2 Stocks to Buy on Overdone AI Infrastructure Spending Fears That Could Rise 30% and 50%, According to One Wall Street Analyst

BNP Paribas analyst Nick Jones says fears that Alphabet and Amazon are overspending on AI infrastructure are "overdone," citing strong backlog-to-capex support and improving revenue per employee. Alphabet raised 2026 capex to $175 billion-$185 billion from $91.4 billion in 2025, while Amazon lifted capex to $200 billion from $131.8 billion last year. The bullish case is reinforced by proprietary chips and internal AI demand, with price targets of $390 for Alphabet and $320 for Amazon implying roughly 30% and 50% upside, respectively.

Analysis

The market is treating incremental capex as a margin tax, but the more important signal is that both franchises are choosing to preempt capacity scarcity before it shows up in customer churn or cloud share loss. That matters because once enterprise AI workloads are designed around a provider’s proprietary silicon and software stack, switching costs rise sharply and future pricing power improves; this is a multi-year compounding lever, not a one-quarter expense line.

The second-order winner is less the chip vendor and more the “capacity ecosystem” around high-utilization AI deployment: advanced networking, power, thermal management, and custom silicon integration. Broadcom is the cleanest direct beneficiary on the hardware side because custom ASIC ramp and interconnect intensity should support backlog durability, while Microsoft is the relative loser in narrative terms if investors continue to question whether its own AI monetization can keep pace with its spending profile. The risk is not that AI demand disappears; it is that utilization lags buildout for 2-4 quarters, creating a temporary FCF reset before revenue catches up.

Consensus is still underestimating how much internal demand absorbs the incremental spend. If these companies are using capacity to improve search relevance, ad targeting, product discovery, and developer tooling, the payback period compresses materially versus a pure external cloud bet. The contrarian setup is that the ‘overspend’ debate may peak exactly when the physical infrastructure is most strategically valuable, creating a better entry point in names with the most optionality and balance sheet flexibility.

Near term, the stocks can stay volatile if investors focus on reported capex growth rather than utilization metrics; that creates a days-to-weeks trading headwind. Over a 6-12 month horizon, backlog conversion and evidence of higher revenue per employee should be the key confirmation signals, while any slowdown in enterprise AI bookings or weaker cloud commentary would be the main reversal trigger.

Key Decisions for Investors

• Go long AMZN into weakness over the next 1-3 weeks; use the recent capex-driven drawdown as entry, with a 6-12 month target tied to cloud share stability and margin durability. Risk/reward favors upside if utilization inflects before consensus expects.
• Go long GOOG on a 3-6 month horizon; the market is mispricing the option value of TPU-led AI monetization and internal efficiency gains. Exit if cloud growth decelerates for two consecutive quarters or capex remains elevated without backlog confirmation.
• Pair trade: long AVGO / short MSFT for 3-6 months. AVGO benefits from custom silicon and networking intensity without the same balance sheet debate, while MSFT remains more exposed to scrutiny over monetization versus spend; hedge beta with equal dollar notional.
• Sell short-dated put spreads on AMZN or GOOG after any 3-5% post-earnings selloff. The goal is to monetize elevated sentiment volatility while keeping defined risk; best if the market keeps overreacting to capex headlines.
• Avoid chasing NVDA here on the headline alone; the incremental read-through is indirect and already heavily owned. Better entry would be on evidence that hyperscaler capex is translating into faster deployment, not just larger budgets.