Huawei Cloud is expanding access to foundation models in South Africa with the introduction of its Model as a Service platform, a move aimed at reducing the cost and complexity developers face when adopting large language models and AI infrastructure.
The company introduced the platform during a workshop at Huawei Office Park in Johannesburg attended by more than 60 enterprise developers, customers and ecosystem partners including Vodacom, SITA, Inzalo and Altron. The event focused on how developers can use Huawei Cloud infrastructure to access, deploy and integrate AI models without building and managing their own compute environments.
The launch reflects a broader challenge facing enterprises experimenting with AI: while interest in foundation models continues to grow, training, hosting and maintaining those systems remains expensive and technically demanding. Huawei Cloud’s approach relies on token based access to prebuilt models running on its AI Compute Service infrastructure, allowing developers to consume models on demand rather than investing upfront in infrastructure.
According to Huawei Cloud, the platform provides access to multiple open source models including Deepseek, GLM and Qwen while supporting model deployment, training and development workflows through a single stack.
“In today’s world, AI is having a profound impact on development behaviours and product paradigms. Therefore, we are excited to launch our MaaS programme, allowing developers access to this technology at low cost. Ultimately, we hope to transform developer workflows by enabling them to unlock MaaS through our token ecosystem,” says Steven Chen, CEO at Huawei Cloud South Africa.
Huawei executives used the workshop to emphasize the role open source models are beginning to play in enterprise AI adoption. Calvin Huang, Director of Cloud Solution Sales at Huawei Cloud South Africa, argued that Chinese open source models are reducing performance gaps with proprietary alternatives while lowering infrastructure requirements.
“With an open ecosystem, you can have quick access to high quality OS models, which offer much lower pricing, our model’s cost is more than tenfold better than closed source solutions, while the model capabilities still meet enterprise needs. We can offer a full OS model portfolio, including Deepseek, GLM, Qwen and more,” he states.
Huang also said Huawei’s platform allows developers to adopt models without significant changes to existing workflows through OpenAI compatible interfaces.
“MaaS also provides for frictionless OpenAI-compatible onboarding – so developers don’t need to rewrite their whole coding stack in order to integrate their work with our top tier large language models (LLMs).”
“This makes it a simple matter for developers to switch from a closed source model to Huawei’s OS model in as little as five minutes – while still cutting up to 60% of their costs.”
Huawei also used live demonstrations to show how the platform handles multi agent workflows and deployment scenarios. Lerato Kekana, Cloud Solution Architect at Huawei, demonstrated how OpenClaw architecture could be deployed through the MaaS environment as part of a session focused on automation and orchestration.
The company says local developer engagement has become part of its rollout strategy. Since introducing MaaS in South Africa, Huawei has used events including the Vodafone Summit, FNB Code Fest and developer competitions to gather feedback and refine token services for local users.
Huang said the company’s position is that developers will increasingly move away from repetitive tasks toward higher value work as AI becomes embedded into engineering workflows.
“LLMs today are ideal for aspects like boilerplate and repetitive coding, natural language interfaces, content generation at scale, data transformation and extraction, prototyping and minimum viable product (MVP) development, as well as debugging and troubleshooting.”
“So what is the role for developers in this AI-driven future? It is vital that developers today focus on high-value, human-centric work; raising the bar on skills; human-AI collaboration; owning quality, trust and accountability; building unique IP and domains; and, ultimately, on advocating and shaping the future,” he concludes.
