Pinterest has taken a bold step to guarantee the tech future by agreeing to pay Amazon Web Services $4 billion over a 13-year period, all the way through 2031. This is the biggest acquisition in the social media firm’s own records and is an extension of the collaboration that was started in 2010. This sort of long term investment is often a mark of a company’s change in product growth and development strategies. It’s clearly about artificial intelligence for Pinterest.
The announcement caused Pinterest to jump almost six percent and Amazon to gain one and a half percent. Investors were obviously onto the idea. What sets this deal apart from normal cloud contracts is the hardware that’s involved. Pinterest will be getting access to Amazon’s custom chip processors: Graviton and Trainium. These are not off the shelf components. They are engineered to efficiently deal with the heavy computation needs of modern AI models. On platforms such as Pinterest, where people discover what they are looking for through images and visual search, fast processing equates to a better user experience.
In its statement, Pinterest’s Chief Technology Officer (CTO) Matt Madrigal stated, “This new partnership with AWS will provide us with the flexibility in compute, optionality in hardware and efficiency in infrastructure to move our AI vision forward and faster.” This is important in the wording as it points to three distinct needs. Flexibility involves the ability to expand or reduce without commitment to a specific infrastructure. With hardware optionality, Pinterest isn’t betting all its chips on one chip, it can select the best chip for the job. The unheralded, unsung hero of any cloud deal is infrastructure efficiency: wasted compute cycles are wasted money.
Over the past several years, Pinterest has been steadily investing in AI tools. The company introduced major enhancements to its Performance+ ad system, which can enable ads to reach individuals without annoying them or being intrusive. For such an advertising technology, real-time decision making is required. The site has mere milliseconds to determine which pins to display and which ads to place next to them when anyone starts looking for home decor ideas or wedding inspiration. That’s possible with the help of AI. AI does need a huge power capacity though, and that’s what this $4 billion deal is meant to provide.
Pinterest has been a partner of AWS since 2010 to enhance the reliability and performance of its core services. That’s a long runway in the tech industry. The two companies know and understand each other’s operations, failure patterns, and scaling challenges after a decade and a half of working together.The two companies are familiar with each other’s operational rhythms, challenges, and failure patterns, as well as their scaling pains and pain gains, after a decade and a half of working together. This new deal is not a fresh one. It is based on 10+ years of developing trust and technical integration. The company will also expand its use of Amazon’s custom silicon chips as a way to get the performance benefits of specialized hardware without relying on the vendor.
The most intriguing part of this agreement is that Pinterest is investing in AWS Trainium for large language models and vision language models. It’s the technologies behind things such as personalized visual search, AI assisted discovery on the platform. A vision language model is like when you’re looking for something fuzzy, such as “cozy living room ideas” and Pinterest magically displays what you were thinking about without you needing to enter the proper keywords. Trainium chips are optimized for training the models which is often the costliest, time-consuming component of AI development.
Pinterest earlier last month predicted revenue for its second quarter would exceed Wall Street estimates. This confidence is probably fueled by positive early results from its investments in artificial intelligence. Logically, when a social media platform can prove this to advertisers with better return on spend, it’s a no-brainer that it will get more budget. Nevertheless, the competition is tough. TikTok and Meta’s Instagram and Facebook aren’t resting either. All of them are investing in AI-powered advertising and discovery. The thing that Pinterest has always enjoyed is its niche – folks go to Pinterest with a plan in mind, frequently planning a renovation or an event. That’s the intention of traffic and its worth. The quicker and more precise the discovery process is, the more valuable it becomes.
This deal has a quieter aspect that you should consider. Cost control is another benefit Pinterest reaps through using multiple silicon-based compute options. When workloads for AI grow out of control, so can cloud computing bills. The long term agreement with hardware targets gives Pinterest pricing predictability. That’s important for a business that must be profitable every single month, but allocates a significant amount of resources for research and development.
