AWS was reusing their existing infrastructure and expertise and opening it up to 3rd parties.

I'm unclear why Amazon needs a custom game engine. They must have some plans to enter this area, maybe make their own VR headsets?

Not really. They mostly built a second parallel system for aws and moved to using it several years after launch. Yes they used knowledge that existed in house but it was a totally separate project and mini company

We imagined that Amazon AWS was letting us use a few of their servers that were idle because it was not the Xmas rush. This was a convenient narrative at the time and it explained the competitive pricing. Reality was that it was a different gig.

I believe that was the goal, but like any large endeavor it took a while to get going.

In a word: Kindle.

Amazon et al may anticipate VR becoming HUGE in the future, so they'd best start creating top-tier easy-access infrastructure & tools for supporting it. Each will want to be THE go-to for customers, just like e-book customers mostly go to Kindle (and a few competitors provide solid alternatives, like iBooks); wasn't that long ago most of us watched the e-book reader market explode onto the scene with lots of competition between B&N, Kobo, Sony, Amazon, Apple, and anyone else that could create a viable e-ink tablet. Similar happened with streaming video, with just a few set-top boxes surviving (Amazon FireTV, Apple TV, Roku, and a few others). All these need a solid, cheap, desirable content-creation platform and distribution system.

Next up is VR: when you're going to render content, what distribution platform will you build for? how will you sell it? what's the go-to store for buying virtual worlds? Bezos & co will want you to choose Amazon Lumberyard &/| Sumerian (and Tim Cook & co will persuade accordingly toward Apple rOS, and Facebook is figuring out how to position Oculus, etc).

But look at all of the other additions on top of AWS now - you're not just getting bare metals, you're getting an entire stack from the machine to the speech output. They have the bare metal, they're just ramping up the value adds in verticals they know can be huge (plus, building... and I hate the word... synergies with Twitch).

All of that CryEngine expertise was probably going to go down the toilet - am going to assume it's been comparatively cheap for them to get a AAA game engine into their locker.

I can think of a couple of reasons why Amazon might be interested in VR, from displaying 3D views of their products to constructing a virtual mall for browsing.

VR online shopping could be a massive thing one day.

Reminds me of the virtual reality access to the market in the movie Valerian.

Amazon owns Twitch and presented the Twitch Launcher not too long ago, which is basically a Steam competitor. Game engines are probably gonna be given for free to devs to build games specifically on this platform.

Given the hype about running games virtualized in AWS, no it is not unclear. Amazon wants a game engine native to their cloud, rather than a PC engine running on virtualized hardware, so people can start thinking really big about VR or games, rather than "one machine at a time" models.

I think this is one of those things that nobody expects that will be obvious in hindsight. Offloading rendering and game logic/AI to a massively parallel server network will allow MMOGs to break new ground in world complexity and persistence, and sell subscriptions to more people that can't afford dedicated gaming machines.

I think it would be doable for most current MMORPGS which aren't very timing sensitive so even if you have a couple hundred ms of lag it could still be playable (especially if the game is designed around this limitation).

VR however is one application where network lag would clearly be totally unacceptable. Can you imagine having even 50ms of lag while turning your head in a game? It seems like out of all the applications of "cloud gaming" this would be the least suitable one. It's not even about the game feeling laggy and floaty, it's about not getting sick while playing it.

I guess they could render a huge FoV for every eye and let the client do the actual tracking to mask some of the latency, but I'm not sure how you can make any translation work that way.

I'm not saying it's impossible and it'll never happen but if that's really the objective it seems they're doing it the wrong way around. Shouldn't they first develop "cloud gaming" infrastructure and then move on to the particular niche of VR?

Ah, yeah I was thinking more generally than VR. The R&D efforts at Oculus to reduce latency from GPU to viewport indicate that network rendering isn't going to work unless you anticipate and buffer the set of every frame the user could possibly require shortly ahead of time; technically possible for low-fidelity graphics I guess but very inefficient (although now that I'm thinking about it you would have to include all possible states of all non-static game assets as well, which creates a combinatorial explosion; I don't see how it could possibly work).

Regarding cloud gaming, Nvidia has a service in beta, but Amazon seems well positioned for a monopoly in this space (for both developer and consumer use cases) and I'm surprised they haven't been pushing hard for it. I've been using Paperspace heavily for cloud gaming for the last six months and it's a very compelling value proposition. All things considered I'd still rather have a dedicated custom PC, but I can't justify spending $2,000 up front for something that is already good enough at $0.60/hr on demand. You can imagine how big the market could be if instead of selling loss-making consoles you could just push a smart TV app that connects to a cloud gaming service and eliminate the console upgrade cycle. It would also be much easier to lure users from other networks if they don't have to invest $300 in a console to switch. I don't know if it would be a losing proposition for Microsoft and Sony but it could be an entry point for a leaner competitor to get a leg up on the incumbents.

Also, they can sell to "casual" gamers. Rent a character and just go. Don't pay for more than you use.