[Arm-netbook] preorder and roadmap clarification

[lkcl luke]

hi folks,

just had a question off-list about how the various stages will work,
so i thought i'd best clarify.

* the NREs (non-recurring expenses) are $2000 per "board development
change".  ok with each successive set, costs can be reduced a bit:
e.g. if all the holes are in the same place on the PCB, and the board
has the same connectors, and it's the same size, you save $500 for not
having to redo the PCB "tooling".  basically, though, it's $2000.  at
this stage, there will be *no* software: you'll get a board, and
that's it.  we will all need to pitch in, munge the lichee v2.6.36
code and get on with it.  (i also have a copy of the u-boot GPL source
code which i will also make available closer to the time).

* assuming that the factory gets things right first time (it could
happen: this is a cut/paste jobbie from the existing Reference Design
after all) and the alpha boards are all ok, we can go straight into
beta with *no* NREs.  that means that the 2nd run would have more
people, but still no OS images etc.  orders for 5 to 50 units are
kinda ok at this point *if* you are a software developer and are
basically happy to get involved doing u-boot, debian-installer, ubuntu
images etc. etc.

* "stable" means, approximately, "kernel's done, u-boot's done,
debian-installer's done, openembedded recipes are done: download,
boot, go".  this is the point at which orders for 1k, 10k and above
would be fine.

now, as the NREs will need to be covered on the alpha stage, it means
that the first few people, of whom there are about 30, will have to be
paying about $100 per board.  i am reluctant to even let 30 people
take that risk: i'd far rather it was 5.  i will see what we can do
with the factory, see if there is any way to do a "pre-alpha" phase -
but i will only raise this with them once they are actually underway.

the second phase - beta - those NREs are _gone_.  so, at this point,
the more people who pitch in at the beta phase, it's basically a PCB
cost + component cost (zero PCB NREs).  at this phase, if there are
100 units in one batch, then let's take some guesses... 100of CPU
pricing would be ooo... *guessing*... $15.  so let's imagine that the
BOM would be $40.  if there are 1,000 units in one batch, then the
pricing would be *guessing again* $30.  at 10,000 units, it would be
$20.  at 100,000 units, it would be close to that magic $15 figure.

so far, if the price was $30, we have 48 people who have said that
they would be willing to pay that cost (at beta only - my
order_processing.py isn't yet sophisticated enough to take into
account if they weren't able to buy at "alpha" stage).  at $25, we
have 53 people.  ok, table:
beta 40 people @ $40
beta 48 people @ $30
beta 53 people @ $25
beta 56 people @ $20

for "stable", we have:
stable 41 people @ $30
stable 61 people @ $20
stable 121 people @ $18

so you can see, these would not be anywhere near the levels of 10,000
or 100,000 pricing - we _do_ however have a couple of people who have
said they'd be willing to do 1k stable orders, which is something that
everyone else could ride on the back of, if nothing else.

but as far as 10k and 100k pricing is concerned, that's where i need
to let _you_ guys handle the software, getting things working on the
alpha and beta CPU boards, and i focus on the next phase, which is
getting ready-made production-quality samples into the hands of
mass-volume clients.

this is all very inter-dependent stuff! :)  and i'm just really proud
of everyone who's helped get things this far - the perorders,
double-checking the linux kernel stuff.

l.