Snapchat could afford to build its own infrastructure if it wanted to. A similar sized Unicorn, DocuSign, has about 2000 employees and probably several hundred million in revenue per year (valuation is around a billion or two, depending on who you ask). They built their own data centers around the world.<p>But DocuSign derives a lot of money from B2B and uptime and location of the servers is important to other Enterprises. DocuSign also started building out its services more than a decade ago before a lot of the AWS or Google Cloud infra got built. So, the decision to build your own infra is as much a decision based on alternatives available. Few alternatives? Build your own infra.<p>By staying on GCP, Snapchat also guarantees its service and uptime will not change appreciably over the next several years. They built on GCP and migrating the whole service off would probably be a gargantuan task (how do you flip a switch and move all your compute overnight without hurting customer experience?). Staying with GCP allows Snap to maintain consistency of service while also buying time to build a transferable version of Snapchat that they could move to other infrastructure after the Google contract is over.<p>Investors on Wallstreet don't like seeing huge changes to company strategy too close to IPO. If GCP has worked for Snap thus far, it is far less risky to investors for Snap to keep on going "business as usual." It's better to overspend to guarantee certainty of service and business health over next few years than do a massive capital investment. Once Snap gets off the ground post-IPO, they can make longer term decisions about their infra.