After having thougth a lot, I arrived to several conclusions :
The "per device" cost is 525€ + shipping
The fixed cost for the whole project (R&D+Loan) is 190k €

The main way to significantly decrease the total cost of a device is to decrease the fixed cost, by selling more devices. I mean, if 1000 Neo900 are sold instead of 500, each customer pays 190€ instead of 380€. In many countries (with 20% VAT), it means more than 230€ saved for each customer. So what about finding a partnership, or launching a crowdfunding (as wicket proposed a few posts above) ? With a crowdfunding, the "per device" cost can be adjusted with the main goal of the project.

One more thing is the pool of donators who will not buy a device.
I'm sure not every donation will end with an order. And I'm sure there will be a few "new" customers who didn't previously donate because they were not aware of this project.
There were 423 donations. How many of them were at least 100€ worth ? Let call this number X.
As soon as there are more than 500-X orders without rebate, each new order (ie without rebate) means for the smallest rebate remaining (ie not used yet) :
- either it will be used, so (380 - its worth) € is gained for the project / could have been removed from the fixed costs
- or it will not be used, so its worth is gained for the project / could have been removed from the fixed costs

What I mean is if we can accurately evaluate the number of buyers (especially new buyers) and it's high enough, the price can fall for everyone, which means more potential buyers.