The AI Bubble Question (avc.xyz)

Wall Street and Silicon Valley are clashing over whether the current AI boom is a sustainable expansion or a bubble waiting to pop. To cut through the rhetoric, VC/blogger Evan O’Donnell built a data-driven model and public dashboard that compares the growth rate of inference token usage (a proxy for LLM inference demand) against infrastructure investment. His approach asks: is spending on GPUs, data centers and orchestration justified by real usage growth? The headline technical takeaway: token consumption growth has slowed to roughly 13% month-over-month as of Sept/Oct (down from 30–40% earlier in the year). O’Donnell’s model suggests current infrastructure spend is rational if that ~13% growth persists, but it leaves little runway if growth decelerates further. A key caveat is timeliness—the dashboard relies on lagged data, so near-real-time trends aren’t visible yet. For VCs, operators and market watchers, the implication is clear: keep a close eye on token/consumption and infra-capacity metrics next quarter—those numbers will determine whether investment levels remain justified or if the market faces a sharp correction.