Defense industry changes and wireless infrastructure/5G opportunities on the horizon.
A Manufacturing Goal Initially Driven by Need
Vertical manufacturing can be defined as a company’s control over the value chain beyond the specific characteristics of a given product or products.
Not only did World War II need what had been learned and developed in prior decades about manufacturing, but the complexity of the equipment and systems it involved required that the sourcing of piece parts had to be conducted by end manufacturers to some degree. This meant that the entire process, from raw materials to finished product, was often controlled in a hierarchical, vertical way. This showed up nowhere more than in electronics, with radar systems being one of the first vertically integrated manufacturing regimens. Many of the required parts were simply not available from existing outside suppliers.
The first mass market for Radio Frequency (RF) power was in wireless infrastructure. As we all know, this is an extremely competitive and cost-sensitive business. To achieve the expected level of cost reduction over time, some degree of vertical integration is needed in the manufacturing process.
Why the Model is Starting to Change
If we move to the current day and look at electronics again, one can divide this vertically integrated world into two sections. Defense, principally defense electronics, and wireless infrastructure/5G. In the former, we see that vertical integration has become firmly in place but not ubiquitous. The two most visible current participants in this form of manufacturing process are Northrop Grumman and Raytheon.
Nowhere does this show up as visibly as in the RF power semiconductors that these companies use in their military systems groups. As mentioned in the preceding section, vertically integrated manufacturing really manifested itself in the RF power sections of radar equipment and systems. Almost always, financial concerns were not the main driving factor. The required technology— principally, Gallium Nitride (GaN) RF power devices—was not readily available commercially at the time. Both companies can make their own GaN semiconductor die and then build them up into Hybrid Microwave Integrated Circuits (HMICs) and then into modules and sub-assemblies.
One thing that has changed with the newest defense procurement processes is that emphasis has shifted toward the actual procurement of equipment and systems. Research and Development (R&D) and development of manufacturing regimens are now not as readily paid for by the government, sothe development of GaN and other advanced technologies into manufacturable devices and modules is now more dependent on company funding. Building these costs into the end systems can certainly be done to some degree, but the historical vertical manufacturing model for RF power, at least for now, will need to be looked at more closely as the financial aspects take on a more prominent role and reimbursement for these costs is not as certain. This reality will require a modified approach.
For wireless infrastructure, technology is important but not the driving factor. Cost is always the issue. This is a very competitive segment and anything that can improve a company’s position is important.
In this case, the vertical integration mechanism is not from devices and into equipment as it is on the defense side. It is, rather, from the bottom up, meaning that the device manufacturing player controls the making of semiconductor wafers and materials as well as the finished part. The two major examples here are Cree/Wolfspeed and Sumitomo Electric Device Innovations USA (SEDU). These two companies’ positioning could drive the market in the future.
What the Future May Hold
On the defense side, look for vertically integrated electronics manufacturers (like Northrop Grumman and Raytheon) to start purchasing more of the RF semiconductors from outside vendors. These will more than likely be some type of surface mount devices but might just be die (especially GaN). This will extend up to HMIC like devices at some level.
As part of the system development process look for some type of a more formal decision-making mechanism to be included in the procurement/manufacturing path to determine whether make-in-house/buy outside gate points.
Merchant suppliers, and other third-party vendors, will start to gear up for participation in this vertically integrated manufacturing regimen to some degree.
Firms like Northrop Grumman and Raytheon have always purchased RF power semiconductors from outside suppliers when it makes sense to do so. The difference now is that the amount of that engagement and the changing environment in defense electronics may accelerate things.
Wireless infrastructure, on the other hand, has different drivers. It is the largest segment in solid-state RF power devices, topping US$1 billion for 2018. At present GaN has a 27% overall share and is expected to continue to increase by several points per year. To a great extent, that penetration level is mainly driven by cost.
The competitive situation is complex, with about 10 vendors in play but with the top 4 controlling over 75% of the total.
When 5G finally heats up, GaN devices will drive the market harder as their principal frequencies are above 3 GHz. At today’s level for 4G and the nascent 5G market factory capacity, GaN wafer supplies can keep pace. In this highly charged atmosphere, though, when the market finally bursts through plant capacity raw material supplies cannot be assumed to be a given.
Vendors such as Cree/Wolfspeed and SEDU could find themselves in an advantageous position as their raw wafer supply is pretty much, if not exclusively, internally produced.
One other thing that plays into this scenario is that vendor-customer relationships are more likely to be some form of buying plant capacity than simply the purchase of parts. A strong and reliable wafer supply is necessary to make this option work.
It is also not unreasonable to envision wafer supply constraints when the market takes off and the number of required devices goes up. This is assuming that the raw wafers will still be Silicon Carbide (SiC) with the GaN structure being an epitaxial device on SiC.
The only disruptive force that could appear is GaN on silicon wafers but, although work is being done to that end, it is still unrealized.