If You're Worried About an AI Bubble, Here's the Tech Stock Portfolio I'd Build

The article argues Nvidia, Amazon and Alphabet offer comparatively resilient exposure to AI despite bubble concerns. Nvidia reported fiscal Q2 revenue growth of 106% to $96 billion and held more than $99 billion in liquidity; Amazon Q2 sales rose 20% to $201 billion, with AWS sales up 37%; Alphabet Q2 revenue grew 24% to $120 billion, including 82% growth in Google Cloud. The piece highlights substantial planned capex—$220 billion for Amazon and $195 billion–$205 billion for Alphabet this year—while presenting their liquidity, profitability and valuations as sources of downside protection.

Analysis

The key risk is treating current earnings multiples as downside protection when AI profits and spending are unusually cyclical. For NVIDIA, a multiple on peak accelerator demand can re-rate sharply if hyperscalers slow orders, shift workloads to custom silicon, or stretch replacement cycles; the cash balance reduces financing risk, not demand or valuation risk. The capex race benefits upstream suppliers of memory, networking, power and cooling, but may ultimately pressure cloud returns and pricing as AWS, Google Cloud and other providers compete to monetize capacity.

For Amazon and Alphabet, the underwriting question is incremental return on infrastructure, not whether they can fund it. Large investment or non-operating gains cited in the article make reported earnings and headline P/Es poor anchors; verify normalized operating income, cash flow, depreciation and capex guidance before acting. Alphabet also faces a two-sided AI effect: cloud demand can rise while AI-mediated search changes ad monetization. Amazon’s exposure is more directly tied to whether AWS growth converts into durable margins after the investment cycle.

Near term, earnings and capex-guidance revisions are likely to matter more than the broad AI narrative. Over 6–18 months, utilization and returns on deployed capacity will determine whether current spending supports earnings or creates excess supply. The contrarian point: “safe AI” is not a valuation hedge; balance-sheet strength limits funding risk but does not prevent multiple compression. No clean outright trade follows without normalized earnings and current valuation data.

Key Decisions for Investors

• Do not add to NVDA solely on the article’s P/E framing. Reassess after its next results: order growth, customer concentration, supply availability and hyperscaler capex revisions are the relevant falsifiers; a sustained slowdown in accelerator demand would undermine the thesis even with ample liquidity.
• Keep AMZN and GOOG as conditional, cash-generative AI exposures rather than treating planned capex as assured value creation. Before increasing either, verify operating cash flow after capex, segment margins and whether cloud growth is translating into returns on new capacity.
• Watch upstream memory, networking and power/cooling suppliers for second-order beneficiaries of continued buildout, but require evidence of order conversion and pricing power; a hyperscaler capex cut would likely reverse that trade quickly.
• Avoid a valuation-only short in PLTR based on the cited multiple. The article supplies no earnings trajectory or current pricing context to establish timing or asymmetry; revisit only with evidence of estimate cuts or fading growth against a still-extreme valuation.