Why Now is the Time for the App Layer | Why Startups Should be TokenMaxxing | Mike Mignano, USV

The AI industry has moved beyond the foundational infrastructure phase. Value is now shifting to the application layer, where mission-driven startups that focus on high-context, defensible workflows can outmaneuver large incumbents.

• Shift to applications — The era of building core model infrastructure is largely complete, and the current opportunity lies in developing software applications that utilize these technologies [07:46].

• Obliteration over automation — USV prefers backing businesses that fundamentally reinvent how a market operates, rather than companies that simply make existing, inefficient business processes slightly faster [32:56].

• The "Rebel Alliance" — The firm is betting on open-weight models, distributed compute, and human-aligned agents as a counterweight to the large, centralized model providers [10:44].

• Startup token strategy — CEOs should continue aggressive "tokenmaxxing"—spending heavily on model usage—because leaner, high-performing engineering teams require every possible advantage to compete against large, well-funded incumbents [23:32].

• Building defensible moats — Startups survive by securing deep, proprietary context, such as clinical notes or corporate meeting data, which is difficult for massive model providers to commoditize or replicate [41:27].

• The energy bet — USV has been investing in energy infrastructure since 2021, as AI growth necessitates new solutions for energy generation and portability [29:32].

  • Small nuclear reactors are an example of this innovation [30:37].
  • Micro data centers placed near power sources solve critical portability problems [31:27].

• Investment discipline — Series A rounds are now frequently priced between $80M and $150M, forcing investors to be highly disciplined regarding ownership and long-term upside [43:23].

• Founder-first approach — Investors should avoid projecting their own product ideas onto founders and instead bet on the team's ability to execute and adapt their own unique vision [51:47].

• How does USV mitigate the risks associated with investing in capital-intensive energy startups?

• Why does the shift toward AI agents lead to the expectation that engineering teams will become smaller?