top of page
Search

The One Trillion Dollar Dispersion

After posting 25% growth in 2021, the consensus forecast for the semiconductor industry is another 10-15% growth for 2022. Even in the face of macroeconomic headwinds, this optimistic outlook has prompted discussions (see here, here, and here) on the prospects for a $1T market by 2030. WSTS estimates for 2022 suggest that the semiconductor market would need to grow at an average compounded rate of 8.5% per year to reach $1T by 2030.


The scenario is a bit ambitious but possible (RCD Advisors forecast is ~7.1%). The semiconductor market had a 6% CAAGR over the last decade. However, tech hardware is now entering a new golden age. Emerging downstream technology trends like vehicle electrification, driver automation, AR/VR, machine learning, and IoT could boost the demand for semiconductors even higher than the historical average.


Assuming a $1Trn market value, the prospects of an extra $444Bn revenue in 2030 are lucrative. And there is no shortage of recommendations available to capture that opportunity. But most of this guidance relies on supplier size and scale advantages. There will certainly be more consolidation in the semiconductor market (see here for Samsung/NXP rumor). But, a $1T semiconductor industry in 2030 (or 2031) could look much different than today. And if it does, relying on competitive advantages related to size or scale may not be enough.

Let's start with three observations:


1. Silicon's Value Growth

Paradoxically, even with all the rhetoric of Moores' law slowing, engineers are rapidly integrating even more functionality into silicon. The result is a widening mismatch of the IC economic value flowing into the tech hardware ecosystem versus everything else. If "software is eating the world", then "semiconductors are eating the hardware industry". The table below illustrates the differences in growth rates across the value chain.


Although the mismatch in value growth has always been true, the forecasts are projecting a much more significant mismatch. Semiconductor suppliers are potentially eating away at the downstream margins of OEMs and taking away BOM value share from other component suppliers at a faster pace. In practice, OEMs have been moving more of their "value-add" into software and services to maintain and grow margins. But the hardware value is leaking into semiconductor components.


The best modern example is cars becoming "a computer on wheels." The concept is reminiscent of how economic value eroded in the PC industry 15 to 20 years ago. Remember when desktops were a "Windows box" with "Intel Inside?" Car makers and tier-one suppliers are now rethinking their supply chains to avoid losing margin to their semiconductor suppliers. The same value leakage occurs in servers (with machine learning processors) and in network switches.


Semiconductors are also becoming an even larger share of the input costs (BOM costs). Today's 5G phased array antennas are an example of RF semiconductors seizing value from traditional passive antenna components. In another example, SiC switches reduce thermal management and energy storage requirements in electric vehicle inverters. RCD Advisors estimate that all non-semiconductor parts collectively represent less than 17% of tech hardware industry value. That percentage will only go down further as semiconductor content continues to grow.


2 Democratizing Design

A foundry industry model emerged long ago, offering semiconductor companies an avenue to outsource production and concentrate on the design of chips. Today, something similar is happening with design. But instead of isolating a part of the value chain into a separate merchant business (like foundries), IC design capability has become much more accessible and diffuse. EDA tools and IP blocks have allowed more companies to design semiconductors and fabricate them at foundries. Apple has moved in this direction. So has Google and Amazon. Even Tesla develops their own ICs (GM and Ford are starting to do the same.)

Tesla's Full Self Driving Chip

There is ample motivation to defend margins from leaking into the silicon of a supplier. In addition to software and services, as EDA tools become more accessible, and fabless design startups proliferate, a part of the $1 T future semiconductor market will likely be custom. As economic value continues to gravitate towards the fourteenth element of the periodic table, every organization in the supply chain will have to reckon with the Coasean "design vs. buy" decision. And often, the optimal solution is for OEMs, ODMs, and other component makers to design these parts themselves.


3 IC Modularization

Organizations are beginning to assemble ICs into modular systems using advanced packaging techniques. As system-in-package (SiP) becomes ubiquitous, the definition of electronic assemblies and ICs blur. The most obvious example is the evolution of network processors (used in data center switches) and co-packaged optics. The network processor is already 50-80% of the network switch costs. If chip makers adopt co-packaged optics, it will further extract value from the box and motherboard assembly.

Copackaged Optics Network Switch Demonstration

Most SiP suppliers today are semiconductor suppliers. But, as the technology becomes standardized (maybe with Intel’s UCIe interface), OEMs and other component makers will also offer SiP solutions. And as motherboard assembly value moves into the IC package, SiP becomes the new market opportunity for other component suppliers to sell into and for EMS organizations to expand their services.

Source: Rockley Photonics


Implications

Rapid value creation in silicon, easily accessible EDA and IP blocks, and modularization are accelerating the dispersion of the semiconductor industry into the rest of the tech hardware supply chain. Eventually, most organizations in the tech hardware industry will design specialized IC solutions (if they can find the engineering talent).


As a result, a $1T semiconductor industry in 2030 (or 2031) will look much different than today. Part of the growth will include custom silicon value. Another part of the growth will consist of system value embedded into a module. Strangely, the supply chain may start to look more like the fragmented plastic parts business. In the plastic part supply chain, CAD design tools allow almost any organization to design custom components.


It helps that the semiconductor market is growing above average. But to capitalize, it is worth stepping back from size and scale assumptions. M&A and strategic investments need different objectives when the industry disperses. The investment premium will be on technologies and processes that enable this dispersion. It isn't solely about licensable intellectual property but rather the technology that allows that property to become more pervasive. It's not necessarily the EDA tools but the innovations that will enable these tools to become easily accessible. Contact RCD Advisors for more on the emerging custom silicon value chain.


コメント


bottom of page