Google Stadia will support “a variety of business models”

Part of that transition will be developers getting used to having the power of Google’s cloud infrastructure, Harrison said. For decades, game developers have had to be “device centric,” Harrison pointed out, designing games for specific and limited home hardware specifications. “With cloud gaming, particularly the idea of compute being sharable across multiple CPUs in a data center, now this transition to gaming being data-centric is going to a really fundamental shift,” he said.

The developers who are devoted to taking advantage of this will be able to implement everything from “distributed physics” to “complex multiplayer going from hundreds to tens of thousands in a very sophisticated world,” he said. In Stadia multiplayer games, “every change that I make to my world can be instantly, in microseconds or less, be distributed to every other client… You can’t do that with a discrete box.”

That sounds a lot like the kind of “revolutionary” features Microsoft promised would be possible on Xbox One through integration with its Azure cloud. The fully destructible environments in Crackdown 3’s multiplayer mode were perhaps the highest-profile implementation of Microsoft’s promise so far, but they ended up being a bit of a letdown in practice.

Elsewhere in the talk, Harrison also started crafting lofty visions of Stadia games with “conversational understanding” of spoken player commands delivered through the controller’s built-in microphone. Google’s research in AI and machine learning could come into play to help enable games with NPCs that “talk back in a way that’s contextual,” he said, making use of data center storage for an “enormous database” of potential conversation options built in real time. Similar machine learning work could also one day help Stadia developers build content more cheaply, with less manpower and time, Harrison suggested.