Interview, Fireside Chat
Why Now is the Time for the App Layer | Why Startups Should be TokenMaxxing | Mike Mignano, USV
- Traditional media is projected to face massive unbundling and become largely defunct, while independent media and self-publishing platforms like Substack are expected to grow significantly larger than anticipated post-2022.
- Within five years, the handing of personal keys to AI agents is predicted to become a de facto standard comparable to the adoption of credit cards, driven by the high value of context in an "always-on" world.
- By 2026, energy remains a core investment theme with a focus on portability solutions involving micro data centers placed directly next to generators, alongside continued infrastructure build-out.
- The application layer is expected to see massive value creation distinct from infrastructure, where companies will focus on "obliterating" markets rather than merely automating workflows, with potential $10 billion to $50 billion valuations for entities in the model routing layer.
- AI adoption will drive a structural shift in enterprises toward "harnesses" (human-aligned agents) to manage incentives and security, while engineering teams may evolve into smaller groups of higher-caliber engineers who delegate lower-level tasks to agents.
- Startups are expected to aggressively maximize token spend for coding advantages against incumbents, whereas enterprises may eventually constrain spend due to the scale of their employee bases.
- Token expenditure scenarios project a potential spike where spending could reach 100% of developer salaries, contrasting with current levels of 3.8% and intermediate scenarios of 20%.
- A divergence is anticipated where non-coding workflows are increasingly handled by open models, supporting a "Rebel Alliance" of open weight models and open source harnesses as a counter to large incumbents.
- Future market dynamics may see the failure of mid-tier content creators, with success concentrated only at the extremes of very small or very large production scales.
- Investment strategies may prioritize massive multiples over ownership percentage at later stages, while smaller funds will face price limits at earliest stages to maintain fund math.
- Risks include the potential for traditional enterprise competitors like Microsoft to leverage bundling to crush AI product rivals, model providers entering application layers to compete directly with portfolio startups, and the threat of token costs fundamentally altering business economics.
- Two opposing technological scenarios are identified: AI becoming a commodity due to intelligence plateaus leading to price competition, or a single entity achieving recursive self-improvement and dominating the market by running away with the win.
- Specific company trajectories include "Bored" as a generational firm, Suno evolving into a user-driven creation platform, and "Fable" returning to enable a "100x engineer" to replace multiple mid-tier roles.
- USV plans to focus on collaboration with founders to produce generational returns over the next five to ten years, while acknowledging that fund constraints will prevent betting on every round from specific points onward.