Outsourced Manufacturing Will Have Little Impact on Optical Suppliers' Operating Margins

One of the great debates taking place right now within the optical components industry is whether to outsource production, and move to a fabless model like most logic chip makers. One one side of the argument, Finisar (FNSR) is staying in-house, while rival JDS Uniphase (JDSU) is preparing to send its manufacturing operations to Fabrinet in China.

Advanced Technology vs. Standard Economics

The components industry is littered with penny stocks, and is filled with remarkably intelligent people who can improve the transmission capabilities of fiber optic technologies, but have no idea how to increase operating margins. As a result, all kinds of experiments with different materials and manufacturing processes are taking place now, and the industry is crossing its fingers and hoping something sticks. On one end, silicon photonics vendors like Intel (INTC), Luxtera, and Lightwire are hoping they can match the chip industry's low materials costs, while on the other, JDS Uniphase is hoping it can match the chip industry's low unit manufacturing costs.

The idea to outsource like chip vendors do makes some sense. By concentrating procurement and manufacturing, there should be some savings in material purchases and shipping, not to mention greater leverage with volume purchase arrangements. But outsourcing cannot get rid of onerous VMI (Vendor Managed Inventory) agreements, which force components suppliers to consign inventory at their customer's manufacturing sites, contract or in-house, even when the customer has made no commitment to purchase everything that ships. With inventory cycles that typically last 70-100 days, this forces components vendors to bear most of the risk of unsold and obsolete inventory, and they will need additional mergers to gain more negotiating leverage with big customers like Cisco (CSCO), Alcatel-Lucent (ALU), and Fujitsu.

More important than VMIs though, is the fact that there is no optical equivalent to the dramatically rising lithography and equipment costs seen in the semiconductor industry. With optics, much of the manufacturing cost gets tied up in assembly and testing, which typically means hiring more people, not buying advanced etching machines. This is reflected in the high asset utilization of components vendors. Finisar's Revenue/PP&E of 6 is higher than the 4.6 put up by fabless semi maker Xilinx (XLNX).

Bringing Mundane Inventory Management to Exciting Science

The optical components industry is still having trouble ridding itself of its science fair culture, and this is reflected in the nonsensical argument that outsourcing somehow puts intellectual property at risk. If this were really the case, the entire datacom and semiconductor industries would have been done in by copycats. Nonetheless, it is also poised to resume its revenue growth with the advent of 100 Gigabit transmission, and further advances in DWDM and Fiber-to-the-Home. And with a labor-intensive manufacturing process, outsourcing cannot raise margins to the extent it does in the capital-intensive semiconductor industry. As a result, focusing on the cost effectiveness of what happens before and after manufacturing, including shipping, purchase agreements, and consignment sales, will be just as important as deciding who will be responsible for final assembly.

Disclosure: No Position

Comments

[Kirk Lindstrom:]

You wrote:

"The optical components industry is still having trouble ridding itself of its science fair culture, and this is reflected in the nonsensical argument that outsourcing somehow puts intellectual property at risk. If this were really the case, the entire datacom and semiconductor industries would have been done in by copycats. "

This may not be true. The IP for Altera (ALTR) or even Intel (INTC) which is going to license and outsource some of its low power, low cost chips, is different than than the IP of Finisar (FNSR).

Much of the IP at Finisar is "how to" that once let out of the bag, is impossible to get back. When I worked at HP fiber optics in the 1980s and 1990s I came up with a manufacturing trick to get good yield on optical transceivers with a couple more db sensitivity than the competition. We never patented the idea because it would be impossible to enforce. We kept it secret and didn't even publish internal papers for internal conferences.

I believe Intel will keep its leading edge chip technology in-house. But it will license its low power, low cost Atom processors to TSMC so others can embed the processor in SOICs and pay Intel a licensing fee.

Finally, I read that many of the component vendors from China have parts to sell in the US but they are having trouble getting them qualified. Knowing how to put the pieces together to make a RELIABLE module is valuable IP, why share it with your competitors which is what you will need to do to get qualified... since what you teach your "partner" will be used to qualify your competition's parts build by the same people.

Apr 30 09:53 AM

[Asian Tiger:]

"....while rival JDS Uniphase (JDSU) is preparing to send its manufacturing operations to Fabrinet..."

