HAS2019: Huawei Lays Out Its AI Service Strategy


In a perfect world for an ICT vendor (and for that matter, a customer) the ability to offer up a full stack of ‘all’ technologies’ to create a full stack of ‘all services’ is the ultimate goal. Here every new customer requirement could be met through an end-to-end life cycle while the incumbent vendor sustains a distinct competitive advantage. Vendors would offer up vertically integrated products and services and share a horizontal management platform across industries. This two by two matrix is all but impenetrable to competition and tightly managed by the vendor. After attending Huawei’s annual analyst summit, HAS2019, in Shenzhen, China I came away with the impression that this was a leading goal of its own ICT strategy. On Huawei’s journey to become a full stack Artificial Intelligence supplier, it plans to execute on that strategy by initially offering a full stack Big Data service running on its full stack Cloud Computing platform.

Let’s give Huawei credit where it is due, there are few if not any ICT vendors left who have the resources, capabilities and product portfolio to deliver this business model. Huawei may be the only one to do so. However, is this the right strategy for today’s ICT and related business services model? Is this what customers want from vendors? How will competitors respond to Huawei’s strategy?

Huawei’s Cloud Computing

Let’s start with cloud computing. The foundation of today’s ICT infrastructure is built upon a robust cloud computing offering. Already we are seeing this market continuously evolve from everything on a single public or private cloud to a world of interconnected multi-cloud platforms supporting hybrid cloud offerings. These complicated permutations for customers would suggest that Huawei’s strategy of offering the full cloud stack removes process friction from the cloud infrastructure management and generates a more simple environment. Here, the combination of Huawei’s full stack cloud platform combined with its full stack ecosystem and services offers up a solution whereby a customer’s transition to a hybrid cloud solution can be accelerated. Huawei’s global cloud presence is significant and growing. Currently it can provide cloud services across 23 geographies within every major region except North America and Australia.

Competitive Viewpoint: Cloud service providers (CSPs) know that future business models will be based off multi- and hybrid cloud environments. For CSPs to be successful they need attract and retain customers to be able to migrate data away from their competitors and onto their own platform.For example AWS already claims that it runs more Microsoft compatible cloud instances on its platform than Microsoft. As customers build next generation business solutions fed by a broad range of data sources, it becomes imperative for their cloud vendor to be able to curate these data sources and feed their own analytics and big data engines to generate artificial intelligence (AI) outcomes. AWS, Microsoft, Google, Alibaba and others will not freely promote Huawei to migrate their customers’ data and vice versa. However, the ability to aggregate varied data sources is an ICT data management software opportunity and this market is mature with several established vendors. In addition these vendors have strong cloud platform positions with Huawei’s competitors.

With these market conditions in play, Huawei should be prepared to accept that while their full stack cloud strategy covers all the needs of their customers, it will be a goal that is almost impossible to act upon. We would suggest that Huawei market its ability to be an open, compatible and global platform offering up a wide range of customer opportunities for the software based data management companies to build their AI-based solutions. To go to market on this will also mean that it can offer up a robust and more focused big data / analytics strategy.

Big Data

Huawei has made the commitment to build an intelligent big data platform and has claimed to be already in 60 countries, signed up more than 1,500 customers and 500 partners. While we cannot verify these numbers, we believe that most of Huawei’s larger big data customers are local telco companies such as China Mobile and China Unicom along with financial institutions such as the Shanghai Stock Exchange, China Construction Bank, China Merchants Bank and Bank of Communications.With Huawei’s known strength in specialized microprocessors, its knowledge in high performance computing, distributed databases, and broad product portfolio in traditional ICT infrastructure (especially storage and compute) we believe that this is an opportunity to be successful.

Competitive Viewpoint: With Huawei’s global cloud computing footprint combined with its strength in processing big data, we believe that this is one of their best opportunities to be the ‘engine of choice’ for many big data/ analytics software vendors. While this again thwarts Huawei’s full stack mission, it makes them a very important and integrated part of a customer’s digital transformation – something that is missing in Huawei’s marketing position and messaging. Digital transformation is only as good as the fuel from digital outcomes based off data analytics. When a customer deploys a multi-cloud strategy, Huawei should position its cloud platform as the analytics engine. However, for the analytics process to be complete, data sources need to be collected from across the end-to-end infrastructure and therefore any vendor needs to develop not just a strong enterprise solution of data analytics but also an edge computing strategy.

