Why most AI companies should consider selling in the next 18 months

Executive overview

Every major tech cycle ends with 95%+ of companies failing. AI is no different — a handful will dominate, the rest will be commoditised or outcompeted by labs.

The core question for every AI founder right now: are you one of the dozen that will matter in 10 years, or are you at your value-maximising moment?

If you're not in the durable handful, now or the next 12–18 months is the best exit window you'll ever have.

The historical pattern

• 1,500–2,000 companies went public during the dot-com bubble; only a dozen or two survived in any meaningful form
• Every technology cycle — including the automotive era 100 years ago — collapses to a small number of winners
• There is no reason the AI cycle will be different

Identifying whether you're a survivor

• Ask: will your product get dramatically better as the underlying model improves — in a way customers still value?
• Depth and breadth of product matters: multiple integrated products embedded in customer workflows are harder to displace
• Workflow embedding is the real moat — change management friction protects incumbents more than raw AI quality
• Proprietary data can help, but data moats are generally overstated
• The second derivative of growth is the signal: if growth is plateauing, a headwind is coming

The lab and infrastructure layer

• OpenAI, Anthropic, Google are likely durable — they occupy the foundational layer
• The lab tier is an oligopoly, not a monopoly — no single player has pulled far enough ahead in capabilities
• Meta, xAI, and others could shift this, but the compute constraint provides a near-term ceiling
• Market structure roughly matches what was predicted: labs aligned with cloud hyperscalers

Exit options for everyone else

• Big tech / hyperscalers: Apple, Amazon, Google, Oracle, Samsung, Tesla, SpaceX — unprecedented buying power from trillion-dollar market caps
• Vertical incumbents: companies like Thomson Reuters for legal or accounting AI
• Competitor mergers: often underused — merging with a neck-and-neck rival stops mutual price destruction and concentrates firepower against larger incumbents (X.com + PayPal is the canonical example)
• 1% of a $3 trillion market cap is $30 billion — large acquisitions that once seemed impossible are now routine

The value-maximising window

• Every company has a peak valuation window, typically 6–12 months
• The window is predictable: spot it when growth starts to plateau before a known headwind arrives
• Founders who are in the durable tier should never sell; everyone else should treat this moment seriously