In an effort to increase technological independence, Chinese companies have begun promoting “AI in a box” solutions for local deployment in customer infrastructure. However, this could pose a threat to the monetization of cloud services provided by giant corporations like Alibaba, Baidu, and Tencent, according to the Financial Times.
The leading player in this arena is Huawei, who has partnered with over a dozen AI start-ups to package their powerful language models coupled with its AI processors and hardware. Among the start-ups are Beijing’s Zhipu AI and iFlytek, a leader in speech synthesis. iFlytek’s founder, Liu Qingfeng, has attested to the AI-box offering “premium performance, superior safety, and ease of management, ready for usage right out of the box”.
The principle behind “AI in a box” allows for the deployment of advanced generative AI models directly to a customer’s local or private cloud infrastructure. Huawei predicts a staggering CNY16.8 million ($2.3 billion) worth of China’s “all-in-one” AI services market by 2024, and a more ambitious projection by Minsheng Securities pegs it at reaching CNY450 million by 2027.
The flurry of tenders from major state-owned companies reflect the growing demand for such technological products. An example can be seen in Chengdu’s “smart city” service provider requesting a CNY2 million bid for the deployment of a grand language model into China Unicom’s infrastructure leveraging Huawei’s equipment.
However, the rise of “AI in a box” could coincide with limiting the ambitions of tech giants like Alibaba, Baidu, and Tencent. These companies have poured massive resources into AI infrastructure and large language models intended to be offered as cloud-based services to corporate customers. Moreover, the proliferation of this local model could exacerbate competition and cause further fragmentation in China’s cloud market.
Despite the aggressive expansion, it’s noted that in the long-run local AI deployment may not be as efficient as using cloud-based AI for access to powerful language models. Additionally, customers must allocate substantial resources for maintaining locally deployed systems.
In conclusion, Chinese start-ups are aggressively promoting “AI in a box” for local deployment in state corporation infrastructures and large-scale businesses. In addition to meeting customers’ need for confidential data protection, it bypasses the shortage of computational power led by the US’s chip export restrictions. However, it brings the monetization plans of tech giants’ cloud services based on advanced AI models into question.
