Prediction: Nvidia Will Outperform Alphabet and Amazon Combined Over the Next 3 Years

The article argues Nvidia still has substantial upside, citing AI infrastructure spending that could rise from $600 billion in 2025 to $3 trillion-$4 trillion annually by 2030. It highlights Alphabet's expectation that 2027 capex will significantly exceed 2026 and says Nvidia's GPUs remain a top choice for AI workloads. The piece is broadly bullish on Nvidia, while noting Amazon and Alphabet remain solid long-term cloud investments despite near-term capex pressure.

Analysis

The cleanest read-through is that the AI capex supercycle is still in its “earnings acceleration” phase for the picks-and-shovels layer, while the hyperscalers are only beginning to feel the margin squeeze from front-loaded infrastructure spend. That creates a two-speed trade: suppliers with pricing power and product differentiation should keep compounding faster than the platform owners over the next 2-4 quarters, even if the latter ultimately capture more of the economic surplus over a 3-5 year horizon. The market is likely still underestimating how long inventory digestion can stay benign when end-demand for compute remains tight.

Second-order effects matter here. If hyperscaler capex keeps stepping up, the most important beneficiary may not be the obvious semiconductor names alone, but the adjacent bottlenecks: power, cooling, networking, advanced packaging, and data-center real estate. That also raises the odds of intermittent supply constraint reratings across the chain, because every upward revision to capex tends to pull forward orders for a broader set of infrastructure inputs before it shows up in reported revenue.

The contrarian risk is that the market is extrapolating linear demand into a highly nonlinear spending regime. If enterprise AI monetization lags capex growth, the hyperscalers will eventually be forced to choose between defending growth and defending margins, which could compress multiples across the complex. For NVDA specifically, the key reversal trigger is not demand collapse but duration disappointment: if 2027-2028 spend visibility becomes less certain, the stock can de-rate well before fundamentals roll over.

The most interesting nuance is that the “better long-term business model” argument for AMZN/GOOGL is real, but it likely belongs in the post-build-out phase, not the current tape. In the nearer term, investors are paying for future option value while absorbing current capex drag, so relative performance should favor NVDA and other infrastructure names until the market starts pricing a lower capex growth rate. That implies the stock-picking opportunity is likely in relative value rather than outright beta.

Key Decisions for Investors

• Stay long NVDA versus a basket of hyperscalers over the next 3-6 months; prefer a relative-value expression (long NVDA / short equal-weight AMZN+GOOGL) because the near-term earnings revision cycle should still favor infrastructure suppliers over capex-heavy platforms.
• Use pullbacks in NVDA to add exposure rather than chase strength; the better entry is after any post-earnings or macro-driven 5-8% drawdown, with a 6-12 month target tied to continued capex revisions higher.
• Build a barbell long in AI infrastructure adjacencies such as networking, power, or cooling beneficiaries versus short lower-growth software or any name where AI spend is being capitalized without clear payback; this captures second-order beneficiaries if capex broadens beyond GPUs.
• If holding AMZN/GOOGL, treat them as 12-24 month compounding holdings but hedge into strength with call overwrites or partial index hedges, because current expectations still underprice the margin drag from sustained capex escalation.
• Avoid shorting NVDA outright unless there is evidence of order slowdown or duration compression; the risk/reward is poor because the next leg lower would likely require a fundamental miss, not just valuation fatigue.