Edge Computing.

Edge computing has many definitions, but for the sake of this document, we are simply implying that any device in a distributed network that has the ability to process data complies with an edge computing model. With this in mind, there has been a rush to build out edge computing capabilities by CSP traditional ICT vendors and of course industry specific Operations Technology (OT) companies. Processing at the edge of the network is much more complicated than many vendors realize or appreciate. While most lead off with offering up security services as their differentiator, very few actually know how to perform data and systems management services. These management services are critical to be able to perform any level of analytics and ultimately generate AI outcomes. AI services will mature into mission critical offerings that drive and support digital transformation in markets such Industry 4.0, autonomous driving, and other latent-sensitive applications.

At HAS2009, Huawei announced its strategy to provide customers with the ability to process data at the edge of the network while emphasizing the importance of synergy between cloud computing, the edge, and the endpoint nodes. A more valuable position would have been to stress the need for synchronization between these three access points. Quoting from a white paper from the Edge Computing Consortium ‘With the rapid development of machine learning, deep learning and edge AI chips, AI inference is extending to edge from cloud’, Huawei announced ‘KubeEdge’.

KubeEdge’s mission is to brig AI to the edge of the network. Key features from KubeEdge include a lightweight container runtime, edge computing with GPU and NPU chipsets, automatically deploy applications at the edge while maintaining edge-cloud synchronization and coordination.

Competitive Viewpoint: Bringing AI to the edge of the network is an ambitious goal for Huawei to attempt. The AI process steps of learning, inference, and training is typically performed at the Enterprise where compute resources are in abundance. Companies such as nVidia, Microsoft, Google and AWS lead the AI market and are carefully trying to understand what it takes to generate AI-based outcomes at the edge of the network.

Offering a AI environment using lightweight containers is a starting point but there is so much more hardware and software work to be done to make this not sound like it’s a science project and a ‘me too’ strategy. Huawei’s current strength is firmly in its chipset road map offerings of Kunpeng (arm-based cpu for datacenter workloads), Ascend (AI chipset based on unified scalable architectures), and CANN (chipset operators API and tools library.

Conclusion and Recommendation

For Huawei and its customers: Huawei’s AI services solutions is predicated on how successful it executes its cloud computing strategy. Having global cloud services are table stakes in this race to curate and process as many varied data sources to ensure the most valuable AI services. However in the emerging multi-cloud / cloud interconnect world, Huawei will be faced with step competition from all of the CSPs who will actively block any attempt to migrate customers’ data to the Huawei Cloud. Huawei’s marketing strategy should not be to position itself as the full stack provider but rather as the AI market enabler for a couple of reasons:

1. Many customers understand the value of vertically integrated solutions (as they work very well in consumer markets) by reducing complexity, improve integration but increase vendor lock-in while reducing their own competitive innovation. However, in the customers’ minds, this equates to more expensive operating models. We recommend customers seek the best of breed at every stage of an AI workflow, and include Huawei where they meet that criterion.

2. Huawei and its customers should look at the places where Huawei’s strong product portfolio in infrastructure and processing create more competitive value for the customer’s AI strategy and partner with the best of breed in the remaining places. A Hybrid AI solution will generate more innovation, attract more developers, and accelerate AI deployments. By taking this approach, Huawei is embracing its ICT strength while acknowledging that the AI/analytics market is being disrupted as many levels by vendors. These vendors include:

a. Snowflake (data warehousing)

b. Tableau (data visualization)

c. Talend (Data Integration)

Finally, any AI solutions market will be driven by the creation of the most intense customer centric ecosystems. End user customer ecosystems are complicated and difficult to manage. However, in the era of creating and delivering AI services, Huawei and its customers should look to partner with the best ecosystems ‘manager’ – one who has the voice of the customer but has the digital experience to create the ecosystems management platform. These platform vendors will not be traditional systems integrators but rather be customer engagement and services vendors. Business models will be based off subscription and consumption frameworks and everyone, including Huawei will be expected to have fully transformed themselves into digital business services companies. This is not an easy task, but one that we feel Huawei has the ability to adapt over time.