The First 5 AI Stocks I'd Buy If I Started From Scratch

The article names Nvidia, Broadcom, Microsoft, Alphabet, and Nebius as the author's top AI stocks to buy now, with Nebius highlighted for potential annual run-rate growth from $1.25 billion at end-2025 to $7 billion-$9 billion by year-end. Nvidia and Broadcom are framed as direct beneficiaries of the AI build-out, while Microsoft and Alphabet are positioned to monetize AI through Azure and Google Cloud. The piece is broadly optimistic on AI infrastructure and hyperscaler demand, but it is opinionated commentary rather than breaking company-specific news.

Analysis

The setup is not a generic “AI wins” basket; it is a barbell between toll collectors and capacity providers. NVDA and AVGO remain the cleanest monetization of capex because they sit upstream of ROI skepticism, meaning they get paid even if downstream AI apps take longer to prove value. That creates an important second-order effect: as hyperscalers and neoclouds race to expand, the real bottleneck is likely to shift from chip supply to power, networking, cooling, and deployment cadence, which should favor adjacent infrastructure suppliers more than application-layer names over the next 6-18 months.

MSFT and GOOG are better framed as call options on usage growth with a built-in defensive characteristic: their AI monetization can compound inside existing cloud and productivity ecosystems without needing a separate “AI winner” narrative. The market still appears to underappreciate that cloud AI demand is sticky once workloads are integrated into enterprise workflows, but that adoption curve is lumpy and can compress near-term margins before revenue catch-up. The main risk is not demand failure; it is capex intensity outrunning monetization, which can cap multiple expansion even if revenue prints remain strong.

NBIS is the only name here where execution risk can overwhelm thematic tailwinds. The market is effectively pricing in a step-function scaling story, so any sign of utilization softness, margin leakage, or customer concentration could drive a sharp de-rating well before fundamentals break. Conversely, if it sustains growth while improving gross margin and financing terms, it can outperform dramatically because the equity is still being valued more like a venture asset than a scaled infrastructure compounder.

The contrarian point: consensus may be overpaying for “AI exposure” while underweighting the fact that power availability, grid interconnects, and depreciation cycles are becoming the real constraints. That argues for owning the picks-and-shovels winners, but being selective on higher-beta capacity names until there is evidence that demand is translating into durable free cash flow rather than just revenue velocity.

Key Decisions for Investors

• Long NVDA / short a basket of high-multiple AI application names for 3-6 months: best risk/reward if AI spend continues but downstream monetization remains delayed; use this as a hedge against AI enthusiasm de-rating outside the infrastructure layer.
• Overweight AVGO vs. MSFT/GOOG on a 6-12 month basis: AVGO offers more direct capex conversion and likely less sensitivity to software monetization delays; target relative outperformance if hyperscaler capex stays elevated.
• Initiate a small, options-defined long NBIS position via 3-6 month calls or call spreads only: upside is asymmetric if scaling continues, but position sizing should assume high drawdown risk on any execution miss or funding wobble.
• Add MSFT and GOOG on pullbacks as lower-volatility AI compounding vehicles over 12+ months: prefer them as core holdings rather than trading vehicles, since the market will likely reward durable cash-flow conversion over headline AI growth.
• Watch for power/grid and data-center supply chain beneficiaries as the next leg of the trade: a basket long on electrical equipment, cooling, and interconnect names can capture the second-order bottleneck if AI capex stays strong.