Thursday, July 30, 2009

Kodak CIS Business In Trouble

Rochester Democrat and Chronicle published few notes from today's Kodak Earnings Call. The company CEO Antonio Perez acknowledged that Kodak's new digital cameras announced this week do NOT use Kodak-designed CMOS chips. And Kodak likely will not put much money into that sensor business. Perez said, “It’s a tough time to make investments in a business with projections pretty poor for the next few years.”

But Kodak is not going to let sensors “die on the vine,” said its spokesman David Lanzillo. “We continue to see a lot of potential with the technology,” he said, adding that Kodak continues to look for business partners or other options for its sensor business.

Update: Seeking Alpha published Kodak Q2'2009 Earnings Call. Here is the quote related to the CIS business status:

Ulysses Janice - Buckman & Reed:

On another subject, CMOS, I saw your new video camera you announced. It’s using a 5-megapixel CMOS. I assume it’s your CMOS.

Antonio M. Perez:

No, it’s not ours, no. It’s not ours.

Ulysses Janice - Buckman & Reed:

Not yours?

Antonio M. Perez:


Ulysses Janice - Buckman & Reed:

How is that project going?

Antonio M. Perez:

CMOS is one of the businesses that is in transformation. If you look at the industry, you know the industry is under distress, as you know. If you look at the comments of the large users, such as Nokia, you know, Motorola, Samsung and all the others, they all predict significant declines for next year so it is a tough time to make investments in the business, you know, whose projection is pretty poor for the next few years. So we are still looking for what is the best alternative for the shareholders as far as monetizing those assets.


  1. It makes Kodak acquiring OV possible now.

  2. From the EK balance sheet, that is impossible.

  3. or OV acquire Kodak

  4. Finally CMOS sensor are projected to make a large increase in average resolution sold and Perez sees a decline?? Maybe CMOS sensors are not Kodak's cup of tea, but they are not declining. With leadership like this is there any doubt that Kodak continues to declone.
    What happens to the Kodak/Motorola sensor supply deal?

  5. isn't the decline he mentioned a decline in unit volume?

    also if you take a lesson from the DRAM business, high unit volumes can still be occuring with the market price below manufacturing cost.

    Many of the same players that are making DRAM are also making CIS. They are accustomed to losing huge amounts of money for extended time periods so that they can get a chance to print money when things swing around again. If you can't ride the troughs to carry you to the next peak, you have no business participating, or at least that is how it looks to this observer.

  6. Part 1 of 2

    If you want to know why Kodak (and other cmos supplier like it) are in trouble, please read below.

    From my experience, for a CMOS sensor company to be successful, continual engineering support, marketing and sales are usually more important than the technology side. I have dealt with all the major cmos sensor suppliers in the past (and some not so major) and have some observations. I am an engineer by training but have been a cold blooded business man for most of my professional career, so I will be approaching this from a business point of view.

    Most CMOS vendors, especially large corporations do not know how to sell a sensor even if their lives depended on it.

    The most common problem is that CMOS sensor vendors do not have intimate knowledge of the whole camera supply chain when they come knocking. If we look at this from the cell phone level, the cell phone people want a discrete camera module that they can use like any other electronic component. They cannot use a sensor no matter how good if it is not delivered in a module form. They need the camera delivered in a form that incorporates optics, packaging, mechanical, electrical aspects, and some kind of evaluation platform for camera testing and also some kind of incoming quality inspection platform during mass production. The cell phone mob also need direct support from the sensor vendor when the colors don't "feel" quite right (which is 100% of the time, image tuning sometimes even goes on during mass production!). It is pretty clear by this point that the camera market is a multi-disiplinary field and everyone in it has to be at the table working together in order to satisfy the cell phone people (the real end customer).

    Looking at this from a module level, the module maker needs to know who to turn to for good quality optics, and what the sensor vendor is planning in terms of module specs. Simple things like height, width, length, aperture and lens composition trends need to be known beforehand and preferably planned by the sensor vendor and the message delivered to the cell phone maker so when component selection time comes around, sensors are not kicked out of the race because some party has not done their homework resulting in something that doesn't quite fit the specs. Getting kicked out for a trivial engineering spec not related to the sensor must be the ultimate insult to the hard working folks designing and fabbing the sensor parts.

  7. Part 2 of 2

    If a sensor vendor's upper management has a "build it and they will come" mentality, then they will not sell any sensors, EVER. This is the kind of business where the crowd designing the silicon needs to get down and dirty, and mingle intimately with all parties way down in the supply chain in order to deliver what the real end customer wants.

