Here are real AI stocks to invest in and speculative ones to avoid (www.cnbc.com)

Stocks tied to AI faced a pullback this week as investors fretted over valuations—especially in speculative plays—but the selloff also hit established, revenue-generating companies driving the AI buildout. Portfolio managers (including Jim Cramer’s Investing Club) argue that core players such as Nvidia, Broadcom, AMD, Meta, Microsoft, GE Vernova and Eaton are being punished despite solid fundamentals because fears about capex and rapid depreciation of AI hardware are being overstated. The broader point: this is a structural investment cycle—data-center expansion, chips and power infrastructure—that many see as underpinning long-term productivity gains akin to prior tech revolutions. Technically, the debate centers on useful life assumptions for AI assets. Bears extrapolate historical depreciation and model GPUs as nearly worthless after ~3 years; management teams counter that software and architecture advances (Nvidia’s CUDA, for example) extend GPU platform life to roughly 5–6 years. Real-world signals—CoreWeave re-contracting H100 accelerators (launched late 2022)—are cited as evidence GPUs retain value. That said, elevated capex and potential overstated useful lives remain real risks. Investors must weigh spreadsheet-driven conservatism against management track records and customer behavior; disclosures note the authors hold positions in many of the named stocks.