A One-Page Site Asks the Question Nobody Wants Answered: Is AI Profitable?

A minimalist single-page site, isaiprofitable.com, poses a pointed question without offering an answer beyond its title. The framing alone has resonated on Hacker News, where it functions as shorthand for ongoing skepticism about whether the trillions in capex, GPU buildouts, and frontier-model training runs are translating into durable margins for the companies bankrolling them.

The joke lands because the underlying numbers are genuinely murky. Hyperscalers bundle AI revenue into cloud line items, model labs report headline revenue without breaking out inference costs against training amortization, and the most-quoted unit economics come from companies with strong incentives to project confidence. Meanwhile, depreciation schedules on H100-class hardware, power contracts, and the cost of serving reasoning models at scale all push the break-even horizon further out than marketing decks suggest.

The page’s appeal is that it crystallizes a question the industry has spent two years avoiding in plain terms. As a piece of commentary it is closer to a protest sign than analysis, but its traction reflects a real shift in tone — investors and engineers alike are starting to demand the income statement, not just the demo.