Huawei eyes AI prowess, invests in compute power

Chinese tech giant unveils ambitious plans to build the architecture needed to meet growing demand for more compute power and artificial intelligence, pledging to do so on an "open" ecosystem" and launching an AI training cluster that it says is 10 seconds faster that the current ResNet-50 record.
Written by Eileen Yu, Senior Contributing Editor

The world needs more computing power and Huawei wants to build the architectures needed to answer the call, spanning processors, networks, devices, and cloud services. It also believes artificial intelligence (AI) will fuel much of this need and is gearing up a "full stack" portfolio to tap the growing enterprise demand. 

Statistical computing, which is needed in dealing with undefinable tasks such as voice and image recognition, will soon become mainstream and Huawei believes AI computing, five years from now, will account for more than 80% of computing power used worldwide.

At the Chinese tech vendor's annual Connect conference Wednesday, deputy chairman Ken Hu noted that training AI algorithms requires a metric ton of computing power, while other more complex applications such as autonomous driving and weather forecasting requires even more compute power. 

Hu added that computing and intelligence would become ubiquitous, where they would be applied across multiple devices including headphones and specialised edge computing. 

This would mean that computing becomes the "bedrock of everything" in future, he said, pointing to Gartner's projections that the computing market will be worth more than $2 trillion by 2023. 

"The industry doesn't have nearly enough computing power to meet demand," he said. "Compute supply relies on processor performance and since we're reaching the limits of Moore's law, if the industry wants to provide a steady and abundant supply of affordable computing power, we need to make breakthroughs in processor architecture."

He added that Huawei would "invest more" to drive its computing strategy, by developing the architecture needed to enable both computing and AI to be readily available. This encompassed development efforts in its Da Vinci chip architecture and portfolio of processors including Kunpeng, Ascend, Kirin, and Honghu, each of which supported specific functions such as AI, smart devices, and smart screens.

He noted that Huawei would not sell its processors directly to customers and, instead, would make them available in the form of cloud services. Its partners also would access them as components, so these could be integrated into their own products as part of a wider solution. These encompassed all of the vendor's hardware, including AI servers and accelerator cards.

Its software, too, would be open source, including server OSes, databases, and AI development frameworks. 

Stressing its aim to build an "open ecosystem", Hu said Huawei would look to offer tools to help its partners develop and port their applications. 

In this aspect, he announced plans for the company to invest another $1.5 billion to beef up its developer programme, making it available to five million developers worldwide. Introduced in 2015, the scheme since had supported more than 1.3 million developers and 14,000 independent software vendors across the globe. 

Eye on AI with 'full stack' portfolio

At the conference, Hu also pointed to Huawei's goal to offer a "full-stack, all-scenario" AI portfolio, which currently comprised its Ascend 310 and 910 processors, ModelArts application development platform, and MindSpore AI computing framework.

It now was adding its Atlas 900 AI training cluster, touted to take 59.8 seconds to train ResNet-50, which was the industry standard for measuring AI training performance. This was 10 seconds faster than the previous world record, according to Huawei. 

Running on Ascend chips, the Atlas 900 also had been deployed on Huawei Cloud as a cluster service, the vendor said. 

Hu said the AI computing platform offered "new possibilities" to various fields of scientific research and business innovation, including astronomy and oil exploration.

Huawei also launched a suite of AI products including the iMaster NCE autonomous driving network management and control system,iMaster NAIE network AI platform, and OceanStor Dorado intelligent all-flash storage platform. 

Hu said: "Today, all industries are rushing to go digital, and artificial intelligence is all the rage. The computing industry is booming, and will only continue to grow. We'll continue to invest. Starting with the most difficult challenge ahead of us--making breakthroughs in architecture, to developing processors--we're going to help expand the industry and build out the ecosystem.

"We strongly believe that facing challenges head-on is the only way to build our competitive strengths [and] the only way to build out the market for our partners," he added. 

Huawei last month shared details of its new Harmony OS software tipped for use in smartphones, smartwatches, and other devices as a potential replacement for Google's Android. This development came after the US government added the Chinese tech vendor to its 'Entity List', which prohibited US companies from transferring technology to companies on the list without a licence from the US government.


Huawei unleashes AI chip, touting more compute power than competitors

Chinese tech giant announces the commercial launch of its Ascend 910 artificial intelligence (AI) chip, touting its low power consumption of 310W, as well as its AI computing framework MindSpore, which will go open source in the first quarter of 2020.

Huawei Cloud ramps up AI efforts with Singapore lab, partners

Chinese tech giant has opened a cloud and artificial intelligence innovation lab in Singapore with resources to help universities and enterprises drive research in these technologies as well as inked agreements with various companies to jointly develop such applications in Asia-Pacific.

Singapore to feel impact of China-US trade dispute if prolonged

Tensions between the two economic powerhouses will lead to fragmented technology markets and, if prolonged, drag down Singapore's economy, says Prime Minister Lee Hsien Loong, who urges both countries to find ways to establish mutual trust.

VMware: China-US trade dispute will split industry into two camps

With the China-US trade dispute unlikely to be resolved soon, technology vendors should be prepared to support two trading blocs--one that is heavily partnered with China and the other with the US, says VMware CEO Pat Gelsinger, while Dell CEO Michael Dell reveals the company has made "adjustments" to due to the conflict.

Alibaba to launch own AI chip next year

Chinese internet giant sets up a semiconductor company and unveils plans to release its own artificial intelligence processor, as it looks to boost support for its cloud and Internet of Things businesses.

Editorial standards