Do Amazon and Alphabet's Combined $420 Billion in Capital Expenditures Make Sense?

Amazon projects roughly $220 billion of 2026 spending and Alphabet forecasts $195 billion-$205 billion, primarily to expand AI and cloud-computing capacity. AWS revenue growth accelerated to 37% year over year in Q2, while Google Cloud revenue rose 82% to $24.8 billion, supporting the view that the capex cycle is justified by demand. The article argues both stocks remain attractive despite forward 2027 P/E ratios of about 25x for Amazon and 23x for Alphabet.

Analysis

The investable question is not whether hyperscaler demand exists, but whether incremental capacity earns above the rapidly rising cost of capital. AMZN and GOOGL can support elevated capex while cloud utilization remains supply-constrained, but the equity rerating requires backlog conversion to revenue and stable operating margins rather than another upward capex revision. Over the next 1-3 months, cloud-bookings commentary, lease commitments, and depreciation guidance matter more than reported cloud growth; each is a leading indicator of whether 2027 earnings estimates need to rise or free-cash-flow expectations need to fall.

GOOGL has the cleaner relative setup if custom silicon adoption shifts workloads away from premium GPU configurations: internal TPU utilization can lower its unit economics while external accelerator sales create a new monetization vector. That is a modest negative at the margin for NVDA only if TPUs become portable across a broader software ecosystem; in the nearer term, it is more likely to pressure alternative accelerator vendors and raise the bar for AMD's data-center share gains. AMZN's risk is more balance-sheet and execution oriented: its larger fulfillment, logistics and cloud investment base makes consolidated FCF more sensitive to construction delays, power constraints, and a slower-than-expected enterprise AI deployment cycle.

Consensus is treating capacity scarcity as durable, yet the eventual risk is synchronized oversupply: AMZN, GOOGL, MSFT and META are all building against similar 2027-28 demand assumptions. A 6-18 month normalization in GPU availability or enterprise AI ROI would compress cloud growth and leave depreciation as a fixed margin headwind, particularly for AMZN. The bullish case is falsified if AWS growth decelerates for two consecutive quarters while capex remains elevated, or if GOOGL's cloud margin fails to expand despite accelerating revenue—evidence that pricing is being competed away rather than capacity being monetized.

Key Decisions for Investors

• Prefer long GOOGL / short AMZN over a 3-6 month horizon. GOOGL has greater operating leverage from cloud mix and potential silicon economics, while AMZN carries higher consolidated FCF sensitivity to capital intensity; reassess if AWS growth exceeds Google Cloud growth for two quarters or AMZN materially raises cloud-margin guidance.
• Maintain a measured long GOOGL rather than chase broad AI beta: add on any post-results decline driven by capex headline risk if Cloud revenue growth and operating margin both remain above prior-quarter levels. Target is estimate-driven multiple expansion; stop/reduce on a material cut to 2027 FCF estimates or evidence TPU sales are cannibalizing, rather than supplementing, high-margin cloud consumption.
• Use NVDA as a hedge/watch rather than an outright short. Consider reducing NVDA exposure only if independent customer disclosures show meaningful TPU workload migration plus a decline in GPU unit pricing or lead times; absent those data, hyperscaler capex remains supportive of accelerator demand despite in-house silicon narratives.
• Set a sector alert around quarterly depreciation growth, data-center lease obligations, and power-delivery delays at AMZN/GOOGL/MSFT/META. If these rise faster than cloud revenue for two reporting periods, rotate from hyperscalers into power and cooling beneficiaries such as VRT and ETN, where backlog conversion is less exposed to end-customer cloud pricing.