There isn't an AI bubble–there are three (www.fastcompany.com)

The author argues AI today isn’t a single “bubble” but three overlapping ones: a speculative asset bubble (eye‑watering valuations — Nvidia ~50x earnings, Tesla ~200x — and investor froth), an infrastructure bubble (huge GPU, power and data‑center buildouts — McKinsey talks of a $7T “race to scale,” with eight 2025 projects already topping $1T) and a hype bubble (promises outpacing reality — an MIT study found ~95% of AI pilots fail to return value). Each bubble has different mechanics and risks: speculation primarily impacts capital markets, infrastructure risks overcapacity, and hype leads companies to pursue the wrong problems with the wrong expectations. For AI/ML practitioners and business leaders this means a clear playbook: don’t flee AI, treat the speculative and infrastructure bubbles as environmental factors (they may even lower future costs), but confront the hype by being systematic. Start problem‑first (map friction points), balance a portfolio of quick wins, strategic projects and moonshots, and integrate systems so models and data streams compound value rather than sit in isolated pilots. Practically, that lets organizations cherry‑pick proven tools, hire experienced talent as the market corrects, and harvest durable ROI while others debate valuations — turning the bubbles’ excesses into operational advantage.