fibeReality predicted a year ago that Huawei Technologies would start producing new types of optical components in-house, which evidently was substantiated in a news report in December. In between that time, there had been feedback from NeoPhotonics to us in September as well as from vendor discussions with the investment community that pointed to Chinese vendors in general moving more towards becoming direct competitors, including Oclaro mentioning in November, DCO module production happening in China, as well as Finisar not “com[ing] away with a warm and fuzzy feeling about 2018 from China demand.” For Fabrinet, the leading Contract Manufacturer (CM) of optical equipment in the world, the loss of a substantial amount of business in China by some of its large customers, including Lumentum Holdings and Acacia Communications, is just one more factor to race ahead as quickly as possible into other spaces outside of traditional communications. In addition, by getting into new businesses, at least theoretically, a shift in its strategy in the direction of design types of work cannot be ruled out.
Of course, it is possible that Huawei and ZTE may decide to use Fabrinet’s facility in China, but at least as of May, 2017, the CM acknowledged at the J.P. Morgan Technology, Media & Telecom Annual Conference, that it “doesn’t have any Chinese customers, per se.” Also, at the event, it was estimated that only 10 percent of its work done in China, remains in the country. We suspect that both system players would be more inclined to potentially work with a CM with a much greater presence in China.
Although Fabrinet has been attempting to position itself in a significantly larger manner at being in the outsourced semiconductor assembly and test space, other companies have a great deal of more experience in this area. Moreover, as of October 2017, we understand that one of Amazon’s optical vendors was told by the customer to allocate capacity as the Macom-Fabrinet project was not on schedule. Nevertheless, it was always a canard that Fabrinet, which has so much familiarity with constructing transceivers, would have any problem in being able to potentially work with Macom’s design.
Speaking of Fabrinet getting into design itself, it would not have made sense with its current telecom/datacom customers. As with other CMs, it has survived on very thin margins, and the only type of design, which tended to be logical has been with guaranteed production. The only exception may have been if a product had been hot, and the customer’s R&D team happened to be missing a particular skillset. However, usually there is no time advantage, pain reduction, or cost benefit for the transceiver company because the design-in for manufacturing prototypes is done on different equipment than for general production.
Yet, Fabrinet’s new CEO comes from Sanmina, which has experience in other market sectors, like automotive, but is most well-known for focusing on Original Design Manufacturing (ODM). In these adjacent markets, especially medical, there can be as long as a five-year design cycle because of the lengthy period for compliance, such as with the US Food and Drug Administration, the Good Manufacturing Practice, etc. Still, for a public company, such as Fabrinet, which obviously, is looked closely for performance each quarter, shifting only partially to a long-term vision type of model may not fly. Perhaps only a pure ODM play allows the necessary flexibility in being able to afford to invest a certain amount of revenue in self-funded, R&D projects.
So, while we have heard good news that a big optical component customer is contemplating at least expanding some of its lines at Fabrinet for one of its businesses, the pressure to diversify certainly becomes more acute as Chinese system houses continue to increase their levels of vertical integration. Naturally, this threat, which could have implications for componentry sales outside of China as well, also means a greater likelihood of consolidation of US componentry vendors, which could result in less attractive pricing terms for the CM by these bigger combinations.
[written by Mark Lutkowitz]