A few weeks on from my spot price post, we're still seeing elevated prices. While typically in these instance types we see 70% savings, we are seeing only about 50% now, at least in my usual zones and AZs.

I found this article which seems to indicate its bigger customers looking to trim bills under inglation pressures, which seems feasible but also a bit strange. I mean, if your workloads are flexible, why wouldn't you have availed yourself of this sooner? Did some businesses have THAT much money to burn? Because these aren't tiny blips, but large and lasting trends.

Like, I could see this make more sense if we saw inter-day fluctuations, but the trends I'm observing don't show that. Remember, spot means you can be interrupted at any time with 2 minutes of warning time, so your workloads should be flexible. This mostly precludes stateful apps. And if you are looking for real savinhs, you'll put your app in sn auto scaling group, which should size up and down throughout the day, depending on load. So why isn't there trends up at start of biz and drops at close? Even in our global economy, There are lots of regional or even city-centric businesses, who have set customers with traditional hours.

My only plausible thought is that this is an internal cost-cutting measure by AWS -- to me, spot has been the place where AWS keeps there "float" of physical servers, to make sure they have plenty of spare capacity. But if they allow that slack to dwindle a bit, it puts upward pressure on spot prices which is what they have been using to regulate demand. By more carefully managing this float, its possible to carry less equipment and also make more money (as spot prices rise), without having a formal price increase. As there is no transparency on how spot pricing is determined, its hard to say, but my thoughts seem feasible.