    There is only 1 company I consider to be good in this regard, and the winner is Omnivision. Disclosure: I do not own OV stocks or receive any compensation from OV nor have I had any monetary or business dealings with OV for the last few years.

    No one in the business compares to OV. Despite OV's drawbacks like their sensors sometimes not working as advertised on the first cut or sometimes changing the secret sauce without informing anyone, they are the king of supply chain relationships in this business and they alone define the industry with everyone else following in their footsteps (often unsuccessfully, sometimes imitating OV so poorly that they shoot themselves in the foot)

    On tier 2 we have SETi, Pixart, Himax Imaging. These are the people who can imitate OV quite well and provide some unique value

    On tier 3 we have Micron, Samsung, Magnachip, ST, Toshiba, Pixelplus, Sony, TASC. There are the people who are either too big to care about the CMOS businses or run the CMOS business like their DRAM/Flash business units. Then there are some who try to imitate OV and shoot themselves several times over.

    Then there are some who are completely hopeless, Kodak, CMOX, Panavision, Foveon. These are the people who don't know how to imitate OV even after someone gives them all the answers on how to start in the business. These people deserve to be flushed out of the market for their lack of business acumen.

  8. amen to the two part explanation of the market dynamics. I see exactly those issues all the time.

    people that buy cmos image sensors in volume want to buy a solution, not an engineering project. Their engineering projects are the products in which these sensors are placed and placed in module form.

    just like pc companies don't buy loose dram, they also don't buy loose image sensors. neither do cell phone companies.

  9. Bold accusations... Let me answer on them from the sensor designer's perspective.

    - Module and lens integration: True, 6 to 10 years ago there were designs disregarding the module and lens considerations. Newer designs mostly account for 4 things:

    1. Array is mostly centered, so module footprint is minimal
    2. Pads located on 2 sides and die shape is made in such a way that the module footprint can be made square and minimal area (COB case module)
    3. CF, ulens and metal shifts are optimized for a specific lens which is the manufacturer's recommended lens
    4. Power dissipation is spread evenly over the die, or, at least, a reasonable effort is made a achieve this. Then the module thermal design can be simplified, to an extent.

    - Sales and support. Talking about 1st tier customers, such as Nokia, Samsung, Sony-Ericsson, LG, Motorola, I do not think they complain about lack of support from the sensor vendors. Some image sensor manufacturers are not build to support smaller customers. Omnivision customer base, on the other hand, has many small customers. May be this is the reason their organization is somewhat optimized for that.

    - About specific companies you've mentioned.
    Panavision is a small company. I think they are about 15 people now and were 9 people not long ago. Having so limited base it's hard, almost impossible, to think about everything and give good support.

    CMOX and TASC were quite small organizations and also had some design problems.

    SETi, well, it's a mystery to me how they can give good support. They are relatively small organization too, so they must have hard limitations. May be you were just lucky to be among selected customers for SETi?

    Kodak might seem big, but their CIS group is small too. As you see Antonio P. does not intend to invest in this.

    Same for Foveon - they are small and split between different projects.

    If you think about it, there is a lot of different tasks to be done in parallel in sensor design:

    - Pixel and process design at device level
    - Pixel optics design, including the metal, ulens and color filter shifts matched to different lenses
    - Analog design: readouts, ADCs and such
    - Digital timing design with a lot of operation modes, different I/Os, etc.
    - Image processing algorithm development and implementation (huge task by itself, not to talk about fitting the existing algorithms to a new pixel versions, which change all the time)
    - Characterization and tracking of hundreds of different optimization lots on the fab
    - Software tools for demo and customer use
    - Productization activities (reliability, process tolerances, yield improvements, statistics, etc.)
    - Sales, support, field applications

    Add few different sensor generations running this pipeline simultaneously. Add also administrative overhead.
    Now, try to map all these tasks to 9 or 15 Panavision employees and it's easy to see why they are "hopeless" in your definition. Same for Foveon and, probably, same for Kodak, even though they are bigger than Panavision.

  10. I am sure things have changed a lot in the past 8 years, but Omnivision used to have very poor customer support. We tried hard at PB to deliver a "whole" product but even then it was difficult with only 130 people. A few years ago I heard Micron had continued the whole product support tradition so I am surprised that they would be listed now in the 3rd group. Actually, I sort of have to doubt the universality of this assessment.

    Nevertheless, the general discussion of supporting the "whole product" and the fact that modules are required these days so the handset guys can just sort of plug them in is fairly valid.

    When I buy my handset I want the whole thing to work together, and not "some assembly required".
    It all flows up (and down).

  11. On tier 3 we have Micron, Samsung, Magnachip, ST, Toshiba, Pixelplus, Sony, TASC. There are the people who are either too big to care about the CMOS businses or run the CMOS business like their DRAM/Flash business units.

