The AI Bubble is Popping: Why Generative AI Companies Are Unprofitable

Generative AI vs. Cloud Infrastructure

• ☁️ Generative AI and Large Language Models (LLMs) are fundamentally different from cloud infrastructure like AWS, Azure, or Google Cloud.
• 💡 Unlike cloud services which provide versatile infrastructure for various applications, LLMs are a feature, not the foundation.
• 🚀 AWS was built out of necessity to handle Amazon's own massive infrastructure needs, scaling a proven demand for cloud-based services.

The Unprofitability of LLM Companies

• 💸 Most LLM companies are deeply unprofitable, with few exceptions like Midjourney (which may no longer be profitable).
• 📈 Only a handful of generative AI companies achieve significant annualized revenue, and many of these have been acquired.
• 📉 Even top-tier companies like OpenAI and Anthropic, despite billions in revenue, are losing billions annually.

Cursor and Replet: A Case Study in Unsustainable Growth

• ⚠️ Cursor's rapid growth was fueled by an unsustainable business model, leading to drastic changes in its terms of service and pricing.
• 💰 The company had to replace its original model with opaque terms and rate limits, effectively restricting user access and increasing costs.
• 📉 Similar pricing model changes have affected companies like Replet, disfavoring users and increasing their expenses.

The Shakedown by Big Model Providers

• 🏦 OpenAI and Anthropic are imposing new terms of business, requiring startups to pay for priority access or face indeterminate delays and rate limits.
• 🤝 These changes force startups to commit to minimum throughputs, ensuring revenue for model providers even if the startup's customers churn.
• 🎯 This dynamic makes it significantly more expensive for AI startups to gain real market traction and scale.

The Illusion of AI Agents and Consumer Startups

• 🤖 The term "agent" is often used deceptively, with most products being advanced chatbots rather than autonomous entities.
• 📉 Even with demos, AI agents show low success rates on tasks, and their implementation is often expensive and inconsistent.
• 🚫 There's a significant lack of profitable consumer AI startups, with many struggling to find a viable business model beyond basic chatbots or search functions.

Troubling Signs in the Generative AI Space

• 📊 Public companies running AI services are making **