Dyack Wrong on 'Cloud Computing' For Games (For At Least Three Reasons)

Denis Dyack of Silicon Knights, in a post at Venture Beat, claims that "cloud computing" will rescue the game industry from the same crash in revenue that the music industry has already undergone, and that film and (possibly) book publishing now faces.

To summarize his argument: linear media consists of a single, shareable file, and cannot be controlled in the Internet age; games are different each time played, and while so far we must distribute executables as single, shareable files, "cloud computing" means we don't have to do that any more. Instead, we can have the game actually running on a remote server, with the player connecting over the Internet, and his computer acting, in essence, as a dumb terminal, receiving and playing the data. Hence, nothing for people to easily copy, and therefore protectible IP.

It's such a wrong-headed argument that it's hard to know where to start, but to begin with an outline of my argument:

1. Dyack doesn't seem to know what cloud computing is, which makes the headline of the article and therefore the attention it is likely to receive somewhat silly; this fact does not, however, invalidate his basic argument.
2. He characterizes linear media as being "commoditized," which is wholly incorrect, and demonstrates a failure to understand where the value does (and does not) lie in non-game media.
3. To the degree that his argument is that server-provided services are harder to pirate than executables, his argument is obvious rather than some dramatic revelation as the result of the growth of clouding computing; and to the degree that his argument supposes a future of dumb terminals connected to servers running games, it's flat out wrong, for obvious reasons of cost-benefit analysis.

So I guess I need to go through each of these issues one by one.

Cloud Computing

Cloud computing is an emerging phenomenon because today, the cost-per-MIPS (million instructions per second) has been inverted. For decades -- basically since the inception of large-scale networks -- the rule has been that server-side MIPS are costlier than client-side MIPS. That's because client-side MIPS were provided by cheap desktop machines, and remote ones by mainframes or other larger and expensive servers. Thus, if you were, say, an investment bank running a trading system, you tried to push as much processing as possible down the tree to the individual user, reserving critical shared data for the server backend.

Today, large data centers can purchase, maintain, and install rack servers at a far lower cost-per-MIPS than individual users -- and at a far lower cost than achievable by small companies running their own, smaller-scale, server farms. Consequently, it's increasingly cost effective for small net businesses in particular to build their infrastructure to take advantage of "cloud computing" initiatives like Amazon Web Services or, say, the Cisco Unified Computing System, rather than hiring IT staff to manage their own server farm.

Specifically, "cloud computing" refers to a cluster of technologies that offer a) remote hosting and provision of computer services, b) in a scalable and load-balanced way, so that a spike in network or processing usage is handled invisibly and without administrative effort by the service provider, c) redudantly, so that the loss of any single server, or even cluster of services, has negligible impact on service provision, d) charging on a usage rather than per-CPU basis.

The advantages Dyack is talking about result from provision of gameplay from remote servers; these advantages exist even if the game provider is hosting and managing its own servers. Consequently, he is not actually talking about cloud computing at all, and the use of the term in his article is misleading.

Nor, incidentally, is cloud computing particularly relevant to a company like, say, Blizzard; given the scale of usage they receive, and their need to configure their server architecture to their own highly unusual (by the standards of other Internet services) needs, they are unlikely ever to transition to a cloud framework provided by another company. It makes business sense for them to run their own data center. From my perspective, Dyack's attempt to connect his argument to "cloud computing" is particularly annoying, because I'd love to see a reasoned argument for how and why cloud computing will and can affect gaming, as it surely will; I've yet to see any major gaming initiative that takes advantage of cloud computing in any major way. (Manifesto hosts our installers via Amazon S3, but this is a fairly minimal and obvious use.)

Commoditification of Linear Media

Dyack argues that linear media (music, film and books) are becoming "commodities" in the world of the Internet, and are therefore doomed to drastic decline in revenues, and that games, since they are different with each playing, are not.

Now here's the thing. A commodity is a fungible substance. Sugar is a commodity; every bag of sugar is chemically identical to every other, and the only reason to prefer one to another is price. And if you're buying brand-name sugar over the house brand you're an idiot.

But no one ever said "I want to listen to some music, whatever's cheapest." Works of arts are inherently not commodities, and never will be, regardless of how they are or ever will be distributed. Blitzkrieg Bop is not interchangeable with the Eroica, Red Harvest is not a reasonable substitute for As I Lay Dying, and Being John Malkovich will never be mistaken for Gone With the Wind.

What digital distribution changes (dratistically) is consumer perception of value. Not long ago, we generally perceived the items we bought as having value because of their physical substance. Thus the cost of books was largely set by the cost of printing ink onto paper, binding the paper into books, shipping the books to bookstores, and supporting the overhead expenses of the retailer. Similarly, the cost of music was the cost of a physical plastic disc with a metal insert with laser-generated pits burned into it that your device at home could render into audible music, along with all the costs entailed in delivering this physical item through the value chain into your hands.

