What might you get by crossing BB-8 with Xavier? The answer isn’t a mash-up between characters from Star Wars and X-Men, but a self-driving car.
For many people the idea of being driven around by their own vehicle feels like science fiction. But the race to commercialise the first driverless car is all too real and Nvidia Corp says it will have its first autonomous model on roads within a year. It is billing its new chipset Xavier as the world’s first AI supercomputer chip, and BB-8 is the name of its test vehicle.
Nvidia’s share price has almost quintupled since July 2015 and Harvard Business Review voted its Taiwan-born founder Huang Jen-hsun as the world’s most effective tech CEO last year (the next Asian in the business rankings was Hon Hai’s Terry Gou at 40).
Based in Silicon Valley, Nvidia designs the kind of application processors that are bringing self-driving cars closer to reality. But the race to be first is intensifying, with NXP Semiconductors – its main competitor – in the process of being acquired by the world’s leading chip designer for mobile phones, Qualcomm.
Also in contention is Intel, which has just announced that it is spending $15.3 billion on Israeli road sensor group Mobileye as part of its own push into self-driving cars. The plan is to combine Intel’s processors with Mobileye’s driver-assist platform. (Intel’s chief executive Brian Krzanich has described it as a merger of “the intelligent eyes of the autonomous car with the intelligent brain that actually drives the car”.)
All four of these firms are American – not something that plays well in China, which has ambitions of its own to lead the world in self-driving and electric cars.
But perhaps the more pressing question is whether Chinese firms can get to the front in making the kinds of chip that provide these cars with their brainpower.
Last week we examined the progress that China has made in eroding American dominance of the semiconductor sector. In the manufacturing of chips (through much-increased capacity at foundries) Chinese firms are starting to have an impact. But when it comes to designing the chips – especially the higher-end performers – they are further behind.
Most of the Chinese efforts to close the gap have focused on chips for PCs and mobile phones, the two platforms which have dominated the semiconductor industry for two decades. But the greater challenge is how to prosper in sectors likely to set the pace in the coming decade, particularly self-driving cars and data centres deploying deep learning, a form of artificial intelligence (AI).
The underlying technology for these new applications is based on developing high-performance computing (HPC), generating a speed and processing power that will smash Moore’s law into smithereens.
And here the Chinese are some way back, with industry titans like Robin Li, founder of Baidu, China’s leading search engine, more than aware of how far they are lagging behind.
Baidu itself is partnering with state-owned Beijing Auto in the push for self-driving vehicles, and a car equipped with Baidu technology is set to debut at the Shanghai auto show in April. But last week Bloomberg quoted from a document that its boss Li has prepared for the government, which calls for subsidies and unified regulations to speed up development of domestic semiconductor skills. “The majority of our country’s laws and policies simply aren’t suitable for the development of self-driving cars,” he complained.
Nvidia is one of the leaders in high-performance computing because of its expertise in graphics processing unit chips (GPUs), which were first used in computer games. These chips deploy parallel process calculations (multiple cores performing simultaneous tasks) at much greater speed and energy efficiency than traditional computer processing units (CPUs), which take on one task at a time. Traditional computers can perform multiple processes by multi-threading but at much slower speeds. And that matters for self-driving cars, where cameras and sensors are transmitting data to GPU-based chipsets, which either steer the cars themselves or warn human drivers of dangers.
These new applications, called Advanced Driving Assistance Systems (ADAS), need to process vast amounts of data – Nvidia’s leading chipset has 7,500 gigaflops of processing power compared to 200 gigaflops for a standard Intel processor in a desktop computer.
China’s semiconductor firms simply aren’t capable of competing at this level. After all, its smartphone makers continue to rely on foreign chips for the manufacturing of most of their models, particularly higher-end phones. And if Chinese chips aren’t up to the task of streaming video or downloading apps on a phone, they aren’t going to be able to deliver the kind of processing power that keeps a car on the road.
But how about making it their mission to catch up quicker on GPUs than CPUs, employing a leapfrog strategy? Perhaps there is an opportunity for some of the country’s more deep-pocketed firms to take advantage of the tumult in the industry by partnering with the new challengers to Intel and Qualcomm, and even snapping up relative newcomers like Nvidia. One problem is that the price tags will be punishing and it’s not clear which of the next-generation chipmakers is going to triumph. And the greater obstacle is that Western governments are already wary of letting their prized assets fall into the hands of Chinese firms.
A deal with a Chinese suitor might hold out the prospect of easier access for the chipmaker to the Chinese market. But the same offer doesn’t seem to have worked in securing tech transfer in the latest chips for smartphones, so it’s hard to see why it might succeed in driverless cars.
Indeed, the global chipmakers may not even need to chase growth in China quite as aggressively in some of the newer applications, where they have more scope to make their first fortune at home.
In the meantime the Chinese firms that are going to rely on the next-generation chips are trying to assess their options by partnering with the companies that produce them. Nvidia is working with Baidu on a plan to roll out a commercially available self-driving car, for instance, and last September the two companies paired up to create a “cloud-to-car” mapping system for self-driving cars.
NXP Semiconductor has just announced that it is collaborating with Geely Auto, based in Hangzhou, to tailor its chips to Geely’s needs.
Another area where more high-performance computing is required is data centres, where artificial intelligence programmes demand massive amounts of processing power. State investment in sectors earmarked as strategic such as Big Data analytics and the Internet of Things might help here. Nvidia has also estimated that a third of the cloud centres around the world already using AI applications are based in China, where companies like Alibaba and Tencent are establishing mega data centres for cloud and machine-learning services.
But again, interest from the Chinese firms is more as the end users than manufacturers of the chips themselves. Advanced Micro Devices, one of the challengers to Intel in chips for servers, is working with Alibaba to improve graphics processing at its cloud-based data centres and AMD has also paired with Tianjin Haiguang Advanced Technology to develop server chips customised for the Chinese market.
It hopes these chips will address some of the data security concerns of the Chinese government. For the same reason Qualcomm has signed a deal with the Guizhou government for a joint venture called Huaxintong Semiconductors which will create customised server chips too. Guizhou may be one of China’s poorest provinces, but it has forged a niche hosting data centres for some of China’s biggest companies including the three telcos and Baidu and Alibaba (the latter has based its cloud computing development there). In true Chinese style, the provincial government has been building a new town – Bainiaohe Digital Town – to house them all.
© ChinTell Ltd. All rights reserved.
Sponsored by HSBC.
The Week in China website and the weekly magazine publications are owned and maintained by ChinTell Limited, Hong Kong. Neither HSBC nor any member of the HSBC group of companies ("HSBC") endorses the contents and/or is involved in selecting, creating or editing the contents of the Week in China website or the Week in China magazine. The views expressed in these publications are solely the views of ChinTell Limited and do not necessarily reflect the views or investment ideas of HSBC. No responsibility will therefore be assumed by HSBC for the contents of these publications or for the errors or omissions therein.