Pinterest is doubling down on Amazon Web Services (AWS) for its artificial intelligence ambitions, announcing a planned $4 billion commitment through 2031. This marks the largest infrastructure deal in the visual discovery platform's history. The company aims to accelerate AI innovation for its more than 600 million monthly users.
Related startups
The expanded partnership, building on a relationship that began in 2010, will see Pinterest utilize AWS's custom silicon. This includes AWS Trainium and Graviton processors to train and run advanced AI models at scale. These models are crucial for powering visual search and personalized discovery features.
Scaling AI with Custom Silicon
Pinterest has long relied on AI to refine its recommendation systems and personalize user experiences. Recent advancements include the development of multi-turn conversational discovery via Pinterest Assistant, powered by open-source vision-language models. To support these demanding workloads, Pinterest is turning to AWS's specialized hardware.
The company plans to use AWS Trainium for large language and vision-language models that drive AI-assisted discovery. Pinterest is also increasing its use of Graviton processors, which already power approximately one-third of its compute infrastructure. This strategic shift targets enhanced efficiency and performance for AI-driven features.
