The problem is not that Google is unreliable. The problem is that the servers are not yours, and thus the services that depend on the servers are not really you

The problem is not that Google is unreliable. The problem is that the servers are not yours, and thus the services that depend on the servers are not really yours. Therefore, if you write software to match, or foster methodologies that depend on a particular product, you risk writing software or patterning activities around a system that can trivially disappear. You have no way of estimating the popularity of adoption (is this android?), or how essential this thing may be to internal Google teams (is this wave?), since it remains something of a trade secret, who is using what, and how much, therefore you can never know when or why Google might declare that its offering is antiquated or unprofitable to Google, or whether it will be deprecated, forcing a migration elsewhere.