AI’s next investment cycle belongs to applications, Pg8

Key Highlights:

• In 2025, companies invested approximately $320 billion in AI infrastructure, while spending on AI applications reached $19 billion, comprising over half of all generative AI spending.
• AI applications demonstrate real market demand, with at least 10 products generating over $1 billion in annual recurring revenue and 50 products exceeding $100 million.
• Meta's $2 billion acquisition of Manus, a Singapore-based AI agent startup, highlights the shift towards valuing effective AI applications with proven business success.
• Coding tools dominated the departmental AI market in 2025, accounting for $4 billion of the $7.3 billion total, with half of all developers using AI coding tools daily.

Detailed Insights:

• The AI industry is transitioning from infrastructure investments to focusing on profitable AI applications due to thin profit margins in foundation model businesses.
• Anthropic has increased its share of enterprise LLM spending to 40% by dominating coding applications, while OpenAI's enterprise share has decreased.
• Generative AI achieved a 34% contribution margin in 2025, its first profitable year, with potential to rise to 67% by 2028 as infrastructure costs decrease.
• Investors are prioritizing AI solutions built for specific verticals like healthcare, finance, and manufacturing that are deeply integrated into workflows and utilize unique data.
• Governments face challenges regarding competition as foundation model providers develop their own applications, potentially impacting independent application builders.
• Policymakers should avoid strict regulations to allow experimentation and innovation in the AI application layer, while focusing on competition rules and acquisition reviews.

Key Concepts Involved:

• Generative AI: AI models that can generate new content, such as text, images, or code.
• Foundation Model: Large AI models trained on vast amounts of data, serving as a base for various applications.
• LLM (Large Language Model): A type of AI model that uses deep learning algorithms to process and generate human language.