Most AI Companies Won't Survive (Tech Investor Explains)

TL;DR

Historically 95-99% of companies in every tech cycle fail, and most AI companies will share that fate within 12-18 months.

Key Points

• 1.Historical tech cycles see 95-99% company failure rates. Of ~2,000 internet-era companies that went public around 1999-2000, only a dozen or two survived — the investor argues AI will follow the same pattern as SaaS, mobile, and crypto.
• 2.Founders of vulnerable AI companies should consider exiting in the next 12-18 months. There is typically a 6-12 month window of peak valuation before headwinds hit, visible when growth's second derivative begins to plateau.
• 3.Core AI labs — OpenAI, Anthropic, Google — are the most durable players. The investor predicted an oligopoly aligned with cloud providers three years ago, and that structure has largely materialized, though Meta and xAI could still shift it.
• 4.Application-layer companies survive if they embed deeply into workflows and proprietary data. The investor highlights Harvey (legal), Abridge (health), and Decagon/Sierra (customer success) as examples — durability comes from change-management lock-in, not just model quality.
• 5.Multiple exit paths exist beyond lab acquisitions. Potential buyers include hyperscalers, vertical incumbents like Thomson Reuters, and even competitor mergers — the investor cites X.com/PayPal and the near Uber/Lyft deal as models, enabled by trillion-dollar market caps where 1% equals $30B in buying power.