"selling, general and administrative expenses (SGA) are predominantly fixed figures in regards to units sold."
but four paragraphs later it explains away Tesla's losses by saying
"The increase in SG&A is to be expected considering Tesla's ramp up of the Gigafactory and its massive growth in general."
Look, either it's independent of sales, or it's dependent. You can't have it both ways.
Also, saying "we don't lose money for each car because I want to deduct the cost of administration, marketing, sales and other general expenses, since those don't really count as expenses" is just disingenious.
What that article might have a point in saying is that Tesla won't lose an additional $20 000 if they sell one more car. I.e. their marginal loss is not $20k. But marginal loss is a completely different quantity from the actual loss per car; it's actually related to the derivative. If we say the loss/profit function has a given value at some point in time, the derivative can of course be completely arbitrary (positive or negative).
Edit: marginal loss equals loss per car only if your loss is linear in sales with a proportionality constant of one. No-one argues this is the case, in fact, everyone agrees (and Tesla is betting heavily) that the second derivative of this function is greater than zero over a large interval (thus it can't be linear). That article is a textbook example of a strawman argument.
Analogy: say going to the zoo costs $50 for your family of 4. But it's a family rebate, so anywhere from 3 to 5 family members costs $50 total. The cost per person for you is $12.50. But the cost of one more ticket if you were to have another child, the marginal cost, is $0. That is, however, irrelevant to you today, because you can't have another child today.
http://seekingalpha.com/article/3968035-tesla-lose-money-eve...