The big difference between 5G and its predecessors is that it is far more than a new mobile network – to deliver on its promises, ‘true 5G’ is a broad platform enabled by a range of interconnected technologies. The convergence of the cloud, especially the edge cloud, with connectivity; the application of AI/ML analytics to everything from network operations to IoT decision making; the enhancement of the capabilities of touchscreens, voice interfaces, augmented reality, cloud storage – all of these are part of the 5G picture.
They are all, of course, enabled by evolution of chip technologies, from RF chips to high end processors, and 5G has been one catalyst in the recent realignment of the semiconductor industry. Events over recent weeks have reinforced several key trends in the sector, which have been visible for several years now.
- Reports that AMD is to buy Xilinx would, if they result in a deal, be just the latest in a string of large mergers and acquisitions, including Intel’s parallel deal to purchase Altera, and Nvidia’s bid to acquire ARM. If AMD does not acquire Xilinx, it is likely that somebody will (pberhaps Qualcomm), given the rising strategic importance of FPGAs (field programmable gate arrays), and programmable chips in general, to support the endless flexibility of 5G.
- ARM itself held its annual developer summit recently and focused heavily on its expansion into high end cloud and network infrastructure processors. This is a market where the adopters of ARM-based designs have long been trying to chip away at Intel’s dominance, and 5G/cloud convergence makes the battle even more strategically critical for both sides, as does the effort to partner with the webscalers for their next generation processors. If Nvidia does succeed in buying ARM, it will have a broad platform reaching from supercomputers to devices, and integrating its own GPU-centric approach with a more Intel-like CPU architecture. Meanwhile, the progress of ARM designs in infrastructure markets is starting to drive disruption of Intel’s incumbency, although ownership by a single chip provider runs a high risk of destroying the famously successful ARM licensing model.
- Another important trend of the 5G era has been network vendors’ change of strategy with regard to their chips. Nokia and Huawei, in particular, have diversified the traditional approach of relying on proprietary ASICs (application-specific integrated circuits) designed by close partners. These are still important, but there is higher usage of merchant chips for some functions, particularly the cloud infrastructure underpinning virtualized network functions such as RAN centralized units (CUs) or 5G cores. But there is also a rising trend to move inhouse efforts on from ASICs, to create full platforms that can even be offered to third parties. Nokia has highlighted the risk of this strategy, having to change course on its 5G base station system-on-chip after poor performance of its first design. But it is also reaping the rewards with the performance differentiation of its FP4 router processors, which are also targeted at the webscalers.
- If vendors are getting more interested in chip design, semiconductor majors are moving up the stack in some sectors. Qualcomm, in particular, is stepping up its efforts to offer reference designs, and even complete smartphones, to facilitate adoption of 5G devices. Reference designs are a time-honored way for chip giants to seed the market, and with clouds hanging over Qualcomm’s future in the Chinese market, thanks to the current geopolitical stand-off, any stimulus for the 5G devices market elsewhere will be important.
- While vendors may still differentiate their products with their own chip designs, there will be fewer elements of networks or devices in which proprietary implementations – even if they enhance performance – will be acceptable. The main focus of the open networks movement has been on interfaces between RAN and xHaul networks, or between different base station elements. But there is also considerable effort going into open interfaces at chip level, whether that comes from an industry forum (Small Cell Forum’s FAPI, for instance), or is vendor-initiated, like Intel’s work on chiplets. The latest effort is OpenRF, which is focused on the radio chips.
All these trends will affect one another and profoundly change the market landscape in the 5G era.