When Snap filed its IPO paperwork last week, hidden with the minutiae of the S-1 was word of a five year, $2 billion contract with Google Cloud Platform to run its infrastructure. While that deal was somewhat surprising in itself, even more so was the fact that Snap felt compelled to list it as a risk factor for future investors.
"We rely on Google Cloud for the vast majority of our computing, storage, bandwidth, and other services. Any disruption of or interference with our use of the Google Cloud operation would negatively affect our operations and seriously harm our business," Snap wrote in the filing.
Snap reports that on average 158 million people use their products daily, generating 2.5 billion snaps. Snap uses Google Cloud Platform to deal with all of the compute resources this kind of intense daily activity requires.
Part of the issue for Snap is they have agreed to pay at least $400 million a year, regardless of whether they use the platform. Given the company's growth trajectory, that doesn't seem like a huge problem. Perhaps that's why Snap seemed more concerned with possible business disruption, but how big a risk are they facing? Let's take a closer look.
GCP downtime question
No business wants to be down, even for a minute, but when computers are involved, there will be times when that happens regardless of the vendor (or if you run your own data center). How much time are we talking about? In 2015 according to numbers from the cloud monitoring service CloudHarmony, Google's cloud products were down for a total of 11 hours and 34 minutes in 2015. As a means of comparison, Azure was down 10 hours and 49 minutes and AWS was an industry best with just 2 hours and 30 minutes of downtime.
To put that into perspective, there are 8,760 hours in a year. All of the cloud vendors did a pretty good job staying up. In addition, these downtimes are totals across all products and regions, so not every customer was affected by every outage.
Regardless, it's worth noting that having access to cloud services provided the resources to run a startup like Snap in the first place. Not only did it give it a place to run its software, it gave it the elastic infrastructure to deal with the variable and fast-growing nature of the business. Consider the capital investment it would have required to build the data centers to scale from zero to 150 million users in five years. It would likely have been prohibitively expensive.
So while being dependent on a single cloud vendor could be considered a risk, it's not as a big a risk as say trying to do all of this itself. At some point, if Snap continues its current growth rate, and continues developing new products and new markets, it might very well decide to use other public cloud vendors. If it reaches sufficient scale, it may even decide it's more cost-effective to begin building its own, but for a five-year old startup, using cloud services makes all the sense in the world. And, while that might present a modicum of risk, it has also allowed them to thrive and get to this point. That seems like a risk a startup like Snap would take every time.
Photo Credit: Roman Boed on Flickr. Used under CC by 2.0 license.