For SpaceX the actual costs they experience are quite low compared to their competition. For example, they don't use rad-hardened CPUs, nor do they use vxworks. This alone saves them millions of dollars. However, it makes sense for them to offer their services as close to the competition's prices as possible, because it maximizes their profits. You see the same thing on their launch services.
> However, it makes sense for them to offer their services as close to the competition's prices as possible, because it maximizes their profits.
This is an important lesson that I learned the hard way with my first startup.
Not only does it maximize their profit, but it also helps reduce the risk of setting off a price war. In my case, we went into the ISP industry in '95 with a price about 1/3 of the larger competition in Norway. What we did not realize - being completely fresh to business - was that these guys did not price their service that way because their costs were that high, but because the market at a time (which was not yet regular consumers) could bear it and their margins were high.
So when we and a couple of other small ISPs launched with so much lower prices in autumn of '95 it too two days, and the big incumbent ISPs followed suit, and eventually pushed the prices much further down. We stuck it out a year before we sold our dial up customer base and focused on hosting and consulting when it was clear that the big providers at that point had seen that their big price drops were well timed for a big growth in the consumer segment and decided they were willing to sustain big losses to take market share.
It took several years before the market segment returned to profitability for anyone, and by then most of the small providers were out of the picture...
You really don't want to rock the boat too much if you can instead build up a war chest, unless you're extremely well funded and/or know for a fact that you can not only beat the competition on price, but can also survive if they decide taking losses for a while is worth it if they can get rid of you.
Thanks for the correction, that's interesting. It's a testament to the utility of vxworks, I suppose, which costs a few hundred thousand dollars per client. Even so, not using rad-hardened CPUs still saves them millions, especially when you factor in backup hardware, test units, etc.
