Atmel sells fabs

http://www.edn.com/index.asp?layout=article&articleid=CA6399448&ref=nbednn

steve wrote:
> http://www.edn.com/index.asp?layout=article&articleid=CA6399448&ref=nbednn
> 

Yes, that's the exit statement.
No mention of what products will be impacted ?


Here is the entry statement, 6yrs ago :

http://findarticles.com/p/articles/mi_m0WVI/is_2000_Sept_25/ai_65568508

Shows these stats, (at end Sept 2000)
Atmel's current fab line-up
                    Wafer size  Processes   Area (sq m)  Status
Colorado Springs, US  6in      0.5 micro m    41,805    Running
Heilbronn, Germany    6in      0.5 micro m   ~10,000    Running
Nantes, France        6in      0.5 micro m   ~10,000    Running
Rousset, France       8in      0.35           36,045    Running, but
                                0.25 micro m,            not complete
                                0.18 micro m
Irving, Texas, US     8in      0.18 micro m   60,385    Running from
                                                         Q1 2001
North Tyneside, UK    8in      0.35,          39,099    Running from
                                0.25 micro m             Q2 2001

-jg

On Fri, 15 Dec 2006 15:45:14 +1300, Jim Granville
<no.spam@designtools.maps.co.nz> wrote:

>steve wrote:
>> http://www.edn.com/index.asp?layout=article&articleid=CA6399448&ref=nbednn
>> 
>
>Yes, that's the exit statement.
>No mention of what products will be impacted ?
>
>
>Here is the entry statement, 6yrs ago :
>
>http://findarticles.com/p/articles/mi_m0WVI/is_2000_Sept_25/ai_65568508
>
>Shows these stats, (at end Sept 2000)
>Atmel's current fab line-up
>                    Wafer size  Processes   Area (sq m)  Status
>Colorado Springs, US  6in      0.5 micro m    41,805    Running
>Heilbronn, Germany    6in      0.5 micro m   ~10,000    Running
>Nantes, France        6in      0.5 micro m   ~10,000    Running
>Rousset, France       8in      0.35           36,045    Running, but
>                                0.25 micro m,            not complete
>                                0.18 micro m
>Irving, Texas, US     8in      0.18 micro m   60,385    Running from
>                                                         Q1 2001
>North Tyneside, UK    8in      0.35,          39,099    Running from
>                                0.25 micro m             Q2 2001

So they are dumping part of their .5u and .35u+.25u facilities.

Jim, there's an old thing I learned a long time ago.  You don't buy
your own stock back.  If the best you can do with your active company
cash is buy back your stock from passive investors, if you can't do
better with the money than those feeble expectations would have from
the money, you simply shouldn't be in business at all.  If worried
about LBOs, go into debt by investing in a new product line or
something.  But don't buy your own stock back.  It's money you've
pried out of their hands and you had better be able to make it work in
an active business better than they can as mere passive investors.  A
company _sells_ stock because it can use the money better than others
can, in some passive savings account.

This smacks of that, to me.  If a company making ICs can't use their
FABs as well as they can use the money they will get for selling them,
they probably shouldn't be in the business of running FABs at all.
This "Fab lite" policy worries me.  What about you?

Jon

Jonathan Kirwan wrote:

> This smacks of that, to me.  If a company making ICs can't use their
> FABs as well as they can use the money they will get for selling them,
> they probably shouldn't be in the business of running FABs at all.
> This "Fab lite" policy worries me.  What about you?

While I don't fundamentally disagree with anything you've said, I'd
like to point out that there is more to making chips than just a bunch
of equipment and some guy feeding it design files. There is recipe
magic and tribal knowledge that might be impossible to transfer to
other facilities.

In that vein, some fabs may be staffed solely to provide the
manufacturing magic that supports a product line that isn't profitable.
Some of these facilities might be in Atmel's portfolio because of
acquisitions; i.e. the money shouldn't have been spent.

The implication from this announcement is that Atmel has excess
production capacity and/or some very underperforming product lines,
neither of which are immediate signs Atmel should get out of the
business; merely signs they need to understand their markets better and
manage their acquisitions with greater care.

