x
AI/Automation

Huawei Commits Up to $20B for Annual R&D, Fleshes Out AI Pitch

LONDON -- Huawei invested US$12 billion in its R&D in 2017 and is committed to invest between $10 billion and $20 billion every year from 2018, stated the company's president of the carrier business group, Ryan Ding, at a pre-MWC 2018 media briefing in London, where he also fleshed out the company's AI strategy.

Such investment talk highlights the gulf in financial muscle between Huawei and its network infrastructure rivals: Its R&D investment levels as a percentage of revenues (around 15%) may not be much different from key rivals such as Ericsson, Nokia and ZTE. But with revenues of $92 billion in 2017 and with further growth to come, it has the opportunity to invest billions of dollars more than those rivals in its tech developments. (See Huawei Hits $92B in 2017 Sales.)

To hammer home the firepower it has, Light Reading's back-of-envelope calculations show that Huawei's 2017 R&D investment was equivalent to more than 50% of the annual revenues of ZTE, almost 50% of Ericsson's 2017 revenues, and about 40% of Nokia's full year sales. (See Germany's 5G Auction & the Gigabit Dream, Nokia Outperforms Ericsson in Mobile but Sees Margin Pressure and Ericsson Stuck in Loss-Making Rut, Offloads Majority Stake in Media Unit.)

Huawei's key focus areas for those R&D investments are not surprising -- 5G, IoT, video and cloud, including investments in making its technology more "intelligent" by integrating analytics and (unspecified) AI capabilities. And not surprisingly, those topics are also front and center for Huawei's plans for the upcoming Mobile World Congress: The company plans to launch more than 20 products during the Barcelona event, with 5G, IoT, video and AI high on the agenda.

Ding says that, in 2017, more than 80% of the company's R&D spend went towards its carrier and enterprise technologies, which have a great deal of overlap and an integrated R&D team, while less than 20% went to R&D focused on devices. Ding added that the investments specifically assigned to 5G R&D would hit $800 million this year. (See Huawei's $800M 5G Budget Piles Pressure on Ericsson, Nokia.)

Intelligence quotient
The company has been touting the role of AI capabilities across its portfolio (carrier, enterprise and devices) for at least a year, but ramped up its AI marketing, including talk of greater investments, related to carrier and enterprise networking towards the end of 2017. (See Huawei Set for 'Intensive' AI Investment.)

Now it's getting more specific. The vendor has developed an AI-enabled compute platform called Atlas that it referenced last year but which it has now committed to launch in 2018. This is designed to be deployed in centralized cloud facilities/data centers to perform enhanced data processing and analytics.

In addition, in the radio access network, Huawei has developed what it calls Wireless Intelligence (WI) to help automate dynamic beam-pattern setting in 5G access environments. Currently, beam forming for 4G/LTE services requires about 200 parameters to be set but, according to Huawei, massive MIMO beam-pattern setting in 5G will involve more than 10,000 parameters, something that cannot be managed manually. Wireless Intelligence, noted Peter Zhou, CMO of Huawei's carrier business group, will identify the optimum parameters for 5G beam forming automatically, based on big data analytics and self-learning capabilities.

Huawei is also pitching the use of AI algorithms in its ‘Network Cloud Engine' transport network management platform to enable optimized configuration and provisioning based on, again, analytics and self-learning. And in its next-generation cloud-native packet core, Huawei plans to add AI capabilities to its Network Function Cloudification (NFC) platform to enable automated, on-demand microservices.

That seems rather ambitious and something that would be some way off, but what is already implemented is a level of self-learning in the vendor's network management systems: Huawei cites HKT, and an unidentified operator in Thailand, as existing users of predictive fault management capabilities and same-day service provisioning capabilities based on new AI and analytics capabilities.

All of which sounds attractive for operators, but is it applicable in many networks? Are these capabilities that, even in the future, could be applied in the multi-vendor network environments that operators want to deploy?

Answering Light Reading's questions on the matter, Ding stated: "We are aware that it is really very challenging to use AI in a multivendor environment and that is why we have already established 20 OpenLabs around the world. In these labs we will work with partners on realizing the end-to-end integration of the solution. In terms of AI applications, we want to retain the complexity on our side and to solve that in OpenLabs, so that for customers we can reduce the complexity of integration for them."

That's all very well, but if Huawei's OpenLabs are, indeed, open to its partners, they are not so open to Huawei's rivals -- the ones most likely to be deployed alongside Huawei in multi-vendor networks. The jury, it seems, is still out on just how applicable such advances might be if not deployed in a single vendor (i.e. Huawei) environment.

That looks like one of the many challenges Huawei will need to overcome as it starts to deploy AI software both internally, to make itself more efficient, and in its products, to help its customers become more efficient and embrace automation. The thing is, Huawei has the R&D dollars to develop a solution.

— Ray Le Maistre, Editor-in-Chief, Light Reading

kq4ym 2/22/2018 | 8:39:44 AM
Re: Relativity No matter the actual numbers the score of billions in R&D as noted will exceed the competition by a long margin allowing Huawei to command a huge advantage in developing new applications and products over the decade including AI which will most likely prove to be the area with the most upside potential.
mendyk 2/8/2018 | 2:11:39 PM
Relativity $10 billion to $20 billion is a pretty big spread.
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE