Alphabet Targets Nvidia with Custom Silicon Push into $300 Billion AI Market

Alphabet's recent move to sell custom silicon directly to customers marks a significant shift in its approach to artificial intelligence (AI) accelerators, positioning it as a more serious threat to Nvidia's dominance.

Nvidia has been the leading player in AI accelerators since it invented the graphics processing unit (GPU) in 1999. The company's chips have become the industry standard for accelerating complex data center workloads due to their ability to perform trillions of calculations per second.

However, Alphabet has been quietly building its own custom silicon solutions, including tensor processing units (TPUs), which are designed specifically for AI workloads. Initially used internally, Alphabet made its TPUs available to Google Cloud customers in 2018 and began selling them directly to clients in the second quarter, increasing its direct competition with Nvidia.

While Nvidia is unlikely to lose its market leadership due to its significant competitive advantage in CUDA, a software platform that allows programmers to write GPU-accelerated applications, Alphabet's TPUs could still become a major player in the AI infrastructure market. According to Gil Luria, head of technology research at D.A. Davidson, custom silicon is a massive growth opportunity for Alphabet and could potentially capture 20% of the AI infrastructure market by 2030.