It used to be fairly simple to segment the chip markets. You had general purpose CPUs, primarily from Intel, used in PCs and servers. There also was a market for high-performance GPUs from Nvidia and AMD for high-end graphics processing and gaming. And then you had the array of ARM based chips powering the mobile device revolution, primarily from Qualcomm, but also from Samsung and MediaTek among others. Finally, you had a number of connectivity and networking focused vendors like Broadcom and Marvel.
In addition, there were functional accelerators like FPGAs, networking, Wi-Fi, etc. that had significant niches to enable. But the arrival of 5G, with a focus on the intensive computing needs of core networks, virtualized RAN infrastructure, edge computing and AI/ML, have fundamentally altered the playing field in the semiconductor market.
Nearly every chip producer sees 5G as its “goldmine” and has positioned to take advantage of the potential opportunity. But 5G is different than previous generations of wireless networks that were primarily built on designed-for-purpose devices that were often custom-built by network equipment vendors (e..g, Cisco, Ericsson, Nokia). The advent of virtualized infrastructure powering the networks that’s taken place over the past 3-5 years, which is crucial to the operation of 5G networks and beyond, has created a battleground that both incumbents and new entries want to stake claim to.
Let’s look at how this new generation of virtualized systems and high performance computing needs have altered the marketing plans of the various chip producers.
Intel
For its part, Intel remains the leading provider of core infrastructure systems, and with its emphasis on targeted systems like Xeon Scaler and FPGA processors (which it acquired when it purchased Altera in 2015), it holds a leading position in v-RAN network transformation and the emerging O-RAN build outs.
AMD
While AMD is making inroads in general servers and cloud enabled data centers, it has yet to have a major impact on the CSP infrastructure market. But that is about to change as it gets ready to digest Xilinx, and create a competitor to the Intel FPGA accelerators so often required to make 5G and especially high performance processing for Massive MIMO work.
Nvidia
Not to be outdone in this space, and already leading in many AI featured systems, Nvidia plans to acquire ARM so that it can target, among other things, the fast-growing needs of virtualized systems for 5G (estimated to be 7-8 million new RAN installations). Nvidia has the acceleration products needed (especially with its Mellanox networking acquisition and its move to DPU), but it needs a general-purpose computing architecture that ARM can provide. There is still some question if the ARM acquisition will be approved, but it is potentially strategic to Nvidia’s long term vision of being a mainstream computing supplier rather than just an accelerator supplier.
Qualcomm
To indicate how hot this market for 5G infrastructure and edge computing is, Qualcomm, long the dominant player in mobile devices, has recently introduced its own offerings to power the exploding RAN market. Its DU X100 card is targeted at the need to empower commercial off-the-shelf (COTS) servers from vendors like Dell and HP, with the ability to become 5G enabled baseband units. They’re leveraging Qualcomm’s extensive skill in wireless technology expertise to produce a high-performance radio unit that plugs into a standard PCIe port, and this capability is critical to enabling O-RAN deployments.
Other specialized chip vendors like Broadcom and Marvel have also seized upon the momentum in the 5G market, as well as the expected explosion in edge processing, to create targeted products for this marketplace.
The MEC effect
The growth of mobile edge computing (MEC) means that there will be immense competition in producing high-performance but less power hungry products, both for on premise needs (e.g., sitting near cell towers or in enterprise locations), as well as in cloud-oriented and often carrier operated distributed data centers.
In smaller, lower power systems, it’s likely that ARM-powered solutions will have a major competitive advantage and vendors like Qualcomm and Nvidia (should it acquire ARM), will be major players. Increasingly at the “mid-range” of the edge, enhanced services like AI and ML will be required (as well as at the high end and core of the networks), and that means intense competition from Intel, Nvidia, and from ARM players with significant AI expertise like Qualcomm.
The edge market is a very diverse one, consisting of products from low- end single-board systems all the way to high end cloud-enabled data center products, so there is a lot of opportunity for everyone to compete. Indeed, in this space even the cloud players are trying to compete with their own chip designs, with AWS creating its Graviton ARM-based solutions, and Microsoft and Google working on similar designs which will be available in their cloud edge services.
The bottom line is that while the chip business has always been a fairly complex mix of solutions for multiple computing needs, and with a fairly large number of chip suppliers especially at the niche functions, 5G and the move to enhanced and highly compute-intensive virtualized wireless networking has fundamentally shifted the strategy and segments in which the primary chip vendors play.
It also signals an increasing utilization of acquisitions of niche players by the mainstream vendors as they build out their function-specific needs (I expect a good deal more M&A activity to come about in the next 1-2 years).
Finally, the new playing field means that it's no longer a game to be won by highly segmented players being good at one or a few things. Chip vendors need to be broad functional providers of products to thrive in the new marketplace.
Look for many more combinations of technologies to come to market from an increasing number of players. Although, in the longer term, I expect the market to consolidate around 3-5 major companies that have the scope and breadth to deliver needed capabilities.
Jack Gold is founder and principal analyst at J. Gold Associates, LLC., an information technology analyst firm based in Northborough, Massachusetts. He has more than 25 years of experience as an analyst and covers the many aspects of business and consumer computing and emerging technologies. He works with many companies, including Intel. Follow Jack on Twitter and on LinkedIn.
His views are his own.
RELATED: Gold: Intel’s Covid-19 model response