I remember once at a Toy Fair having a buyer from a major retailer tell me that "games have no intrinsic value" -- a statement which both blew my mind as to its obtuseness, and to the degree that it was revelatory of consumer perception. A boardgame does not have value in the same sense that a 10' length of 2x4 does; its value is in the play, and the enjoyment and intellectual challenge you gain thereby, not in whatever bits of cardboard and plastic (or bits and bytes) you purchase in order to play it. The consumer perception of value in popular works of art was, in fact, an incorrect one; the cost of all that plastic and paper and packaging, the cost of the marketing and distribution and retailing efforts, is in fact friction. It's not relevant to the real value of the work -- nor, in fact, to its real cost, which is measured in the cost of creation.

What Internet distribution actually does is reduce the level of friction, disintermediate those who used to manufacture the physical object needed to embody the artwork created by others, and to some degree disintermediate the people who perform marketing, distribution, and retailing services. (Not as much in the latter regard as it should, perhaps, but that's a subject of another rant.) Consumers, rightly, perceive that if they aren't buying plastic and metal and cardboard, if they just want the work in itself, then the only people they really should be paying are the creators thereof, and the cost the consumer pays should therefore be much smaller.

Every mass artform has its publishers (or labels or movie studios, whatever you wish to call them). Publishers provide four services: development funding, manufacturing, distribution, and marketing. Manufacturing is now irrelevant. If your distribution is mainly through iTunes or Amazon, then you can go direct to them, and the chokehold publishers used to have on distribution, with their vast teams of sales reps calling on thousands of individual accounts, is no longer relevant. If technology makes production vastly easier and cheaper -- as it has in music and to a lesser degree in film, and not yet much in games -- then development funding is less critical. Marketing remains critical -- and yet, this suggests a future in which publishers, rather than taking the lion's share of the consumer dollar, cease to exist, or rather, become contract marketing houses, with the artist receiving a considerably larger share of a -- much reduced -- consumer dollar.

Dyack argues that linear forms are forced to offer their products at values increasingly close to zero by the forces of digital distribution; I don't think that's true. I believe consumers perceive value in such works, and (as demonstrated by the success of iTunes) are willing to pay what they consider reasonable rates for efficient delivery, and a positive shopping experience, for linear media. What's only beginning is the disaggregation of the intermediaries -- and while it's clear that, say, the music labels are, and will continue, to suffer, it's not at all clear that either artists or consumers will, in the end, suffer at all. Indeed, they may both benefit greatly.

And that's good. They're the ones who matter. Everything in between them is just overhead.

Why Dyack is Wrong: The User's MIPS are Free

Dyack argues that since gameplay is different each time, you can structure a system in which games are processed on remote servers, and therefore there is no single shareable set of bits that users can transfer to one another; therefore, they must continue to pay for gameplay; therefore, games have a sustainable business model when "commoditized" media do not.

Part of this is self-evident, and part of it wrong. Let's look at each part in turn.

Korea has the most vibrant market for MMOs of any country in the world. Part of the reason for this is doubtless purely cultural. But another part of the reason is that, until recent years, Korea did not permit the important of consoles from Japan (their hated foe, even today), and therefore console gaming has never amounted to much there; and part of the reason is that laws against piracy of software were never enforced in Korea. Thus, Korean developers could not make money doing either console or conventional PC games (the latter because they would be pirated); the only way to make money was to do a server-driven, online game, and charge for access to the server.

Thus, sure, if your game requires server access, you have a pressure point you can use to continue making customers pay. And for games like, say, World of Warcraft, consumers will understand that you incur substantial costs by operating those servers, and it's reasonable to pay for access to them.

But Dyack seems to believe that the basic business model is extensible beyond MMOs, which clearly require a server component, to, say, games like Oblivion. In his model, even a single-player game would run primarily on a remote server, in order to protect the code from replication, with the user's client doing little more than receiving and rendering data.

But you know what? Even MMOs push as much processing as possible down to the client, reserving the servers for handling multiplayer interactions and providing a central data store (the latter for security reasons and to prevent hacking). Last time I checked, the WoW client was in excess of 6GB, including all the graphics needed for play, with client-server communication limited to the bare minimum needed.

There's a simple reason for that; even if, on a cost-per-MIPS basis, server farms are now more cost-effective than desktops, the user has already bought the damn desktop (or laptop or whatever). That's sunk cost. And using the user's machine costs the service provider zero. It's still more efficient to push as much as possible to the client, because the cost of client MIPS is zilch.

I've seen any number of pitches for services where, say, graphics are rendered remotely and then dispatched to the user; this makes no sense, and never will. Why not have the client do that processing? And the bandwidth costs of streaming high quality interactive video to clients is extremely non-trivial; let the client do the rendering.

Moreover, how stupid are customers? Won't they realize that they reason they require Internet access to play your game, and can't play it offline, and can't own the executable outright, even though this is a wholly single-player experience that by no stretch of the imagination requires Internet access, is purely so you can make sure they're paying you?

Nope. Sorry. Single-player games, or small-scale multiplayer games where clients connect directly instead of through a remote server, are indeed more like Dyack's "commoditized linear media" than they are like MMOs. You can't, and shouldn't, force an artificial net intermediary. The argument fails.

In summary:

1. Dyack may think he's talking about cloud computing, but he clearly isn't.
2. No artwork is ever a commodity.
3. The need for access to a server to play provides business model protection only to a restricted category of games, not to games as a whole.