⛏️
πŸ”¬
🏭
🧊
πŸ’Ž
☁️
🌊
🧠
Models
⚑
πŸ“Š
πŸ”’
🌍
CHAPTER 08

Foundation Models

The most competitive chokepoint \u2014 for now

FOUNDATIONMODELSLAYER 08

At the top sit the AI models consumers and enterprises use. This is the most competitive layer β€” but concentration is increasing as training costs soar into the billions.

Enterprise LLM Market
πŸ‡ΊπŸ‡ΈAnthropic (Claude)40%
πŸ‡ΊπŸ‡ΈOpenAI (GPT)27%
πŸ‡ΊπŸ‡ΈGoogle (Gemini)21%
Others12%
Consumer Chatbot Traffic
πŸ‡ΊπŸ‡ΈChatGPT68%
πŸ‡ΊπŸ‡ΈGoogle Gemini18%
πŸ‡ΊπŸ‡ΈClaude3%
Others11%

Global AI investment hit $202 billion in 2025. China's DeepSeek demonstrated competitive models at significantly lower cost, showing alternative approaches are possible.

Frontier AI Model Development
πŸ‡ΊπŸ‡Έ US & AlliesπŸ‡¨πŸ‡³ China
Key Relationships
OpenAI ↔ Microsoft
$13B investment. GPT in Office, Copilot, Bing.
Anthropic ↔ Amazon / Google
$8B from Amazon, $2B+ from Google. Multi-cloud strategy.
Google: Full Stack
TPUs β†’ Gemini β†’ Search, Android, Workspace.
DeepSeek / Baidu (China)
Competitive capability. Cost-efficient training. Compute access remains the constraint.