When OnLive was announced, I was in the ranks of the doubters. Bluntly - I'm still right there, and while it's clear that some of the company's technology is impressive, I still don't think that this is the shape of any future that's on the cards in the next few years.
The reason is that while OnLive and other such services have proven that their technology can work, assuming certain fairly expensive criteria are met in terms of server farms and network quality, they have not, to my mind, answered the more relevant criticisms - the economic ones.
The basic economic theory behind Cloud gaming goes like this. Right now, the consumer pays for a piece of hardware which can play games - a significant up-front cost - and then buys the games for it. He or she is also probably paying for a broadband connection, since gaming has become an increasingly (but by no means exclusively) connected experience.
The technology will find application, but the services presently envisaged are doomed to failure.
Cloud gaming moves the costs around. In this new model, the publisher or provider buys the gaming hardware - all of it - and installs it en masse into data centres. The consumer then effectively leases that hardware through a monthly fee, or perhaps through inflated prices on the games that they buy on the service, allowing the provider to recoup their large investment in hardware.
The idea is that since consumers have no up-front cost, it'll open up a market that doesn't want to invest in expensive gaming hardware, and that the provider will be able to benefit from economies of scale which individual consumers buying hardware cannot tap into. For publishers, meanwhile, it's a piracy-free environment, since code is never downloaded onto the users' computer and thus can't be copied.
There are, however, immense problems with all of those assumptions about the Cloud gaming model. Firstly, the trend towards increasing power and decreasing cost in consumer electronics continues apace - and one of the most powerful recent trends on that front has come in the form of GPUs, with vastly powerful GPUs, capable of superb 3D graphics, becoming relatively commonplace in multi-purpose consumer hardware where once they were the preserve only of expensive gaming PCs and consoles.
As such, Cloud gaming is between a rock and a hard place. Top end users who have the cutting edge graphics hardware in their PCs won't want it, because there will always be a trade-off in terms of latency and graphical fidelity, which they won't tolerate. Meanwhile, users willing to tolerate that trade-off are also likely to be willing to put up with the less than cutting edge graphics offered by their existing computers - or, indeed, their phones or their tablet devices.
In other words, the idea that having to buy expensive hardware is a barrier to gaming is increasingly obsolete. The proliferation of consumer devices that can play great games is putting an end to that - making it hard to see where the mass market for Cloud gaming will come from. Some of the mooted ideas for business models might turn that around, such as the concept of having blockbuster games available at launch for a flat (but presumably high) monthly fee, but this idea too is flawed, not least in that getting publishers to come on board with such a revenue-slashing concept is a tricky proposition.
The mooted economies of scale, too, don't quite add up. Consider comments from GameStop this week about streaming console games to their customers over a Cloud gaming service - something which would involve literally placing racks and racks of consoles into data centres. In theory, you could save money because you don't need a console per user - if one user plays between 5pm and 8pm, and another logs in at 8.30pm, you only need one system.
In practice, of course, you need to have enough consoles (or blades, or VMs, for the PC titles) to meet peak demand - so if 80% of your users tend to log on to play games on a Wednesday evening at 7pm, you need to have 4 consoles for every 5 users on the service. Not a great economy - and you can't even use the clever trick beloved of other Cloud services, which spread "peak load" out around the clock thanks to servicing many different timezones. No such luck; Cloud gaming needs servers physically located close to the user, so the machine that one player uses after dinner in London isn't going to be any use to you after dinnertime in New York.
As for being piracy-free, this truly is a pipe dream. Sure, you can't make a copy of a game off a streaming service - but that assumes that you're going to release a game that only ever appears on a streaming service, and is never actually launched in any downloadable or physical media form. It's meaningless for your game to be "unpirateable" on OnLive if someone can rip the DVD of the retail version and post it on BitTorrent - and the idea of Cloud gaming reaching a market share large enough to justify a Cloud-only launch is science fiction at this point.
The technology is impressive, and it may have some limited applications - I suspect that Gaikai's creators, who are largely focused on offering game demos streaming over the 'net, are keenly aware of that. The notion of full-on Cloud gaming, though, remains an expensive bust, and I can't help but wonder if heavily funded companies such as OnLive aren't aiming more at an exit strategy involving acquisition for their technology than at any really significant market share.
Besides - if we want to think about how gaming can take advantage of the whole concept of the Cloud, then this is the wrong way to go about it. The fact is that game designers, not publishers or distributors, wrapped their heads around the uses of Cloud technology long ago. When you play an MMORPG, regardless of where you log in, you're playing your own character in your own game world - because it's all stored in a data centre.
The client does what the client does best - renders nice 3D graphics and animations. The server does what the server does best - stores and processes a ton of data and squirts the results down the line to the client, a line which, of course, is a hell of a lot less demanding than the vast bandwidth and low latency needed to stream crisp HD video. Increasingly, this approach, pioneered by MMORPGs, has found fertile ground in many other game genres where persistent, server-stored characters and profiles are increasingly popular.
That's how the Cloud will be used in games in years to come. The idea of removing 3D rendering tasks from the client is a nonsense, at a time when 3D rendering chipsets are increasingly cheap and powerful, but when bandwidth available to consumers remains hugely variable and, increasingly, capped by ISPs. The technology will find application, but the services presently envisaged are doomed to failure. GameStop should really be looking elsewhere for its salvation.
If you work in the games industry and want more views, and up-to-date news relevant to your business, read our sister website GamesIndustry.biz, where you can find this weekly editorial column as soon as it is posted.