I’m surprised he took this take, because honestly, mobile would be the biggest winner there, but it’s square-peg, round-hole. Most people wouldn’t use their phones for console games if they could, because it’s a different form, with a device that has a different purpose. The way most people play mobile games & console games are very different: sporadic short sessions during the day is different from a 2-3 hour gaming session.

Maybe it can help close a gap between high end & low end users, but I don’t think streaming gaming is going to have a significant impact on the total addressable market. We’ve already seen platforms not succeed in that space. Why would it work now? If you hope it addresses the low end, why would you think they have budget room for another subscription?

I don’t see Strauss’ take panning out here.

The “advent of streaming,” Zelnick said, is something that Take-Two believes is “right around the corner, in terms of having something that’s low latency and really works well for consumers. Machines that weren’t game machines before will become game machines.”

I mean, if you’re comparing it to, say, a local PC, it’s higher latency, less configurability, less reliability, and games will probably stop being available at some point. I am also skeptical that we have the bandwidth to move everyone to unicast high-res video streaming at once. You maybe gain on a troubleshooting front and it might be somewhat cheaper on a hardware front due to shared costs and people not using the systems 24/7 (though I suspect that people would want datacenters as near as possible due to latency, so you can’t be sharing hardware with a country who isn’t awake at the same time, say).

If there’s a real case for streaming gaming, I think that it has to be for mobile-only users. Mobile hardware is just too power- and cooling-constrained to be competitive with larger, fixed-in-place systems with wall power, so using a phone as a display for a remote server might make more sense, if it lets users without a PC play games that they wouldn’t otherwise be able to play.

I’m also really skeptical that the current memory crisis is going to be that much of a factor. Yeah, okay, so it costs more to buy a PC in 2026 and 2027…but it also costs more to buy a server to do streaming gaming on. That server might get higher capacity utilization, so it might benefit the streaming providers somewhat relative to everyone having a PC, but I’m pretty skeptical that that’s going to be a sea change.

I think that outside of mobile, if streaming games are going to be the future, that you’d see a major shift to console games doing it first. They have limited expandability, so don’t give up some of the things that you would if moving away from a local PC. Basically, instead of a PS6, say, yould just provide PS6-class games to PS5 users, with the PS5 being a thin client for a game running in a datacenter. But we haven’t seen a major transition of that sort.

There are genres for which the latency isn’t such a big deal. I think that it’ll always be a poor match with twitch games. If you’re playing a point-and-click adventure, though, latency may not matter much.

But my bet is going to be that in 2029, at the end of his three year period, most PC games continue to be local.

I think that if people really feel priced out of newer hardware during the memory shortage, they’re more likely to just keep playing less-hardware-demanding games locally than to transition to streaming games.

more economical to fly roundtrip from the UK to buy an MSRP GPU at Nvidia’s QuakeCon booth than it is to buy one anywhere online

That was for a limited number, for a high-end GPU, and for an Nvidia GPU. Most people are not using 5090s.

EDIT: There are some changes that could be made to game engines, if you’re doing it at the game dev level and developing a new game, to try to mitigate streaming latency, stuff that you can’t do if you’re trying to retrofit streaming to existing games. So, like, existing multiplayer latency-sensitive network games typically make use of client-side prediction. Like, my client will make some guesses about what other players are doing, since it won’t actually know until a packet reaches the server from them and then the server sends a packet to me. It might (reasonably) assume that the player will, say, keep the same buttons that they presently have down. So it makes that assumption and renders the world as if that assumption were true…but it remembers the last time it got a state update from that player, and it retains all the game world state up until that point. If it turns out that one of its predictions was wrong in a way that invalidates its assumptions, then it recomputes game state from that point in time. I see what looks like a short “glitch” maybe, but most of the time, latency appears to be reduced. There are variants on this, but this is roughly how things work.

It should be technically possible — though not normally done today that I’m aware of — for a streaming game server to make server-side predictions about my input and to render the game world and start streaming input to me as if that had been my input. If it gets an update from me and turns out that it was wrong, then it rolls back that world state and recomputes from that point. Similar approach to client-side prediction.

Prediction isn’t magic. It can’t make latency simply go away. You’d tend to see a lot more brief “glitches” this way after you push a button (or, if it’s a mobile phone, probably tap on the screen). But…you could potentially reduce the apparent impact of latency to some degree.

If you add thin client software to the mix that can do things like render simple visual effects and play some preset sounds, where the game instructs the thin client how to respond in advance, then when you push a button, you could potentially have the game software immediately play a “button pressed” sound and then start a visual effect like the screen flashing or something like that, even if it takes more time to get a rendered frame reflecting the input back from the other side. That could potentially improve perceived responsiveness.