Alibaba Integrates DeepSeek AI Models into Cloud Services, Competing with Global Tech Giants

Alibaba Integrates DeepSeek AI Models into Cloud Services to Rival Microsoft and Amazon

Alibaba Group Holding's cloud-computing division recently announced the availability of DeepSeek's AI models on its platform, marking a strategic move to compete with other tech giants like Microsoft, Amazon, and Huawei, who have also integrated the Chinese start-up’s open-source systems for their clientele.

Alibaba Cloud issued a statement on WeChat emphasizing the platform's capabilities, where users can seamlessly conduct processes such as training, deployment, and inference without writing a single line of code. This advancement is aimed at providing developers and enterprises with a more streamlined and efficient experience in AI development and application.

Through the PAI Model Gallery, Alibaba Cloud users can now access and deploy DeepSeek’s models to enhance their reasoning and text-generating applications. The gallery showcases DeepSeek's cutting-edge AI models, including DeepSeek-V3 and DeepSeek-R1. These models are highlighted for their cost-effective development compared to other large language models (LLMs) created by major AI companies. Also available are distilled versions, like the DeepSeek-R1-Distill-Qwen-7B, which offer similar performance with reduced computational expenses.

Large language models are crucial to services such as OpenAI's ChatGPT, leveraging open source to enable extensive modification and distribution. The technique known as distillation, employed by DeepSeek and other companies, involves training smaller models to emulate larger ones, drastically cutting down resource needs while maintaining performance standards.

This initiative by Alibaba Cloud aligns with a rising trend among significant tech firms to integrate DeepSeek's models, enhancing value for their users. A similar effort was seen with Huawei’s cloud-computing operations, which linked with AI upstart SiliconFlow to make DeepSeek's V3 and R1 models accessible via their Ascend platform during the Lunar New Year.

Additionally, Tencent Holdings took steps to support DeepSeek's R1 model on its cloud, simplifying the setup process for users to under three minutes. Nvidia provided users access to DeepSeek-R1 through its NIM microservice, lauding the model’s exceptional reasoning, inference speed, and accuracy.

Microsoft recently extended R1 compatibility to its Azure cloud service, facilitating AI app development on Copilot+ PCs via GitHub, while Amazon incorporated R1 in its Amazon Web Services platform for developer application creation.

Despite these developments, some experts remain skeptical. Fudan University’s Computer Science Professor Zheng Xiaoqing noted DeepSeek’s reported training cost savings might be overstated, as they omit earlier architectural and algorithmic research expenses. He attributed DeepSeek's accomplishments primarily to engineering optimization, suggesting limited impact on chip procurement and shipment dynamics.