Nvidia’s intimation that it will license its graphics technology to third parties highlights broader challenges for wireless chipmakers. The decline of the traditional PC is a given, so it is essential to win share in the mobile space. However, Nvidia has made progress in the new ‘post-PC’ form factors like tablets, it has not yet converted this into a really viable business. This is largely because most mobile volumes still come from handsets, and in that space, Qualcomm increasingly gives its rivals very little oxygen. An Intel success in the mobile space, which now looks unexpectedly likely, will leave other players racing for new markets, ones where the CPU power/performance race is not the primary driver of success. These will include cross-platform technologies in specialized areas like graphics (like Nvidia and AMD), and of course the Internet of Things (IoT), which will provide an important home for advanced modem makers.
Certainly, only Qualcomm and MediaTek can safely rely on smartphones alone (and neither of them is doing so). Integrating LTE modems with its Tegra processors will help Nvidia address smartphones more aggressively, but even as it does, Qualcomm is moving onto its turf. A recent CNet benchtest of the Snapdragon 800 put the latest Apple and Samsung Exynos processors into the shade in Nvidia’s most sensitive area, graphics performance (though there was no comparison made with Tegra 4). Against such a backdrop, it is not just Nvidia, but all mobile chip providers, which need to decide where their next growth market will lie.
Some have already done so (and some, like ST-Ericsson, were drowned before they could work it out). Texas Instruments has virtually exited smartphones ‘ first basebands, then processors, leaving it mainly present via its connectivity chips. It is building up its massive analog business and has also, tellingly, said it plans to adapt its OMAP mobile processor architecture for the internet of things. Freescale also backed out of the smartphone space and is similarly turning many of its mobile processor efforts towards the IoT, where it will combine them with its existing strengths in embedded and automotive chips, to reflect the rising levels of processing power involved in machine-to-machine (M2M) applications. Broadcom and Marvell remain determined to make their mark in handsets, though they too are likely to be squeezed by Qualcomm outside of the connectivity space.
Even more squeezed by Qualcomm in handsets are the modem specialists, given the inexorable trend for smartphones to feature integrated modem/processor platforms like Snapdragon. Some of the single-trick ponies have been acquired (Icera by Nvidia, Beceem by Broadcom, and so on), while others, like Sequans and Altair, are building their strategies around the evolution of an LTE-only market. Some of that will lie in high end gadgets whose cost and power limitations are less rigid than those of a handset, so all-in-one chipsets will be less essential. Carriers which are investing in national LTE coverage will soon start to sell 4G-only products ‘ not handsets, because of the roaming issues, but data-driven gadgets like tablets.
But the larger LTE-only market will revolve around the virgin networks being created for the internet of things ‘ the evolution of the current M2M space that sees virtually every appliance acquiring an internet connection and needing more powerful connectivity than in previous, low-rate M2M systems. For instance, last fall Sequans introduced Streamlite LTE, a parallel range to its Streamrich LTE line of 4G products, which targets high end consumer devices. The new low power family is aimed at consumer electronics and M2M products and kicked off with the SQN3101 Firefly, an LTE baseband that delivers up to 50Mbps of throughput with ultra-low power consumption.
Of course, there are many challenges in adapting a platform and a business model which was conceived for big-ticket mobile devices to a world of millions of ultra-low cost gadgets, and much of the profit will lie in the products that control all those sensors and embedded controllers rather than the chips themselves.
For LTE to be present in millions of connected devices, it needs to be far cheaper ‘ a $10 increment to the product cost, like embedded Wi-Fi, rather than its current $150 or more, as Altair puts it. This is where simple, low power, LTE-only chips like Streamlite LTE, or Altair’s promised family of next generation offerings will come into their own, seeking to undercut the current high margins that Qualcomm can command in the embedded module sector (it supplies its modems to IoT module leader Sierra Wireless, as well as Novatel and others).
