The market is pricing AI infrastructure faster than it's pricing AI outcomes. Stripe's reported $7B-plus acquisition of OpenRouter values a payments layer for AI API calls at a level that assumes the underlying models generate durable revenue; meanwhile, Goldman and Intel back Higgsfield at $5.4B on the bet that AI video tools will capture marketing spend. But the actual demand signal remains muddled. Zuckerberg's public case for superintelligence for everyone landed poorly among the people who actually build these systems, suggesting the founder-led narrative around transformative AI is decoupling from practitioner skepticism. Anthropic's Dario Amodei frames backlash as a trust crisis rather than a legitimacy crisis, a framing that lets him avoid the harder question: whether the thing being distrusted is sound. The infrastructure play makes sense if you believe models will consolidate around a few dominant architectures and that routing, integration, and governance layers will become commodities. The content play makes sense if you believe marketing budgets will shift to AI-generated video faster than they shift to anything else. But neither bet requires that the AI itself works as advertised, only that enough capital and corporate inertia will push adoption forward anyway. That's the real signal underneath today's noise: the money is moving faster than the evidence.
Sloane Duvall