Sep. 14 at 8:45 PM
Pinterest has unveiled a new AI infrastructure layer developed with NVIDIA using Blackwell GPUs to support multimodal products combining vision and language, including visual search, content understanding and safety systems. The platform integrates NVIDIA Blackwell GPUs, NVIDIA Dynamo and Pinterest’s proprietary visual embeddings.
The new system allows Pinterest Assistant to process 25 times more visual context per request. Pinterest reported roughly 85x faster response startup and 7.3x lower overall latency when using precomputed visual representations instead of repeatedly processing raw images. The infrastructure supports more than 80 billion monthly searches and other functions including ranking, OCR, safety and signal generation.
The initiative builds on nearly five years of collaboration and more than 14,000 NVIDIA GPUs, giving Pinterest a standardized foundation for expanding multimodal and conversational AI across its platform.
$PINS $NVDA