Tunguz: Google is well-positioned for AI leadership, but these bottlenecks still remain

The evolution of Google’s AI and search capabilities is reshaping the company’s strategic direction, with significant implications for developers, hardware infrastructure, and monetization.

Key developments in efficiency: Google has achieved remarkable improvements in both algorithmic and hardware efficiency for AI operations.

• AI algorithm inference costs have decreased by a factor of 1000 over three years
• Google’s data centers now deliver 4x more computing power per unit of electricity compared to five years ago
• Cloud customers are using more than 8x the compute capacity for AI training and inference compared to 18 months ago

Developer adoption and platform growth: Google’s AI platforms are experiencing significant momentum in the developer community.

• 4.4 million developers are now using Gemini models, double the number from six months ago
• Vertex AI usage has increased 20x over the last year
• Google’s models now represent 2 of the top 6 models according to OpenRouter, with adoption rates exceeding OpenAI’s

Emerging AI capabilities: Google is actively developing new ways for users to interact with AI systems.

• Project Mariner, currently in testing, enables AI to understand and complete tasks directly from browser screens
• Circle to Search feature is driving more than 10% of searches among early adopters
• Multimodal interactions combining video, voice, and visual inputs are being developed for future workflows

Monetization and advertising: Contrary to previous concerns, AI-driven search is showing promising financial results.

• AI Overview ads on mobile are achieving similar monetization rates to traditional search
• These results challenge predictions about reduced profitability in AI-enhanced search
• The successful monetization provides a foundation for further innovation in AI-driven advertising

Infrastructure challenges: Google faces significant hardware deployment demands to support AI growth.

• The company has committed $75 billion in capital expenditure for cloud infrastructure
• Current demand exceeds available capacity
• Similar capacity constraints are being experienced across the cloud computing industry

Looking ahead to 2025: The integration of AI into core products and services will fundamentally change user interactions.

• New modalities for computer interaction will become mainstream
• AI agents will enhance both deep research and local process automation
• The combination of hardware improvements and AI capabilities will enable more sophisticated applications

Strategic implications: The successful monetization of AI features and significant infrastructure investments suggest Google is well-positioned for the AI transition, though hardware constraints may temporarily limit growth potential in the near term. The focus will be on delivering these capabilities at scale while maintaining efficiency gains in both algorithms and hardware.