    Maybe Eric can expand on this, but I always thought that Samsung and probably Toshiba operate their sensor businesses pretty independently of everything else. Do does the sensor division report up to a guy who's also responsible for memory?

    Also, nothing was said about STM. Don't they supply whole modules?

  12. Part 1 of 2

    Cold blooded business man again.

    I'm a demanding individual but also a straight talker. I did not intending to accuse anyone. I apologize if my post offended anyone.

    As mentioned before, I feel that sensor companies need to mingle within the industry. image sensor and EF has also mentioned that a lack of people power to perform a multitude of tasks seem to hamper their ability to provide all round support. I think it is a fair assessment and a business reality. Here is the real kicker, most non-techical people (and these are the majority you will be dealing with constantly on the business side of things) do not understand or care how much sweat and tears goes into the product. They only want to pay a limited amount of dollars to get a component that lets them churn out x number of handsets on time with minimal risk to themselves. I don't expect technical people to be responsible for the business side, but certainly the business types or upper management within a sensor company should be taking care of this.

    I feel that consistent winners in the sensor race are the ones who go out of their way to make external alliances and deals, leveraging external parties and resources outside their organization to achieve a better levels of support for their end customers.

    Alliances I'm talking about might be

    1. Fab/technology (pixel, process, BSI, microlens, etc)
    2. IP providers (EDOF, image processing cores)
    3. EV Board providers (evaluation platforms, either in house, or external)
    4. Packaging/testing (wafer level, chip level)
    5. Optics (lenses, wafer optics, etc)
    6. Actuators (VCM, liquid, stepping motor, mems, etc)
    7. Module makers (COB, Flipchip, SMT, COF, you name it)
    8. Sales and Distribution (preferably with some kind of front line technical support for module/cell phone companies and more importantly mass production customer support)
    9. Other chip vendors, who in combination enable various markets (cellphone, digital cameras, video cameras, embedded notebook camera, standalone PC camera, security systems, game machines)

  13. Part 2 of 2

    Now, lets talk about time frames.

    Rewind back to 2000, when the CMOS sensor business was still in its infancy. As a module maker or system maker, It was possible to accept some risk and take up some of the burden of producing a working module in order to rush the next big thing into the market. The sensor could be hugely off-center, the sensor designer could know nothing about the packaging of the die, but that was ok, various companies were willing to take up the slack and package the sensor despite it being layed out not suitable for certain packaging requirements. The module makers can screw up, the optics were not quite right and everyone else in the value chain was allowed to screw up but were still able to design their modules into the cellphone despite weird outer dimensions or having numerous passive/active components protuding on the camera module. No problems, we'd just change the shape of the cell phone slightly to compensate.

    The situation I mentioned above persisted until around 2003, then suddenly companies like OV upped the ante and started to knit a more coherent story out of all the various industry players in the supply chain. When this happened, cell phone companies suddenly realized they were working too hard all along and demanded full service from their suppliers. You will also find at this time that cellphones got REALLY thin. I have no doubt someone in the sensor industry got quite a lashing from Motorola, demanding that they produce a suitable module for their RAZR phone.

    Ever since 2004, the barriers of entry for the CMOS sensor market have become higher and higher annually as the technology side and the business side have become entrenched into camps. In addtion to dealing with process node shrinks, race to the smaller pixel and smaller die sizes, the sensor companies also needed to cultivate their own loyal supply chain in order win business. To put it simply, the business became a game for grown-ups, and if you were a kid on the block, you'd better find a niche that the adults weren't interested in.

    Back to the present day. If I were a cellphone maker now, I would not accept any risk associated with a camera module . This mean everything contained within it, the sensor, lens, the packaging method, any patent issues, reliability issues, image quality and final pricing of the module. I would also not lift a single finger to help out my suppliers because there are people who do a supurb job at giving me what I want. Even if I wanted to help out a startup, I would not be able to justify to my boss why an unproven solution was chosen over an established one unless the value was monumental.

    Since there have been a few new CMOS sensor companies established in the last couple of years, I hope my comments can help anyone who wants to break into the field. The ultra large volume markets are hot, crowded and existing players all have huge portfolio of products and allies to call upon to get the job done. Tread carefully and you may be able to gain a beach head in the industry

  14. Anyone care to speculate what Perez means by "monetizing those assets"

  15. To cold blooded business man,

    I don't think you put those sensor vendors in right segment.

