Steve Dekorte just posted a thought-provoking little piece about the new cloud-based gaming service, OnLive. Steve’s math in the post is off by an order of magnitude, unfortunately. 10 Gbit/s is actually 10,000 Mbit/s, so if each HDTV subscriber is using 5 MBit/s, then a 10 Gbit/s uplink can sustain 2,000 concurrent HDTV subscribers, not the 200 that Steve derives. In the video, Steve Perlman also talks about how they are not going to get just a single 100 Gbit/s connection to a datacenter, as is more typical, but rather around 10 10 Gbit/s connections from various providers in order to lower per-subscriber latency. Given an aggregate bandwidth of 100 Gbit/s per datacenter, that would allow for 20,000 concurrent HDTV subscribers for a single datacenter at the best.

Lets assume that bandwidth costs are $5/Mbit/month, which is probably on the low side but maybe achievable. 10 10Gbit/s connections would then cost them $5 * 10 * 10,000 Mbit/s = $500,000 USD per month per datacenter. This means they need $25 in gross revenue per subscriber per month just to break even on bandwidth, assuming all subscribers are using HDTV. That doesn’t count any other costs, either. Ouch.

But lets make it more realistic. Lets say that the HDTV-to-standard definition TV ratio is 1:1. So that means that they can sustain 10,000 concurrent HDTV subscribers whilst also sustaining ~33K subscribers using standard definition. With ~43K paying customers per month, they will only need around $11.50 per subscriber per month to break even in bandwidth costs. This is much more doable for them, since I think the most they can get away with charging and still drive adoption is $30/month, maybe $35/month. Factoring in manpower, hardware and R&D costs, however, still leaves them most likely taking a loss on the monthly subscription.

The problem with the service is that HDTVs are cheap these days (and getting cheaper all the time) and the electronics manufacturers are pushing them very hard. They are selling quite well, so HDTV percentage will continue to rise as percentage of their subscribers. An HDTV subscriber costs OnLive 3 1/3 times the amount of a standard definition subscriber in terms of bandwidth, but there’s no way they can actually charge HDTV users 3 1/3 times the monthly subscription to make up for it.

All is not lost, though. Its typical today for console manufacturers to lose money on each console sold and then make it up on game licensing revenue. In this regard, If OnLive is charging $30/month/subscriber and the total cost to OnLive is $50/subscriber/month including manpower, hardware, etc, then that would yield a total loss of $20/subscriber/month, or $240/subscriber/year. The issue here for OnLive is that Microsoft and Sony lose that money only once for the lifetime of the console, whereas OnLive keeps leaking the cost every month the subscriber stays with them. If Microsoft is losing $200 per console sold, then OnLive would lose the same amount per subscriber after just 10 months of loyalty. The console guys are also amortizing that loss over 3 years (the expected lifetime of the console) while OnLive just continues to eat the cost year after year. So, if you take that accounting into account, Microsoft is losing $100 per year per console but OnLive is still losing $240.

Financially, its looking as if OnLive is in a bit of a worse position than the traditional console manufacturers. With the consumer market moving to the much more expensive (for OnLive) HDTV and bandwidth costs at the level they are today, they will have to find a way to sell many more games and rentals on their service than is typical for the console market. It does seem to me as if they are already thinking this way, given Steve’s mention of “episodic games” in the video, where one would would pay multiple times for pieces of a game. Personally, I’m hoping that they do figure out a way to do this, or reduce bandwidth costs, or both. I would love to see this kind of service succeed.