
AI investment and infrastructure development hinge on four critical bottlenecks: context feedback, compute, capital, and culture. While scaling laws remain valid in domains like material science, the industry currently suffers from an "infrastructure wastage crisis" caused by a lack of compute fungibility and open standards. Rather than relying on general-purpose foundation models, the next generation of $100 billion companies will be "frontier systems" firms that vertically integrate physical data collection, specialized compute, and mission-driven culture. Anj Midha, a founding investor in Anthropic and founder of AMP, argues that sovereign, locally managed infrastructure is essential for mission-critical workloads, particularly in Europe. To sustain innovation, the ecosystem must move toward "optimal competition" and adopt coordinated defensive protocols—an "Iron Dome" for inference—to protect against adversarial distillation attacks while shifting venture capital toward deep, long-term co-founding partnerships.
Sign in to continue reading, translating and more.
Continue