    To me, there are only 2 segment:
    1) for teir one handset makers (like Nokia, Samsung...)
    The players should be STM, Toshiba, Aptina, Samsung, Sony, and "half of OV"
    2) for 3rd world customers including China
    OV,SETi and Galxycore (to me, Pixelplus is a past. Pixart has no chance. Himax has the good team from ESS but, sorry, still has no chance.)
    3) ex-handset market
    OV (good at NB, though is lossing its advantage)
    Aptina (start to get in NB)
    Sony (still strong in security market)

  16. I think Perez made it very clear, he is backing out of the non-profitable business selling low end, low asp, low margin handset cmos imaging market where noone is making money these days.

    They will continue to receive proceeds from their IP about $300 million annually for sometime however.

  17. To cold blooded business man

    God bless you!
    I hope that your comments are being read by the managers of what you call hopeless companies (Kodak, CMOX, Panavision, Foveon). None of these companies have the scale to compete and they bring no real value. However, as you mention, there may be a place for image sensor startups as niche players.

    Now related to Kodak…..
    Two years ago Kodak announced that all of it's camera manufacturing had been outsourced to Flextronics. Does Kodak have a say about what image sensors should be used for its cameras? I hope not. If they do and they prefer not to use their sensors... well that is an interesting way of supporting your own business.

  18. How much do you think that Kodak’s CIS business is worth? Why would anyone buy them?

  19. The main asset of Kodak CIS business might be its patents. Some of them are quite fundamental and might have good payoff.

    On another matter, CMOX is out of business for 2 years or so - don't build on its managers to draw any conclusion from this discussion.

  20. 130 people at Photobit is a lot of people. Did revenue support that many people? Does Kodak even have that many people dedicated to CMOS imagers/whole product?

  21. 130 is not really a lot when you are supporting a number of products. And I mean support - with sales/mktng, FAEs, product qual, reliability, FA, software, etc. etc. Plus we had ongoing contract work and niche/custom products. So, we were busy. The revenue certainly supported that many people and we had a reasonable gross margin. As I posted a few days ago, PB Tech was nicely profitable but PBC was running net negative due mostly to R&D expenses. This is typical in any start-up with a steep ramp.

    There was an article in the June 23, 1997 LA Times Business section about the relationship at that time between Photobit, Kodak and Motorola. PB was doing design work for Kodak, and Kodak was using Moto for fab. Kodak could have made this work but they were so intent on secrecy and compartmentalization that they suffocated any rate of progress. By the time the first collaborative chip saw first light, the technology had already gone thru about 4 generation iterations in other PB products.

    Kodak "coulda been a contender" but they blew their chance at least twice, and more like 3 or 4 times thru poor management in my opinion. If only Mike McCreary and a few technical greats like Tom Lee had remained, perhaps things would have been different.

    Today I think Kodak is committed to imaging, which is no longer chemical, it is electronic. But they are not a semiconductor company so image sensors should probably a long term IP development operation (future imaging systems)and not part of vertical product integration strategy.

    Kodak's real screw up was not in image sensors. It was failing to forecast and accept the advent of digital imaging as THE imaging business and then lose the print business to HP.

    When the market rebounds someone should "buy" the image sensor technology team at Kodak along with a license to Kodak IP. They also need to close down their very old semiconductor fab.

    Design talent at Kodak, fab talent at IBM. A good core for a niche market play.

  22. > Kodak's real screw up was not in image sensors. It was failing to forecast and accept the advent of digital imaging as THE imaging business...

    Here is another view on Kodak digital transition:

  23. Indeed as I said, Kodak was there from the beginning. Upper management just did not accept digital imaging as THE business for their future.

    And as for so-called analysts...ha!

  24. "There is only 1 company I consider to be good in this regard, and the winner is Omnivision. Disclosure: I do not own OV stocks or receive any compensation from OV nor have I had any monetary or business dealings with OV for the last few years."

    I agree that Omnivision is well positioned for the immediate future with BSI, Wafer Level Optics, Wave Front Coding, modules, etc.

    Kodak problems:

    Kodak's problems are an absolute parallel to Control Data of the 1980s.
    CDC stood high in the middle of the computer business yet the missed the minicomputer, the super computer, the personal computer, and the internet.
    Simply a lack of vision and unwilliness to cannabalize an existing business in favor of a promising future business.

  25. CDC indeed bought into so many businesses that they lost focus. CDC ran their business like Noah's Ark. If it belonged in the animal kingdom, then it should be on the Ark. This ruined their fiscal planning and they essentially knocked themselves out.

    Kodak's Antonio P. is probably a lot more business savvy compared to 80's businessmen, and probably realizes this, and it looks like he is trying to stop Kodak from imploding on itself. Cash is King, and marketshare is for fools in these difficult times


All comments are moderated to avoid spam and personal attacks.