But there is huge growth potential in the fledgling segment, which will attract the mobile players as well as specialist providers from the automotive and industrial segments. The IoT will surpass 25bn units and $4 trillion in 2020, predicts research by IDC, and that only applies to devices with 32-bit microprocessors, not sensor nodes or those with only microcontrollers (the latter, itself, a big market for ARM and its licensees).
And in the medium term, this market has the potential to evolve rapidly, supporting new applications and opening doors for many innovators. To achieve the full potential of the IoT, rather than merely connecting up more devices in the conventional way, there needs to be new thinking about the services supported, the architectures, the security systems, and the business models for all concerned.
This is not just about equipping existing non-smart devices, from cars to refrigerators to heart monitors, with internet access to cloud services. It is about creating a whole new ‘fabric’ of sensors which connects everything imaginable, and from the vast amounts of data collected, delivers hugely granular levels of intelligence. As Intel put it recently, a user does not want to have to tell a coffee maker they want a drink; they need it to understand intelligently when coffee is required.
Singapore is one of several countries which are buying into this vision, starting with smart city initiatives and expanding beyond that into a ‘sensor fabric’ which will support not just the internet of things, but the fabled ‘internet of everything’. The island state’s Infocomm Development Authority says it wants to be the first country in the world to be covered by such a fabric, based around RFID tags, wearable sensors, environmental sensors and smart meters.
Others see the IoT evolving into a platform for immersive entertainment. A recent event at Stanford University highlighted a project at rival MIT, which is creating a giant display around a huge cloud of tiny drones, each carrying an LED, and resulting in a hugely interactive content experience, with users both viewing and contributing video in real time.
Even further out than that, perhaps, some projects are focused on ‘body area networks’, an area where Qualcomm is doing significant research. This will evolve from the use of sensors attached to the body to monitor factors like heart rate or movement, in the healthcare and fitness markets; and from the emerging trend for ‘wearable computing’ in Google Glass or Apple’s connected clothes ideas. Those foundations could lead in future to ad hoc networks based on everybody being equipped with sensors and transmitters. One initiative in this area was announced recently by French institute CEA-Leti, which sees applications in group indoor navigation, health monitoring on a massive scale and in collaborative communications. Its Cormoran project would see body-attached devices doing more than just exchange data with the internet, but actually routing and relaying it to nearby people, creating a huge distributed mesh network.
These ideas are fascinating but, even putting security and privacy implications aside, unlikely to become real revenue sources for chip suppliers in the near future. While most mobile silicon players are keeping an eye on IoT, they also need to secure themselves some kind of role in consumer devices and the post-PC landscape. A popular route is the spread of ARM-based processors into the server and network infrastructure segments. But those markets will become ultra-competitive too, so the best prospects of achieving a niche with decent margins may entail some more flexible thinking – working around Qualcomm and Intel, perhaps by licensing technology or focusing on specialized functions such as advanced video and augmented reality.
Back to Nvidia ‘ it is preparing to license its famous graphics technology to other processor makers to incorporate in smartphones or larger computers. It will offer both its GPU cores and patents related to ‘visual computing’, it said earlier this month, and expects products with its technology to appear from 2015. “It’s not practical to build silicon or systems to address every part of the expanding market,’ wrote EVP and general counsel David Shannon in a corporate blog post. “Adopting a new business approach will allow us to address the universe of devices.”
It will initially license the GPU core based on its Kepler architecture, which is currently used in its own GeForce, Quadro and Tesla graphics products (for PCs, workstations and supercomputers, respectively). The next generation of the Tegra mobile platform will also harness Kepler.
The move will, however, create new competitive tensions for Nvidia, even with ARM, whose processor cores it licenses ‘ but which has its own GPU platform, Mali. Another IP-only graphics provider, Imagination Technologies, which supports the iPhone and some Samsung devices, has recently seen its shares under pressure and will not welcome a new rival. But complex business models and relationships are the only way to build up sufficient presence in competitive areas like mobility ‘ Nvidia’s CEO Jen-Hsun Huang said: “I think the shape of companies in the future will look increasingly hybrid. Microsoft licenses software, they sell devices, they have services. All of the above. I think you’ll see more and more companies that are like that.”