The exact reason for JDSU using CM is that JDSU's management didn't know how to manage its own Shenzhen facility. The most executives, directors and manages at JDSU's transceiver business, unfortunately didn't understand the optical business.

Apr 30 10:46 AM

[numinary:]

I understand that investors cannot appreciate the difficulty of manufacturing fiber optics, but I still struggle not to sigh when I read comparisons to the IC industry. This is similar to comparing a NYC hedge fund to a local community bank -- scale is different, customers are different, regulations/standards are different -- but they both deal with money...

Some of this is inherent in the physics of the devices, some of it is inherent in the different market sizes -- 2 orders of magnitude is a big deal...

Vertical integration for optics suppliers can be a key differentiation. It is important for the optics suppliers to understand what technologies matter most to keep in-house. For example, perhaps Finisar feels that their differentiation is at the transceiver manufacturing level so they don't want to let that "know-how" out to competitors thru outsourcing of manufacturing. JDSU might feel that their differentiation is in the fabrication of optical devices and the manufacturing at the transceiver/sub-system level is not unique.

The author seems to lump all of the manufacturing together in one clump and not appreciate that different manufacturing steps have different technological requirements. The manufacturing that is going on at Fabrinet for JDSU does not involve the chip (iii-v semiconductor such as InP) fabrication, but rather PCB manufacturing, and some chip packaging. Finisar keeps it all in-house because they have more high volume products with which to ride out waves in the industry whereas JDSU is more of a telecom play and tends to have more peaks & valleys which correspond to network build-outs.

Please do a little more homework on the industry before posting next time. All the same, I take your comments about the science fair culture to heart and do not deny that some of that is present...

Apr 30 08:25 PM

[BasaBasi:]

Established technologies at lower bit-rates may be sourced from China or other lower cost regions but anything above 2.5Gb/s particularly for use in Single Mode applications is not straightforward to build with acceptable yield & performance. This applies more to the laser die themselves but complete transceivers are not trivial to assemble or test either.

I worked on the first 10Gb/s serial transceiver in the late 90s and remember the pain... though we were the industry leader in the end. Until the dot.com bubble burst of course!

All this gets me thinking about AMD and whether their decision to outsource is a terrible strategic error that means they will not have any degree of technology leadership in the coming years.

May 01 04:09 AM

[siliconkanth:]

I think the caption is catchy but I find little substance. I am a PhD in silicon photonics and have been following the trends very closely.

1. Primary argument, that the capital intensive industries alone (as
opposed to labor intensive) benefit from off-shoring is probably
incorrect. For example the best off shored operations, BPO (telephone
based support, medical transcriptions, tax filings) need little
capital to start. In fact, all off shoring to India is done in labor
intensive as opposed to infrastructure intensive businesses.

2. Secondary argument is that since vendors (JDSU eg) need to maintain the stock at Cisco etc it cuts into the advantage of off shoring. The need for maintaining stock is not a feature of optical components alone. The industrial ASICs to Cell phone chips/power supply electronics must be going through similar requirements.

All said and done, off-shored photonics companies competing with
on-shore photonics companies will face same costs for maintaining the stockpiles.

3. Another thing to consider is that one can do fab-less operation
without going off shore. There are a number of electronics companies
who do their fab in US. Silicon photonics allows this and thats what Luxtera and Lightera are trying.

4. Third argument, that the optical lithography does not need upgrading, I find hard to believe. As the leading fabs upgrade, the cost of a particular resolution process follows a U shape. The newest being the most costly and the older again being more expensive as the equipment makers (applied materials etc) phase out the old process machinery. So even though optical fabs dont need the latest resolution processing they need to stay at the bottom of the U-shape of the cost of fabs. So I doubt this statement as well.

5. And finally, there is little substance to my and his arguments except
may be some logic. My hope is that when numbers are put in, taking
40-100 million out from the initial 5 years cost for the fab
cuts the investment needed substantially. This is very helpful for a
start up but not necessarily for an established player like JDSU.

PS: I hope the author run the numbers before writing this article and compared the results for off-shored, non-offshored, established and upstart companies. It will be great if you share them else there is no credibility to anything mentioned in the article. If not as 'numinary' mentions above, I suggest the author do more homework next time.

May 05 01:54 AM