AI's Trillion-Dollar Opportunity: Sequoia AI Ascent 2025 Keynote

Sequoia partners Packer Radi, Sonia, and Constantine argue AI’s TAM is 10x+ cloud, agents will form a full economy, and founders must run at maximum velocity now.

• AI’s addressable market attacks both software AND services budgets — at least 10x larger than cloud computing’s starting TAM.
• Cost per token dropped 99% in 12–18 months; gross margins will improve as price moves from tool to outcome.
• ChatGPT daily/monthly active user ratio now approaching Reddit levels — hype-to-reality gap has closed.
• Vertical agents (Expo in pentesting, Traversal in DevOps, Meter in networking) are already outperforming best human specialists.
• Sequoia predicts an agent economy where agents transfer resources and value, not just information — requiring persistent identity, MCP-style protocols, and new security layers.
• Pre-training scaling is slowing after 9–10 orders of magnitude since AlexNet; inference-time reasoning, synthetic data, and tool use are the new scaling vectors.
• First one-person unicorn hasn’t happened yet, but companies are scaling faster with fewer people than ever before.
• Founders must adopt a stochastic mindset: AI outputs are probabilistic, not deterministic, requiring a management rather than IC engineering approach.

2025-05-07 · Watch on YouTube