When it comes to cloud domination, there’s a lot of talk around what the Big Three are doing. In the West, Amazon (AWS), Microsoft (Azure) and Google (GCP) lead the public cloud market, and it would be easy to assume that this ranking is pretty much the same wherever you go. But then there’s China, with its own cloud market that will triple in size by 2027, according to projections from the China Academy of Information and Communications Technology, a government think tank. China now has the world’s second-largest cloud market after the US, with Alibaba Cloud, Huawei Cloud, Tencent Cloud and Baidu being the main players. But unlike the rest of the world, China is not dependant on the US hyperscalers and is actively expanding its cloud companies’ footprints across Asia-Pacific, Europe, Latin America and Africa.
In 2024, Huawei opened the first major cloud region in Egypt. The year before, Alibaba launched its cloud regions in South Africa and Mozambique through a partnership with BCX. Under the African Local Public (ALP) Cloud model, BCX can now offer Alibaba’s Aspara Stack – the same platform trusted by Chinese government agencies, state-owned banks and telcos.
According to a BCX spokesperson, the ALP Cloud gives regional customers a way to keep critical data under local laws and plan cloud spend in local currency, which can help offset currency swings and cross-border compliance concerns. What sets the Chinese cloud market apart is that it’s never about selling cloud services in isolation. Building the physical infrastructure to support cloud growth has always been part of the same strategy, such as fibre networks, stable power, robust datacentres, and high-speed connectivity to keep cloud platforms up and running. Huawei, for example, spent decades laying the groundwork for this ecosystem by supplying core network equipment, rolling out mobile base stations, expanding fibre routes and partnering with local telcos to build reliable national coverage. It was the first cloud provider to launch a datacentre in South Africa in 2018, but that only happened after being in the country for almost three decades.
And then the sanctions came, and they gave us 30 days to move completely out of their cloud.
Calvin Huang, Huawei Cloud
“The only cloud vendor doing hardware and software at the same time is Huawei,” says Calvin Huang, head of solution architecture for Huawei Cloud South Africa. He says this gives Huawei more freedom to design systems that run efficiently, while keeping costs under control. The company builds its own servers, switches and storage, then develops the cloud software layer to match the hardware’s performance needs. In practice, this means better coordination between components such as CPUs, memory and NPUs, and fewer limits when new technologies need to be integrated.
He sees this control as part of how Huawei manages the basics that make cloud usable in the first place – stable power, resilient cooling and the network infrastructure to reach customers wherever they are. “If you don’t have connectivity, you are hardly [able] to enjoy the benefit of cloud,” he adds. “We want to build a cloud and a network that will boost the digital economy of this country.”
Keeping hardware production in-house also affects how Huawei charges for its services. By avoiding margins from suppliers, costs can be planned more precisely and there is room to offer flexible terms rather than fixed minimum commitments. Huang says this is important for organisations that want to test or move workloads gradually, particularly in cases where budgets are tight and there is a need to show measurable returns rather than just lifting everything into the cloud at once. In sectors like financial services and the public sector, this combination of network reach, local datacentres and control over the technology stack has made a substantial difference. BCX says the same principle underpins its partnership with Alibaba Cloud, offering local hosting and billing that gives African organisations more control over where their data sits and what it costs. Organisations want to know exactly where their data sits and who manages it, which is easier to demonstrate when the core infrastructure is owned and run by the same company.
In China, strict privacy and security laws mean regulated companies must keep personal and sensitive information on the mainland, under local jurisdiction. This backdrop has shaped how companies in China actually use cloud services. While consumer-facing businesses such as Alibaba and JD.com have driven much of China’s early cloud boom, handling huge spikes in e-commerce, payments and live-streamed sales, the next wave is expected to come from sectors like manufacturing, logistics and heavy industry. According to the ‘Cloud in China: The outlook for 2025’ McKinsey report, around 32% of industrial IT workloads in China are expected to shift to the cloud by 2025. The demand for private-cloud customisation is very high in China, and this can constrain scalability and profitability. McKinsey points out that local companies often choose private cloud because they have security concerns about public cloud customisation. Regulated industries like banking prefer to keep critical systems in-house, which is one of the reasons hybrid cloud has become so widespread in China. Vendor lock-in is very much on the radar and multicloud has become the norm, which means Chinese organisations often blend providers to reduce dependence on any single platform.
According to Forrester’s ‘State of Cloud in China 2025’ report, 97% of Chinese organisations already use hybrid cloud and 92% run multiple public clouds. “We don’t use technology to lock in our customers; we encourage them to adopt multicloud,” says Huang, who explains that one of the biggest differences between Chinese and US hyperscalers is that they don’t treat cloud vendors as exclusive technology providers. “They’re free to choose which cloud vendor they want to use. That’s why most of the American players don’t get a lot of market share in China,” he says, “especially in government or finance. They really need to escape from American technology, as there’s too much uncertainty.”
Huang relates the story about how Huawei was once a big AWS customer – almost in the top five. “It was before we had cloud. And then the sanctions came, and they gave us 30 days to move completely out of their cloud,” says Huang. “This is a real thing, so we think it’s very important to give the world another choice.”
CHINA’S AI STRATEGY
For China’s cloud ecosystem, AI is no longer just another workload. It is becoming the defining driver of how new clusters are funded, where they are located and how cloud providers pitch their value. Training LLMs demands uninterrupted access to thousands of GPUs for days or weeks, which helps fill capacity that once served more conventional enterprise needs. Keeping that training inside China’s borders is now standard practice, partly to comply with strict data security rules, partly because domestic providers want AI demand to fill racks that once served more conventional enterprise needs. Cloud giants such as Alibaba, Tencent and Huawei, along with the big state-owned telecoms, have spent years building out hyperscale capacity. Some of that capacity has sat idle or underused, but the surge in AI workloads is beginning to close the gap. DeepSeek’s training runs, for example, tapped clusters across multiple providers to secure enough GPUs to build models tuned specifically for Chinese language and local applications. Huawei Cloud has leaned into this shift, positioning itself as a strategic option for local developers working around advanced chip restrictions. While China still faces limits on cutting-edge chips, domestic cloud operators have enough capacity to funnel resources where they are needed most. Huawei’s leadership continues to argue that China can train large models at scale without foreign compute if local capacity is fully used.
Energy costs and location choices matter, too. Many AI focused clusters are built inland, closer to cheaper hydropower and industrial rates that help offset the huge electricity draw of large training runs. As more renewable energy comes online, local governments and operators see big AI jobs as an anchor customer that justifies new grid investment and steadies power demand. Alibaba has committed to invest $53bn in AI and cloud, more than its total spend over the past decade. Once cautious after regulatory crackdowns and slowing profits, Alibaba is now betting that AI training and deployment will keep its racks busy and secure its place in China’s next tech phase. That said, not every cluster is running at full tilt yet. As MIT researcher JS Tan points out, China’s domestic cloud sector still carries structural gaps that have kept parts of this massive capacity underused for years. Many businesses have been slow to adopt the software tools and higher-value cloud services that keep racks filled long-term. Price wars, a shallow domestic enterprise software market and low IT spending all limit how far providers can move beyond basic, low-margin infrastructure. This makes large AI training jobs one of the few reliable ways to soak up idle capacity and justify new investment, but moving workloads and securing stable power still takes coordination that does not always keep pace with the speed at which labs want to launch new models. DeepSeek, Baichuan and Zhipu are all developing large language models that lean on the same domestic clusters. Each new project adds pressure to keep GPUs local, secure affordable power and prove that China’s cloud can deliver the scale its AI ambitions demand.
* Article first published on brainstorm.itweb.co.za
Share