However I am worried. I'm designing in AVR micros and DataFlash memory,
and a couple of other divisions use Atmel's RF transceivers. While I
think AVR will be "core", I'm not so sure about DataFlash (which has
lots of competition) and the RF stuff (which is niche-y and expensive).

Jonathan Kirwan wrote:
> On Fri, 15 Dec 2006 15:45:14 +1300, Jim Granville
> <no.spam@designtools.maps.co.nz> wrote:
> 
> 
>>steve wrote:
>>
>>>http://www.edn.com/index.asp?layout=article&articleid=CA6399448&ref=nbednn
>>>
>>
>>Yes, that's the exit statement.
>>No mention of what products will be impacted ?
>>
>>
>>Here is the entry statement, 6yrs ago :
>>
>>http://findarticles.com/p/articles/mi_m0WVI/is_2000_Sept_25/ai_65568508
>>
>>Shows these stats, (at end Sept 2000)
>>Atmel's current fab line-up
>>                   Wafer size  Processes   Area (sq m)  Status
>>Colorado Springs, US  6in      0.5 micro m    41,805    Running
>>Heilbronn, Germany    6in      0.5 micro m   ~10,000    Running
>>Nantes, France        6in      0.5 micro m   ~10,000    Running
>>Rousset, France       8in      0.35           36,045    Running, but
>>                               0.25 micro m,            not complete
>>                               0.18 micro m
>>Irving, Texas, US     8in      0.18 micro m   60,385    Running from
>>                                                        Q1 2001
>>North Tyneside, UK    8in      0.35,          39,099    Running from
>>                               0.25 micro m             Q2 2001
> 
> 
> So they are dumping part of their .5u and .35u+.25u facilities.

Note that was at Sept 2000, so they will be more advanced now.
Heilbronn, IIRC, is their Automotive products.
The Nantes fab I think went about 1 year ago.
This is a bit more up to date
http://www.cadence.com/company/success_stories/atmel1_ss.pdf

> 
> Jim, there's an old thing I learned a long time ago.  You don't buy
> your own stock back.  If the best you can do with your active company
> cash is buy back your stock from passive investors, if you can't do
> better with the money than those feeble expectations would have from
> the money, you simply shouldn't be in business at all.  If worried
> about LBOs, go into debt by investing in a new product line or
> something.  But don't buy your own stock back.  It's money you've
> pried out of their hands and you had better be able to make it work in
> an active business better than they can as mere passive investors.  A
> company _sells_ stock because it can use the money better than others
> can, in some passive savings account.
> 
> This smacks of that, to me.  If a company making ICs can't use their
> FABs as well as they can use the money they will get for selling them,
> they probably shouldn't be in the business of running FABs at all.
> This "Fab lite" policy worries me.  What about you?

  Well, it is the opposite trend to someone like Microchip. We might all
grumble about the PIC core, but you have to be impressed by Microchip's
ability to make money from what is not leading edge stuff, and to
make money on one of the lowest ASPs in the industry.

  The other end of the scale is someone like SiLabs (fabless)
and companies like Zilog, and even Freescale are also moving "fab lite".
I believe all Zilog's new Flash devices are foundry.

  Where it bites, is when your designs use a chip that was in
the FAB that closed, and it used a process that has not been moved to a 
foundry :(

  Some recent-memory examples of this, were Philips Coolrunner cplds, and
AMD SPLDs. In those cases, the fab-rug was pulled, terminating products 
earlier than normal.

-jg

"larwe" <zwsdotcom@gmail.com> wrote in message 
news:1166153215.051641.146840@80g2000cwy.googlegroups.com...
>
> Jonathan Kirwan wrote:

> While I don't fundamentally disagree with anything you've said, I'd
> like to point out that there is more to making chips than just a bunch
> of equipment and some guy feeding it design files. There is recipe
> magic and tribal knowledge that might be impossible to transfer to
> other facilities.

This is so true. For such a high tech process it depends on really skilled 
people to make it work.

"larwe" wrote:

...
> However I am worried. I'm designing in AVR micros and DataFlash memory,
> and a couple of other divisions use Atmel's RF transceivers. While I
> think AVR will be "core", I'm not so sure about DataFlash (which has
> lots of competition) and the RF stuff (which is niche-y and expensive).

