Why it matters
The strategic read
Adds another named enterprise buyer to the AWS custom-silicon capacity lane. Pinterest is not just signing a generic SaaS agreement; it is locking a multi-billion-dollar AI infrastructure path around Trainium, Graviton, model hosting, inference, EKS modernization and AI-assisted discovery for more than 600 million monthly users.
Source-supported terms
What was disclosed
- 01
Amazon said Pinterest announced a planned $4B AWS commitment through 2031, the largest infrastructure deal in Pinterest history.
- 02
The agreement extends a Pinterest/AWS relationship that began in 2010 and supports AI training, inference and platform infrastructure.
- 03
Pinterest will use AWS Trainium to host and run large language models and vision-language models for personalized visual search and AI-assisted discovery.
- 04
Pinterest will expand its use of AWS Graviton, which Amazon says already powers roughly one-third of Pinterest compute infrastructure.
- 05
Amazon said the AI models supported by the agreement power visual search and discovery for more than 600 million monthly users.
- 06
Pinterest is also using the expanded agreement to move from traditional EC2-based environments toward Kubernetes-based architecture on Amazon EKS.
- 07
DCD reported the commitment as a new $4B infrastructure agreement with AWS running through 2031.
- 08
Pinterest CTO Matt Madrigal said the expanded AWS commitment gives Pinterest compute flexibility, hardware optionality and infrastructure efficiency for its AI vision.
Disclosure boundary
What is known—and what is not
Primary Amazon / AWS announcement plus DCD near-primary coverage. Public details disclose value, term, AI workload purpose, custom-silicon path, existing Graviton footprint and EKS modernization direction. Full cloud-services agreement, annual spend ramp, reserved-capacity mechanics, Trainium generation mix, region allocation, data-center sites, pricing, termination rights, service-level terms and workload-level capacity commitments are not public.