For years, flat-rate AI subscriptions heavily subsidized power users, masking the true cost of token consumption. As providers shift to usage-based billing, this subsidy is ending. Enterprises now face "sticker shock" and must justify AI spend with clear ROI, moving from rampant experimentation to cost-conscious implementation.

AI application-layer companies are knowingly accepting negative gross margins by reselling expensive model inference. Their strategy is to first lock in users with a superior UX, then solve the cost problem later through vertical integration or cheaper models.

The aggressive price-cutting for AI APIs by companies like OpenAI and Meta is not about immediate profitability. It's compared to the early days of Uber, which subsidized rides to capture the market from taxis, suggesting a long-term play for dominance over short-term revenue.

The perceived constraint on AI compute isn't a true supply issue, but a consequence of VC-funded companies pricing their services below cost to fuel growth. This creates artificial demand that masks the true, profitable market size until unit economics are forced.

Unprofitable AI models mirror Uber's early strategy. By subsidizing services, they integrate into workflows and create dependency. Once users rely on the tool (e.g., a law firm replacing an associate), prices can be increased dramatically to reflect the massive value created, ultimately achieving profitability.

The narrative of "off the charts" AI demand is misleading. Major AI providers like OpenAI are "burning tens of billions of dollars," indicating they are not charging the true cost for their services. A realistic picture of demand will only emerge once they are forced to price for profitability, which could significantly cool the market.

To capture market share, AI labs are offering access to their latest models at prices far below their actual cost. This creates a short-term "price war" that benefits users with heavily subsidized access but highlights the industry's shaky unit economics.

Current unprofitability in some AI applications, like subsidizing tokens for coding, is a deliberate strategy. Similar to Uber's early city-by-city expansion, AI labs are subsidizing usage to rapidly gain market share, gather data, and build a powerful flywheel effect that will serve as a long-term competitive moat.

AI companies like OpenAI are losing money on their popular subscription plans. The computational cost (inference) to serve a user, especially a power user, often exceeds the subscription fee. This subsidized model is propped up by venture capital and is not sustainable long-term.