admin The agreement, announced on 4 June, is the largest infrastructure commitment in Pinterest’s history and places AWS custom silicon at the centre of the company’s next phase of product development. Pinterest plans to increase its use of AWS Trainium for training and running large language models and vision-language models, while widening deployment of Graviton processors, which already support about a third of its compute infrastructure.
The deal gives Pinterest added capacity for AI workloads at a time when image-led search and automated advertising are becoming central to the company’s growth strategy. Its systems must interpret billions of images, connect them with user intent and turn browsing activity into actionable recommendations for fashion, home design, food, beauty and retail categories.
Matt Madrigal, Pinterest’s chief technology officer, said the company was “heavily investing in AI to make discovery more personal, visual and actionable” for users. He said the expanded AWS commitment would provide “compute flexibility, hardware optionality and infrastructure efficiency” and help improve consumer experiences and advertiser performance through proprietary models and open-source models.
Pinterest and AWS have worked together since 2010, helping optimise large-scale data lake infrastructure and core service reliability. The agreement extends that relationship across AI model training, inference and platform infrastructure, deepening AWS’s role as Pinterest’s preferred cloud services provider.
The company also plans to continue shifting from traditional EC2-based environments to a Kubernetes-based architecture on Amazon Elastic Kubernetes Service. The migration is intended to improve developer velocity, reliability and efficiency as AI-assisted products become more data-intensive.
Pinterest’s AI push is linked to its broader effort to define itself less as a conventional social network and more as a visual search and shopping engine. Its proprietary Taste Graph helps connect user interests with personalised recommendations, while transformer-based models and multimodal systems have improved the way the platform ranks, retrieves and serves content. Pinterest Assistant, a conversational discovery feature powered by open-source vision-language models, forms part of that shift.
The agreement follows a stronger start to 2026 for Pinterest. The company reported first-quarter revenue of US$1.008 billion, up 18 per cent year on year, with global monthly active users rising 11 per cent to a record 631 million. It posted a GAAP net loss of US$74 million, adjusted EBITDA of US$207 million, operating cash flow of US$328 million and free cash flow of US$312 million for the quarter.
Bill Ready has argued that Pinterest occupies a distinctive position because users often arrive with purchasing or planning intent rather than purely for entertainment. That distinction has become more important as advertisers demand measurable returns and platforms race to connect search, recommendations and commerce. AI-powered products such as Pinterest Performance+ are designed to automate campaign optimisation and make advertising more effective across search and shopping surfaces.
For AWS, the Pinterest commitment adds to a wave of large enterprise cloud deals tied to AI infrastructure and custom chips. Snowflake signed a US$6 billion AWS agreement in May covering AI and data workloads, including Graviton processors. Amazon has also stepped up financing for its infrastructure buildout, securing a US$17.5 billion delayed-draw loan facility in June as hyperscalers pour capital into data centres, chips, networking, energy and cooling systems.
