[jzl]

On top of all that, Google has said that they plan to roll out over 7500 edge computing clusters just for Stadia, all for tackling the latency problem. (Not clear if those clusters can be put in existing Google facilities or not.) Having so many edge clusters runs counter to the ability to amortize costs by ensuring fully subscribed hardware usage. Each edge cluster will be as costly as the required peak usage for each location. And I imagine Stadia will have a very distinct pattern of peaks and troughs of usage relative to local time.