Drawing any conclusions from G-Cloud sales data is difficult at the best of times for a number of reasons:
1) Spend numbers are largely 'self-reported'
2) Because of 1, reporting tends to lag reality such that reported sales in any given month tends to go up over time (i.e. recent months' spending tends to be under reported)
3) Reported sales of AWS and Azure may or may not reliably include sales via re-sellers and partners
4) Comparing sales of Microsoft (i.e. Azure and Office 365) against sales of AWS makes no real sense since you are not comparing like with like - if you could split out just Azure spend, then that would be a reasonable comparison with AWS spend - but I suspect that isn't possible for some of the reasons above.
As to, "why doesn't the UK gov build its own cloud?". Ha! Can you imagine. If cloud was simply a place to run VMs that might just about be a possibility. But if you think that is what cloud is, I suggest you go and look at what the likes of AWS, Azure and Google actually offer these days. There isn't a cat in hells chance of government producing anything in the same ball-park... not even on the same planet.