I'm Not Convinced the AI Bubble Exists

The central claim is that AI’s enormous valuations may reflect genuine growth rather than a speculative bubble. The industry has reportedly spent about $1.9 trillion and committed to roughly $3 trillion more, but much of this is one-time investment in data centres, not a permanent operating cost. Anthropic’s filings suggest it earns $2.08 for each dollar spent on computing, while its revenue rose from $41 million in 2023 to $16.3 billion in the first half of 2026. The author argues that even a rough estimate of $250 billion in annual repayments on $5 trillion of infrastructure costs is not implausible for an industry with major revenue streams, valuable equity, and options to reduce spending.

Future profits could grow as AI is adopted in programming, mathematics, professional paperwork, and autonomous vehicles; the argument is that widespread use can expand even without further model breakthroughs. Competition and the lack of a durable technical moat do not necessarily prevent profits: consumer habits and companies’ costly integrations can retain customers. Nor does the author see an imminent funding crisis, given potential IPOs, public debt markets, and growing operating profits. Finally, public firms such as Alphabet, Meta, and NVIDIA have price-to-earnings ratios around the 20s, unlike the unprofitable or extravagantly valued companies of the dot-com era. Alphabet’s growth also appears to come more from Google Services than from cloud computing, weakening the case that AI infrastructure spending alone is inflating its earnings.