Warren Buffett Bought Alphabet Stock About a Year Ago. 3 Reasons It’s Still a Buy Today

Alphabet reported Q2 revenue of $119.8 billion, up 24% year over year, while Google Cloud revenue rose 82% to $24.8 billion and its backlog reached $514 billion. The company raised 2026 capital-expenditure guidance to $195 billion-$205 billion from $180 billion-$190 billion, signaling continued AI-infrastructure expansion. A federal judge also rejected the U.S. government's push to force a sale of AdX, reducing a major antitrust-remedy risk, although additional legal cases remain.

Analysis

The investable issue is whether incremental AI infrastructure spending converts into durable Cloud operating income rather than a lower-return capacity race. GOOG’s valuation can support multiple expansion only if backlog conversion lifts cloud margins faster than depreciation and power/network costs; otherwise, elevated capex raises free-cash-flow sensitivity and leaves the stock exposed to any enterprise AI-demand digestion. The article’s reported backlog and growth figures should be independently verified against filings before they are incorporated into estimates.

Competitive dynamics favor GOOG where its proprietary models, TPUs and data stack lower customer inference cost, but the same vertical integration is a second-order headwind for merchant accelerator demand at the margin. That is more relevant to NVDA than the headline suggests: sustained Google TPU deployment would not impair NVIDIA’s near-term revenue, but could cap the upside case embedded in hyperscaler GPU procurement over the next 6-18 months. A higher-capex equilibrium also benefits data-center power, networking and cooling suppliers more reliably than it benefits cloud equity holders, whose returns depend on utilization.

The regulatory outcome is not a clean all-clear. Restrictions on default distribution, data use, ad-tech interoperability, or AI product bundling could reduce Search’s traffic-acquisition advantage and raise monetization friction without requiring a breakup. The market may be underpricing this slower, margin-dilutive remedy risk because it is less dramatic than a divestiture; the key 1-3 month catalyst is remedy detail and any change in management’s traffic-acquisition-cost or advertising-margin outlook.

Waymo should be treated as embedded optionality, not a near-term valuation driver. Its strategic value rises with demonstrated unit economics and utilization, while TSLA’s autonomous-driving narrative faces a relative disadvantage if commercial robotaxi deployment remains concentrated in a better-capitalized, safety-validated competitor. Conversely, a broad regulatory or safety setback would compress autonomy optionality across both names.

Key Decisions for Investors

• Maintain or initiate a 6-12 month long GOOG position only on evidence that Cloud operating-margin expansion and backlog conversion offset the capex step-up; size as a core quality exposure, with thesis invalidated by two consecutive quarters of decelerating Cloud growth plus worsening consolidated free-cash-flow conversion.
• Use a 3-6 month pair: long GOOG / short TSLA in equal beta-adjusted dollars. The trade isolates enterprise-AI monetization and commercial autonomy execution against a more expectation-sensitive autonomy valuation; cover the short if TSLA demonstrates sustained paid autonomous-ride utilization or GOOG reports material Waymo losses without utilization progress.
• Do not chase NVDA solely on hyperscaler capex headlines. Set an alert for evidence of materially rising TPU/internal-silicon share in Google’s AI fleet or reduced external accelerator procurement commentary; that would support a tactical NVDA underweight versus broader AI infrastructure suppliers over 6-18 months.
• Ahead of regulatory remedy milestones, consider trimming GOOG delta or owning 3-6 month downside puts if implied volatility is below its litigation-event range. The risk/reward improves only if option premium does not already price a meaningful adverse remedy; a benign remedy with no distribution or data-use restrictions is the clear falsifier.