fibeReality retraced the last three years of the six-inch 3D Sensing (3DS) VCSEL wafer marketplace as part of our preparation efforts in presenting at an EPIC conference last week. As a result of our latest intelligence gathering efforts, it became evident that simultaneous to the great achievement by Lumentum Holdings (along with its partners) in shipping more apertures within a two-month period in the fall of 2017 than in the entire history of such devices, there was the beginning of a marketplace “pop,” which gradually worsened. In particular, there were a couple of vital matters left out of what otherwise seemed to be a very successful story.
After initially dismissing Finisar because it remained stubborn about sticking with the four-inch VCSEL diameter, Apple was forced to go back to the supplier to renegotiate a contract to obtain a rather substantial number of wafers at the smaller size just to satisfy the initial iPhone 10 orders. At least at the very beginning, Finisar turned out to be correct with its prognostication about a hefty amount of yield lost in moving to the larger size. More importantly, it led to the vendor getting back into the game with the announcement of the new, huge fab in Sherman, TX.
Subsequent to the preliminary rush for components, we understand that it was discovered that an engineer apparently involved with the Lumentum effort happened to make a simple, mathematical error. Panic set in as there was a great deal more inventory in the pipeline than previously presumed. Suffice it to say that momentum took a big hit.
Logically, we assume that had there been widespread knowledge of this evidence, it may have partially tempered some of the enthusiasm of so many players going after this seemingly attractive 940nm space, such as in aggressively constructing new fabs, while paying inadequate attention to other VCSEL market segments. Even Lumentum got carried away by the excitement. As yields have significantly improved since the beginning days at companies like WIN Semiconductors and Skyworks, of course, the large overcapacity situation has worsened.
In addition, the EPI firm, IQE also was way off on demand, which did not materialize, and overbuilt its facilities. Moreover, we have found out that the plan for II-VI’s Warren, New Jersey plant is to devote resources for a GaN EPI layer for SiC.
While we expect progressive growth over the next few years at 940nm, its extent is the biggest unknown factor. A lot of it will depend on whether “killer” parts become available on smartphones, or on Apple potentially introducing a totally new consumer offering.
Given that price pressures are bound to become greater, we believe that no more than a doubling of the VCSEL market size for this phone space should be anticipated. Furthermore, the idea that the automobile sector will come to the rescue in a really meaningful way anytime soon for such sensors should be abandoned.
The market cycles in the auto realm are excruciatingly long, and work on more primitive types of networking will frequently continue to take precedence. We are not bullish on VCSELs involved with autonomous driving for use on the outside of the vehicle.
For our fast-growing, totally separate, quick update blog, which is exclusively on fibeReality’s LinkedIn page, please follow us here.
As always, fibeReality does not recommend any securities, and this writer does not invest in any companies being analyzed by us.
[written by Mark Lutkowitz]