Here are real AI stocks to invest in and speculative ones to avoid

Concerns over valuations in AI-linked names knocked markets this week, and the selloff extended beyond speculative “nuclear” and neocloud plays to core, revenue-generating companies such as Nvidia, Broadcom, GE Vernova and Eaton—stocks the author says they own—raising questions about capital expenditures and accelerated depreciation of AI hardware. The central debate is whether bears’ spreadsheet-driven assumptions (chips losing most value in ~3 years) or technology leaders’ claims are correct: Nvidia’s Jensen Huang says CUDA advances have extended GPU platform life to five-to-six years and CoreWeave re-contracted for H100s, while AMD’s Lisa Su says customers are beginning to see returns on their AI investments. The piece concludes that if managements’ views on useful life and durable demand hold, recent weakness may be overdone and long-term investors should favor high-quality AI players, but warns that elevated capex and depreciation remain key risks to monitor.

Analysis

Equity markets were knocked this week by valuation concerns centered on AI-linked names, with the selloff extending beyond speculative "nuclear" and neocloud plays to core, revenue-generating companies cited in the article — Nvidia (NVDA), Broadcom (AVGO), GE Vernova (GEV) and Eaton (ETN) — stocks the author says they own; the piece notes a partial rebound on Friday and the provided sentiment score is moderately positive (0.45) while market impact is modest (0.35). The central contention is methodological: bears apply spreadsheet-driven depreciation assumptions that render chips near-worthless in ~3 years, while management teams argue software and platform advances have extended useful life. Jensen Huang claims CUDA has stretched GPU platform life to five-to-six years and CoreWeave re-contracted for H100s (released late 2022), and AMD's Lisa Su says customers are beginning to see returns on AI investments, evidence the author uses to defend current valuations. The principal risk remains elevated capex and the potential for overstated useful lives; investors should monitor capex guidance, depreciation policy changes, re-contracting activity for H100-class GPUs, and ongoing management commentary from NVDA, AMD, META and MSFT as the primary indicators that will validate or invalidate the bullish operational thesis.

Key Decisions for Investors

• Consider selectively adding to high-quality, revenue-generating AI exposures (NVDA, AVGO, AMD, MSFT, META) on meaningful pullbacks if you have a multi-year horizon, given management commentary on extended useful lives and evidence of re-contracting for H100s, but size positions to reflect capex/depreciation risk
• Reduce or hedge exposure to highly speculative 'neocloud' and nuclear AI names where valuations appear most stretched and downside from accelerated depreciation is highest
• Monitor three leading indicators to reassess positions: corporate capex guidance and depreciation policy updates, frequency and scale of GPU re-contracting (e.g., H100 deals), and direct management commentary from GPU and cloud leaders
• Use phased entries or protective hedges (tranches, options, stop-losses) to manage episodic volatility tied to narrative shifts between spreadsheet-driven bears and management guidance