Something the article misses is complementary services. A lot of companies don’t care at all about X vs 2X for server cost, but if you lack certain features aro

Something the article misses is complementary services. A lot of companies don’t care at all about X vs 2X for server cost, but if you lack certain features around say managed Spark job execution or serverless container deployments, it’s a total deal-breaker. AWS, Azure and GCP tend to have the widest coverage of complementary services, along with all the other stuff like volume discounts and credits. I believe in many cases it could be perfectly rational to pay 2X more on just the server portion. I also think support is another big issue, but GCP makes me second-guess myself, since they are the third largest provider but their support system is (literally) to just say “fuck you, read our upsell-laden docs.” A good example is the way the GKE SLA page mentions a bunch of beta features of kubernetes will void your SLA credits, when in reality Anthos is not remotely close to being ready for production use by most teams and they have no choice but to rely on beta features in GKE. Multi-cluster ingress is a good example of this - “just switch to Anthos” as a proposed solution is literally equivalent to saying, “fuck you.”