog growth is not accompanied by a sustained step-up in cloud growth rates over the next 2-3 quarters, the market will start treating capex as front-loaded expense rather than strategic moat-building. That is especially dangerous for the name with the weakest current cash conversion, because falling free cash flow tends to force a choice between defending growth and defending shareholder returns.

The consensus is probably underestimating how asymmetric the reaction function is. Good numbers likely get faded because investors already own the optimism, while even a modest miss on monetization can trigger multiple compression across the group. The cleaner contrarian setup is that the strongest backlog visibility may actually belong to the stock with the worst recent price action, making it the best relative-value long if management can show credible acceleration; otherwise, leadership shifts toward the names with the most explicit monetization metrics and the least narrative risk.

Time horizon matters: this is a days-to-weeks catalyst around earnings, but the real allocation decision is over months, as depreciation and margin pressure catch up with spend. If the market starts to price in a 2027 normalization of capex intensity without commensurate revenue yield, the current AI infrastructure trade can quickly rotate from growth premium to value trap.

Key Decisions for Investors

• Maintain a tactical long MSFT / short AMZN pair into earnings and through the print if you want to express relative monetization quality; the trade works if the market rewards proof of backlog-to-revenue conversion over raw capex scale. Risk: AMZN shows accelerating AWS reacceleration, which would compress the spread quickly.
• Buy near-dated AMZN puts or a put spread into the event if implied volatility remains below the likely post-earnings move. This is a defined-risk way to express concern that capex intensity is outrunning free-cash-flow conversion; stop if management gives explicit guidance tying spend to revenue inflection within 1-2 quarters.
• Go long GOOGL vs a basket of the weaker AI hardware beneficiaries for 4-8 weeks if cloud backlog and monetization commentary remain robust. This captures the market’s preference for visible demand while avoiding the name most exposed to incremental capex skepticism.
• Reduce exposure to the highest-beta AI infrastructure suppliers for 2-3 weeks if the hyperscaler commentary is mixed, because any hint of capex optimization tends to flow first into ordering discipline. A modest disappointment can trigger 10-15% downside in the supply chain even when the hyperscalers themselves hold up.
• For investors already long the complex, hedge with short-dated QQQ calls or a small SPY put spread around the earnings window. The goal is to offset a correlation spike if the market interprets one weak read-through as evidence that the whole AI spend cycle is entering payback mode.