A blog post about why private cloud is failing got some traction, prompting some follow up articles. Regardless of the statistical value in the actual survey, an interesting observation was:
Of the 140 companies Bittman surveyed, the most common reason for dissatisfaction (noted by 31 percent of respondents) is that too much emphasis was placed on cost-cutting, not on providing agility in creating, spinning up and down capabilities as needed.
If you are looking to set up your own infrastructure through buying/leasing hardware and co-locating in data centres, it's relatively straightforward to calculate a massive cost saving when you compare compute, storage and networking to public cloud pricing. The common pattern is starting on public cloud and then migrating to colo as your company scales. I wrote over a year ago about how Moz saved over $500k/month by doing this.
However, the key here is that the savings are on raw compute, storage and networking. As you start to broaden your definition of "cloud" to include other IaaS and SaaS products, the cost savings diminish. Can you build your own equivalent of Google's BigQuery? Can you build your own equivalent of Amazon's build and deployment pipeline management products? Yes you could, but can you match their features and pricing? Probably not.
The whole point of outsourcing is to hand over part of your business that you do not consider "core", technology that you don't need to own, to a specialist. A dedicated vendor can do a much better job in every aspect - features, updates, security, support and of course, pricing. This is the whole model of SaaS! With a broad definition of "cloud", these are not commodity products and so it's no wonder that projects trying to replicate AWS or Google Cloud are failing.