All rights reserved by @TwitterMore than one and a half years back, I made a tweet asking people
When startups grow, do they stay in the cloud or move to their own datacenter?
I got multiple responses with some saying that they will continue to stay in the cloud as they don’t have to incur Capex and they can stay lean and others saying that they will move to their own datacenters. I was always curious to know the thinking that goes on among those who matter in such successful startups. If they move away from the cloud, why do they want to do it and why successful startups with high growth don’t see the cloud in the same way they saw when they were just two people with dreams to change the world upside down.
There is another clue to this question today with an announcement by Twitter that they are planning to build datacenters in the state of Utah and move towards the end of this year.
Later this year, Twitter is moving our technical operations infrastructure into a new, custom-built data center in the Salt Lake City area. We’re excited about the move for several reasons.
Well, this is not all that surprising for me because we have seen many such companies move to their own datacenters after achieving a certain level of growth. In fact, Twitter is not even moving from cloud to their own datacenter. Rather, they are moving from a managed hosting to their own datacenter. This is a pretty normal evolution. But, what surprised me was one of the reasons they gave for their move.
Second, Twitter will have full control over network and systems configuration, with a much larger footprint in a building designed specifically around our unique power and cooling needs. Twitter will be able to define and manage to a finer grained SLA on the service as we are managing and monitoring at all layers.
This justification can come from any company moving from the cloud to their own datacenter too. If every successful consumer focussed company is going to demand full control over network and system configuration and they want to be the person defining the SLAs, it is a big trouble for the public clouds. When consumer focussed companies are so obsessed with control and SLAs, there is no way enterprises are going to embrace public clouds anytime soon. Either we need a change in mindset of the companies in both the consumer space and the enterprise space or public cloud providers have to do more to lure them towards clouds. I would love to hear from startups, SaaS companies, enterprise IT folks, etc. on this topic. Please add your comment below or contact me through this form.
PS: Thanks to William Vambenepe for getting me started on this topic.
Additional Reading:
CloudAve is exclusively sponsored by
Krishnan Subramanian

Krish dons several avatars including entrepreneur in exile, analyst cum researcher, technology evangelist, blogger, ex-physicist, social/political commentator, etc.. My main focus is research and analysis on various high impact topics in the fields of Open Source, Cloud Computing and the interface between them. I also evangelize Open Source and Cloud Computing in various media outlets, blogs and other public forums. I offer strategic advise to both Cloud Computing and Open Source providers and, also, help other companies take advantage of Open Source and Cloud Computing. In my opinion, Open Source commoditized software and Cloud Computing commoditized computing resources. A combination of these two developments offers a strong competitive advantage to companies of all sizes and shapes. Due to various factors, including fear, the adoption of both Open Source and Cloud Computing are relatively slow in the business sector. So, I take it upon myself to clear any confusion in this regard and educate, enrich and advise users/customers to take advantage of the benefits offered by these technologies. I am also a managing partner in two consulting companies based in India. I blog about Open Source topics at and Cloud Computing related topics at

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.