The RF stuff went in the sale to E2V,  although I have had a written 
statement from E2V saying that they will continue the RF side (I've designed 
into several products), I'm not too convinced. Time to look elsewhere for 
longevity I think.
Jim 

larwe wrote:
> Jonathan Kirwan wrote:

> However I am worried. I'm designing in AVR micros and DataFlash memory,
> and a couple of other divisions use Atmel's RF transceivers. While I
> think AVR will be "core", I'm not so sure about DataFlash (which has
> lots of competition) and the RF stuff (which is niche-y and expensive).

It is a concern.  I've been burned with
fabless flash.  Might be good to look at
ST's nand flash.

On 14 Dec 2006 19:26:55 -0800, "larwe" <zwsdotcom@gmail.com> wrote:

>Jonathan Kirwan wrote:
>
>> This smacks of that, to me.  If a company making ICs can't use their
>> FABs as well as they can use the money they will get for selling them,
>> they probably shouldn't be in the business of running FABs at all.
>> This "Fab lite" policy worries me.  What about you?
>
>While I don't fundamentally disagree with anything you've said, I'd
>like to point out that there is more to making chips than just a bunch
>of equipment and some guy feeding it design files. There is recipe
>magic and tribal knowledge that might be impossible to transfer to
>other facilities.
>
>In that vein, some fabs may be staffed solely to provide the
>manufacturing magic that supports a product line that isn't profitable.
>Some of these facilities might be in Atmel's portfolio because of
>acquisitions; i.e. the money shouldn't have been spent.

Hard to say.  These details _may_ apply, or not.  And I wouldn't know,
because I'm not privy to the details.  But looking from the outside,
we cannot belabor them, either.  All we can do is see what the big
picture is looking like.  I've boiled it down to the fundamentals.
That means I've also boiled away a lot of reality in doing so.  But
this is the nugget that stands out like a sore thumb when I am done
with it.

>The implication from this announcement is that Atmel has excess
>production capacity and/or some very underperforming product lines,
>neither of which are immediate signs Atmel should get out of the
>business; merely signs they need to understand their markets better and
>manage their acquisitions with greater care.
>
>However I am worried. I'm designing in AVR micros and DataFlash memory,
>and a couple of other divisions use Atmel's RF transceivers. While I
>think AVR will be "core", I'm not so sure about DataFlash (which has
>lots of competition) and the RF stuff (which is niche-y and expensive).

Each of us has to act like a spider in a web and 'feel' the vibrations
from this.  How we each respond is our own business.

Here's my "spider sense" at work:

One of the really sad things here is that we are talking about 1300
employees, as well.  It takes a lot of effort to get employees into a
company, give them desks, walls, phone service, as well as all the
time invested in the various relationships.  All this goes away.  A
company should prefer to keep their employees and take risk at going
into new markets with them.  If the company is obviously overstaffed,
then these employees obviously have the time to apply themselves to a
new business opportunity.  And they are already hired, already have
all the time investment in them, have desks, and so on.  Way ahead of
competition, that way.  The only question is whether or not you
believe in yourself as a company enough to go for it.

Tektronix is a case in point.  Luckily, this is a case where I got a
chance to talk with one of the board members on a personal 1:1 basis
for a few hours circa 1985.  I had only left the company about two
years before and in that time I'd noticed that their employee loyalty
was way down.  When I was there, a boss need only merely suggest the
idea of staying a weekend or an evening to get something resolved and
folks did it in droves.  That had changed and I was bothered by that.

Tek had over $650 million at US Bank in 1982 (10k report), undedicated
to anything in particular.  Not salaries, etc.  It was just an asset.
They had no long term debt, no short term debt -- to speak of.  They
owned property around the state of Oregon, including Redmond, Bend,
and elsewhere, as well.  And they were worried about at least two
things.  One was an LBO.  The other was the fact that their gross
revenue per employee was an atrocious $65k.  This, at a time when IBM
was at about $125k.

