Jesus, the figures posted here are so far off its not even funny.
I've tried to avoid posting in this thread, but the sheer amount of misinformation posted here is absurd. How do I know this? I do this stuff for a living working for an even larger organization than 95% of all megacorps out there.
Chances are, Blizzard is using a TON of very high end servers. Even with a 25% discount, these servers run in the neighborhood of $1M (see IBM P690 for an example). Now we have to assume they are clustered, this brings in approximately $15k per cluster license. Chances are, they have close to 40 of these servers at the start. Thats $40+(15k*40)= 40.6M in server costs alone. This is just for the hardware. Maintenance on this class of servers runs in the neighborhood of $10k a year. So 40 * $10k = $4M a year in maintenance fees. now that we have the server hardware, we need a place to put all this stuff. These days, that means a SAN. After buying all the fibre, SAN switches and storage devices (they will need disk and tape), we are looking at nearly $4M in costs up front for a project of this size. These types of setups have a shelf-life of 3-5 years if planned properly (which is impossible).
Now, your average software developer earns abouy 60k a year. Blizzard probably has probably had more than 100 people working on this project total (managers, supervisors, programmers, artists, technical writers, QA people, sys admins, technical support, etc). For the sake of argument, lets assume 45k a year is the average pay of all the people involved. So 100 * 45k = $4.5M for one year.
Monthly costs for bandwith are likely to exceed $30k (this includes the connections and additional costs their ISP will forward on to them). Chances are they have two primary OC-3 connections and two secondary ones. This means they are spending $360k a year in external networking fees. Now they have to have a way for the servers to talk to each other. You average high-end Cisco switches run $50k - they are going to need at least two, $100k then. Plus they are going to need redundant routers as well. You average Cisco enterprise class routers run $30k, you're going to need two of these as well = $60k. Now they are going to need firewalls. Cisco PIX firewalls run $15k each, they are going to need four (two for each incoming line set). Thats another $60k.
Now consider they need to have a place to put all these things in. Buildings aren't cheap. Figure $3M for a building with a fully built server room (they don't just throw these things on concrete floors). PDU's (power units) run about $125k each. They will need at least four of them to power 40 servers. Thats another $500k. Yearly power costs for such a setup will likely run $80k a year (power companies pre-bill for large environments).
Now, we have a partial lists of all the costs (I know I'm ignoring some things like property taxes, debt, marketing (advertising), installation costs, money taken by parent company, etc). we also have to ignore the potential for doubling our network applicances (most places plan initially for full redundancy) however, their number of appliances could be 2 or more times what I listed depending on their utilization. Figure the first two years of development did not include the server costs as listed (they would probably have used servers that were already present for development work), they also wouldn't need the outside connections until around 1 year before BETA. The building would have been built early because it takes time to build them.
We also know WoW has been in development for around 4 years.
Now that we have at least some of the costs and timeframes laid out, we can come to a pretty rough (but more accurate than anything else so far) estimate of their current costs to date.
Servers: 40.6M
Maint: 4.0M
SAN: 4.0M
Staff: 4.5M X 4 = 18M
Network: 580K
Building: 3M
This yields us a total of: $70.18M in up front costs to date (estimated) for Blizzard. Now we also have to assume that the $360k a year for their OC-3s will keep recurring. Also, they will be upgrading and replacing servers as necessary. Lets assume four servers per year will be replaced with another server costing a similar amount, thats another $4M a year in recurring costs (very conservative estimate). Their staff costs are recurring, another $4M per year. Remember they have to pay taxes on all their profit too since they are an American company, assume $20% corporate taxes (includes state, Federal and any other local taxes, anyone in accounting knows this is much higher). For the sake of me not wanting to keep parsing these out, lets assume thats their full list of recurring yearly costs.
Now assume Blizzard charges $12.95 a month (a good compromise between high and low figures and a likely target). We can safely assume they will have 400k active accounts per month. This gives us $5.18M a month in income. Now we have to take taxes out first, they come to 1.036M per month. So they are bringing in about $4.144M a month in profit. Now we figure out their yearly profit: $49.728M But wait, we forgot to take into account their expenses!
$70.18M total up front.
$8.36M for years two+ for recurring costs.
This means that in year two (the first year of release) they have $78.54M in expenses already accrued.
So, they have made $49.728M in profit in year two (first release year, second year of having the major expenditures). This means they will still be in negative earnings for post-release year two by $28.812M, release year three will be their first profitable year. But this is assuming they don't open a Europoean, or Asian market as well, then they have to pay taxes on the profit they make over there, and as we all know, Europrean and Asian taxes are insane compared to ours. Also, anyone who has worked for a company owned by another can testify that parent companies love to take massive bites out of the profit of their child companies.
Is Blizzard sitting on a cash cow? Yes, but chances are good they aren't pocketing much on this project until years 3 to 4 after release. Most software companies build their business model for endeavors like this expecting to not make any profit at all until years 3 or 4. Years 1 and 2 of post-release are considered as the years to make up for previous losses.
All these figures are estimates, but very good estimates based on my experience in the industry as well as knowing how well other MMORPG's have fared in the past (Everquest still holds the MMORPG record at 550k active subscriptions at any one time - most have fared nowhere near as well, usually hovering between 200k and 400k actives).
Believe it or don't believe it, but know Blizzard has absolutely no intention of making a profit on this game until years 2 and 3 - if even then considering they have to develop, and ship expansions in order to keep their playerbase. It's basic business theory.