600 million monthly users, 5 billion searches per month, $4 billion committed to AWS through 2031. The largest infrastructure investment in Pinterest's 16-year history — and it is entirely about AI-powered visual discovery.
The architecture: AWS Trainium to train and run the large language models and vision-language models that power personalised visual search and AI-assisted discovery. Graviton — which already powers roughly a third of Pinterest's entire compute infrastructure — expanding to run more of the discovery systems serving hundreds of millions of users simultaneously. And a major infrastructure modernisation running underneath both: migration from traditional EC2 to Kubernetes on Amazon EKS, improving developer velocity, operational reliability, and efficiency at global scale.
The products being built on that infrastructure: Taste Graph — Pinterest's proprietary AI system mapping individual taste across categories, styles, and intent, helping users move from open inspiration to personalised, actionable results. Pinterest Assistant — multi-turn conversational discovery launched recently, powered by open-source vision-language models optimised for scale. And the next generation of recommendation systems evolving from traditional retrieval to transformer-based generative models.
Pinterest CTO Matt Madrigal: "This expanded commitment with AWS gives us the compute flexibility, hardware optionality, and infrastructure efficiency to accelerate our AI vision for the next generation of visual discovery on Pinterest. This strategic partnership will help accelerate AI innovation at Pinterest, improving both our consumer experience and advertiser performance."
AWS SVP Dave Brown: "AWS is the best place to do AI at this scale, and we're committed to helping Pinterest's teams move faster and think bigger."
Now read this against the full visual commerce picture. Visual search queries on mobile grew 312% year-over-year. Google launched Style-Matching Intelligence. ASOS put its catalogue inside ChatGPT with shoppable video. Hey Savi and PayPal closed the first agentic commerce loop in UK fashion. Phia raised $35.5 million for AI shopping intelligence. Every company in commerce is racing toward the same moment — the instant a consumer sees something they want and converts that visual impulse into a transaction.
Pinterest owns that moment better than anyone. 600 million monthly users. Fashion as core use case. The $4 billion builds the AI to close the loop.
The global clothing sector is down 63% in Bitcoin terms since April 2023. A $10,000 investment in the fashion basket returned $9,730, against $17,400 from the S&P 500, $26,510 from Bitcoin, and $23,170 from gold - https://smarttimes.net/smart-fashion-council