So the board decided that they needed to buy their own stock back with
all that cash sitting in the bank.  That would reduce their exposure
as an LBO (not so much, but some.)  And they also decided to take the
non-risky approach to solve the revenue/employee by reducing the
number of employees.  It was either that, or increase their revenue.
(Revenue isn't directly controllable, but getting rid of employees
is.)

They offered early retirement, as I learned from meeting with this
board member, to some 810 long time employees.  The board expected to
see about 200 accept, give or take.  They got 795 acceptances.  Just
about everyone who'd been their for 20 years or more.  Gutted right
away.  In less than one year.

In addition, they now would fire people they didn't have a ready slot
for instead of moving them around to see if they could fit.  And in
the process of going from 24,000 employees to some 14,000 (memory
fades here on exact numbers, but this is ballpark) in a few years,
they also lost their internal reserve of attitude.  Employees didn't
consider themselves an integral member of a community working together
where they all sunk or swam together, but from a much narrower view
where either side might pick up and leave the other to fend for
themselves.  Two weeks notice became the norm, on both sides.

The right stand to take, in the face of being an LBO target and in the
face of poor revenue/employee, especially in the case of Tek where
they had ready access to more than half a billion dollars entirely
within their control and probably the ability to borrow another
billion and a half almost at a whim, and with the excess unused land
they owned, they could have walked into any business line they wanted
to -- I don't care what it is -- with 10,000 excess employees already
hired and in their desks, two billion dollars of fluid, working
capitalization, and moved into anyplace in force to bring their
revenue line up.  The shift from excess cash in the bank to being in
debt would have immediately removed them as an LBO target.  And they
would be doing what any company should be doing with all those
resources available.

Any company would have given their eye teeth and their own children
for an opportunity like this.  And in this case, Tek would not have
needed to go begging for this kind of capitalized entry.  They would
have been in full control, flush with cash, already bursting with
employees ready to go and in their desks.  Going into business just
doesn't get any better than that.  Period.

What did happen was wrong on its face.  It demonstrated little more
than their own lack of confidence in their abilities to manage their
own business.  (They weren't buying back their stock because they were
taking themselves private, they were doing so simply because they
wanted to get rid of an excess of cash that might be used against them
in an LBO attempt.)  An active business manager offers stock because
they believe in their ability to create more with that money than
others can.  Buying back stock simply to spend cash is a statement
that you can't think of a better way to use it.  And that is sad.

They needed to dodge an LBO and they needed to improve a key ratio.
They had the cash, no debt, the ability to borrow, and the existing
staff already hired up and in desks.  In that situation, they chose
use their cash to buy back stock and to turn their backs on their
already huge investment in existing employees.  This neither removed
the LBO risk (they still weren't in debt) much nor did it move in the
right business direction (revenue growth.)  They could have shouldered
into any business they wanted to grab, with cash aplenty to spend, put
themselves into debt (substantially reducing LBO risks) and worked in
the right direction of growing their revenue by entering some market.

If anything does, this tells me that they didn't believe in
themselves.

In talking with this board member, he largely agreed with this
analysis and instead tried to explain to me about the conservative
nature of some of the board members -- professor from Reed college,
and so on.  Oh, well.  Folks would have begged for a chance like they
had.  And they completely blew the opportunity.  (At the time I was
talking with him, this wasn't 20/20 hindsight.  Tek was still in the
middle of restructuring and downsizing and selling off parts of its
business.  But time has born out this analysis, anyway.)

...

I still imagine the bigger picture here is significant.  Atmel is not
taking what looks like some modest step -- but is making a sea change
that represents an internal belief about itself.  And while I agree
completely that details matter and that it could be other than I
imagine here, my spider sense is tingling.

Jon

Jonathan Kirwan wrote:
> One of the really sad things here is that we are talking about 1300
> employees, as well.  It takes a lot of effort to get employees into a
> company, give them desks, walls, phone service, as well as all the
> time invested in the various relationships.  All this goes away.
<snip>

  I think the norm is to sell the fabs as 'going concerns', (or
similar to the Nantes sale) - so that means not all the jobs
go away.
  Indeed, as you see when Atmel brought Tynside 6 years ago,
it came with a loading commitment from Infineon.
  I think that is also common, and in both the buyers and sellers 
interests in the short term at least.

-jg