The key takeaways from a massive day of tech earnings

Big Tech continues to raise AI spending, with Microsoft planning $190B in capital expenditures this year and Amazon previously guiding to $200B. Amazon posted its best AWS growth since 2022, with sales up 28%, while Google said AI plans helped drive subscription demand and Meta and Microsoft both signaled continued AI-led investment. The offset is near-term margin and cash flow pressure, including Amazon's 95% drop in free cash flow and Microsoft's expectation that headcount will decline year over year.

Analysis

The market is starting to treat AI capex as a second-order balance-sheet arms race rather than an ROI story. Near term, that favors the infrastructure stack: hyperscale cloud, networking, power, and custom silicon vendors should keep taking share as the big platforms pull spend forward, but the spend itself is becoming less differentiating and more necessary just to defend position. That makes the likely winners the toll collectors upstream of compute demand, while the marginal loser is free cash flow quality at the platforms if revenue monetization lags by even one budgeting cycle.

The more interesting implication is margin compression through internalization. If Amazon eventually commercializes its in-house chips externally, that is a direct signal that custom silicon is moving from cost optimization to a semi-fungible product category, which can pressure third-party accelerator pricing and reduce the moat of pure-play hardware vendors over 12-24 months. At the same time, Google’s subscription momentum suggests AI monetization is starting to show up in consumer ARPU before ad monetization does, which is important because it provides a cleaner path to offset capex than Meta or Amazon currently have.

The labor comments are a bigger tell than the capex numbers: management is telegraphing that AI should raise output per employee, not just revenue per user. That means operating leverage may come later than bulls expect, because companies will likely reinvest productivity gains back into model training and product expansion instead of immediately widening margins. For the broad market, this keeps a lid on near-term earnings revisions even as top-line growth improves, which argues for selective exposure rather than chasing the entire mega-cap AI basket.

Contrarian take: consensus is underestimating the duration of the capex cycle and overestimating how quickly it translates into free cash flow. If AI demand is real but monetization is delayed, the next 2-3 quarters could produce periodic drawdowns in the most expensive names whenever investors start pricing in slower payback. That creates a better setup for buying on weakness than for initiating fresh outright longs after earnings gaps.

Key Decisions for Investors

• Long MSFT vs. short a basket of AI-capex laggards for 3-6 months; MSFT has the cleanest enterprise monetization path and can absorb headcount reduction without sacrificing growth, while the short basket should include names with weaker cash conversion and higher execution risk.
• Buy AMZN on 5-8% post-earnings pullbacks over the next 1-2 quarters; the market is focusing on free cash flow stress, but AWS growth acceleration plus eventual Trainium externalization creates a 12-24 month optionality runway. Tighten risk if AWS growth decelerates below high-teens.
• Pair long GOOGL / short META for 6-12 months; Google is showing nearer-term monetization from AI-adjacent subscriptions, while Meta’s AI spending is still more of a cost narrative. The pair benefits if investors keep rewarding visible cash-flow support over open-ended capex.
• Watch semiconductor and networking suppliers for a tactical long into the next earnings cycle, especially on weakness; the second-order beneficiary is the infrastructure layer, not the platforms. Use 3-6 month horizons and take profits quickly if hyperscaler capex guides flatten.
• Avoid chasing the highest-multiple AI platform names into strength; sell covered calls or trim into post-earnings rallies where capex guide upgrades outpace monetization evidence, because the risk is a 1-2 quarter FCF disappointment rather than a business-model break.