According to an analysis by SemiAnalysis, this transition reflects a strategic pivot at Google’s parent company, Alphabet, to prioritize infrastructure and cloud revenue over maintaining the lead in raw model capabilities.
This shift matters because Google Cloud Platform (GCP) is emerging as a more significant economic driver than the company’s first-party AI models.
While the Gemini model generated an estimated $12 billion in annual recurring revenue in mid-2026, SemiAnalysis projects that by the end of 2027, GCP could reach over $73 billion in third-party AI infrastructure sales and $120 billion from selling its proprietary Tensor Processing Units (TPUs), which are specialized AI chips.
This indicates that Google is focusing on becoming the foundational "electrical grid" for the AI industry, providing the essential computing power for other companies' applications rather than competing solely on model performance.
The new strategy emphasizes the diffusion of technology—spreading AI tools across the broader economy—rather than concentrated invention.
Google is leveraging its massive distribution network, including Android and Search, to deploy smaller, more efficient AI models at scale while selling its compute capacity to external customers and competitors.
By prioritizing high-margin infrastructure sales and capital expenditure on its cloud business, Google aims to capture value as a universal platform for AI, even if it cedes the lead in developing the world's most powerful individual models.