AI大模型賽道太擁擠，隻投兩票，我會選Claude和Gemini，OpenAI出局。OpenAI強相關的公司，大家小心。

Anthropic (Claude)

Constitutional AI · Claude Code · Enterprise

Why it survives: Anthropic has cracked the hardest problem in AI monetization — enterprise willingness to pay. At $14B ARR growing 10× per year, with positive cash flow projected by 2027, it's the only pure-play AI lab on a path to self-sustainability. Winning 70% of head-to-head enterprise deals means customers actively choose it over cheaper alternatives. That's a quality moat, not just a distribution moat. And its cloud-agnostic positioning (AWS + GCP + Azure) means it can't be platform-locked out.

Google (Gemini)

Alphabet · DeepMind · TPUs

Why it survives: Google is the only player that doesn't need AI to be profitable — it's already a $400B/yr revenue machine. That financial permanence means it can subsidize AI indefinitely, continue building custom silicon (TPUs), and embed Gemini into 3 billion devices without needing a return for years. No other player has this runway. Distribution alone makes Google unkillable in AI.

Why OpenAI is the odd one out in a two-player world

OpenAI has the biggest brand — but brand alone doesn't win a 20-year race. It burns $14–17B per year and has never turned a profit. Its enterprise win rate is being eroded by Claude, and its consumer share is being eaten by Gemini — both margins being squeezed simultaneously. The IPO will buy time, but the capital structure is fragile. In a brutal consolidation scenario, OpenAI likely gets acquired (most probably by Microsoft) rather than surviving as an independent entity.