David Perry - Part One

Well, there are two different approaches - one, you build a network that's gigantic and hope people come, and the other is to build a network based on demand. If you advertise for traffic - imagine I was to go in all the movie theatres, or put banner advertising all over Brighton, for example - and all these people show up, there better be servers for them. If there isn't, then all that peak traffic that came in is going to be wasted.

You can't have that, so you have to build a network that's ready for the peak traffic, and there will be lulls in that traffic, and you'll have servers eating power... it's quite a complex equation. If you were making that order, how many would you order on day one? That's the peak traffic solution, that's what OnLive is going to have to solve.

It's a little bit old school thinking to be honest, because it's a brute force solution that's very, very expensive. It's going to cost... I don't know.

Let me give you an analogy - you just signed up to this service and you're paying a subscription now. In that subscription you try to log in during peak, and it says "Sorry, our servers are full" - how happy are you going to be?

That's a cancelled subscription, and they can't afford that - it's such a high cost. To get a paying player on the internet costs a lot of money.

So for us we're the exact opposite - we're doing a scaling solution. Say we only have one server in the whole world, and that server can deliver about 3500 new players to a publisher during a month, based on the number of hours, if they do one hour each.

It's going to be up to the player to decide if they want to continue playing that game, but assuming it's good, you'll have a healthy conversion from playing to wanting to continue.

So just like banner advertising on the internet, there's only a certain amount of views available, a certain amount of inventory, and people end up bidding for that inventory - it's how Google works, it's how Facebook works. You bid for inventory - there are 3500 players available, how much do you want to pay? The publishers can decide - it might be too expensive for them, it might be really worth it.

So we then say: "My God, they ordered 10,000... we need more servers." And we just keep on buying servers - but we only buy them based on demand. Why does that work? Because it keeps the cost down for everybody. We have no servers running, and I didn't spend USD 150 million, with the interest on USD 150 million burning away as I hope people are going to show up.

That's basically the model - every time we hit maximum capacity, we order more servers.