> you either carry a shitload of capacity that stands idle 90% of the time
But costs less. THAT is the point. (A quad-core is almost there, dual hex xeons will provide 10x capacity. Not enough? Add more cores. At $2000/mo, the payoff is pretty easy in less than a year.)
Look: You guys have a business which is not yet breaking even. Perhaps that will fix itself with a bunch more customers. But as with all businesses, once your invoices start showing up on the CFO's screen, he is going ask to design you out. He will walk down to the office of his head IT geek, who is always trying to justify his salary, and say: What does it cost to take this in-house?
You think the IT geek is going to say: Oh no. We need to have that in the cloud.
I think perhaps you underestimate the magnitude of the difference between the lower and upper bounds of the capacity demands placed on a video transcoding service, and the difficulty of building one in the first place.
But costs less. THAT is the point. (A quad-core is almost there, dual hex xeons will provide 10x capacity. Not enough? Add more cores. At $2000/mo, the payoff is pretty easy in less than a year.)
Look: You guys have a business which is not yet breaking even. Perhaps that will fix itself with a bunch more customers. But as with all businesses, once your invoices start showing up on the CFO's screen, he is going ask to design you out. He will walk down to the office of his head IT geek, who is always trying to justify his salary, and say: What does it cost to take this in-house?
You think the IT geek is going to say: Oh no. We need to have that in the cloud.
Really?