Is the Huawei version of ChatGPT coming? The product is mainly aimed at ToB/G customers, and industry chain related stocks are expected to benefit from it

On June 3rd, Cailian News Agency reported that titanium media recently learned exclusively from Huawei that Huawei will release a multimodal, billions level large model product called "Pangu Chat" that directly targets ChatGPT. It is expected that Huawei Pangu Chat will be released to the public and tested internally at the Huawei Cloud Developer Conference (HDC. Cloud2023) held on July 7th this year. The products are mainly aimed at ToB/G government and enterprise clients.

With ChatGPT rapidly becoming popular globally, the craze for AI large model tracks has surged from abroad to China, and has rapidly heated up accordingly. At present, giants such as Alibaba, Tencent, Baidu, and 360 are accelerating the layout of big models. Huawei's entry means that in the field of big models, giant enterprises such as Alibaba and Tencent will face strong competitors.

Huawei has been laying out large-scale models since 2020. Pangu large-scale model is a series composed of CV (computer vision) large-scale model, scientific computing large-scale model, NLP (natural language processing) large-scale model, voice large-scale model and multimodal large-scale model, which will be officially released in April 2021.

At present, the Pangu Big Model has been validated in over 100 industry scenarios, including energy, retail, finance, industry, healthcare, environment, logistics, etc. The application of the Pangu Big Model can be seen from both industry and scenario dimensions:

In terms of industry: The CV model can be applied to fields such as industrial quality inspection, logistics warehouse monitoring, and fashion assisted design. The NLP big model can be applied in fields such as intelligent document search, intelligent ERP, and small language big models. Scientific computing models can be applied to fields such as meteorological forecasting and wave prediction.

In terms of scenarios: The Pangu Mining Large Model has achieved functions such as panoramic video splicing of fully mechanized mining scenes, ensuring safe production; The CV large model has achieved an accuracy higher than that of human inspectors in the wayside image detection system for freight car faults; The meteorological large model has achieved accuracy exceeding traditional forecasting methods.

The core positioning of Huawei's Cloud Pan Gu Big Model is to empower various industries, divided into three levels: L0 is a basic universal big model similar to GPT3; L1 is a large-scale industry model that combines basic models with industry data for mixed training. It has published large-scale models for industries such as mining, meteorology, pharmaceuticals, molecules, electricity, waves, and finance; L2 is a deployment model generated by deploying L1 in specific downstream business scenarios, such as financial OCR models, power inspection models, etc.

Guojin Securities believes that the Pangu Big Model has strong generalization ability and is suitable for a large number of complex industry scenarios. With the opening of the Pangu Big Model, developers from all walks of life no longer need to "start from scratch" and only need to find the models they need on the cloud. The Pangu Big Model serves as a universal "wheel" for AI in all industries, helping accelerate intelligent transformation in all industries.

Dongwu Securities believes that the advantages of Huawei's Pangu Big Model lie in talent reserves and controllable computing power, which is expected to become a leading big model in China. Its ecological industry chain is expected to accelerate development. Dongwu Securities has sorted out the relevant stocks of Huawei's Pangu Big Model industry chain as follows:

It is worth mentioning that compared to the previous market's emphasis on high-frequency C-end creative applications in the AI application market, "Pangu Chat" mainly targets ToB/G government and enterprise customers, which can be said to be a "new path". CICC believes that there are also many scenarios where large models can be utilized in heavy applications on the B-end:

Integrating AI modules in general enterprise service software (ERP, PLM, MES, etc.) is expected to achieve a change in interaction mode, lower the application threshold of the software, and improve usage efficiency. In the industrial field of vertical subdivision, a large amount of underutilized data has also been accumulated. The application of AI technologies such as CV and NLP is expected to empower industrial scenarios, improve intelligence levels, and achieve cost reduction and efficiency increase in production.

The landing scenarios for large models on the B-end are also rich, and CICC suggests continuing to pay attention to investment opportunities in general enterprise service tracks, vertical segmented industry tracks, and other industrial scenarios:

General Enterprise Service Track: Recommend Saiyi Information (a major partner of Huawei MetaERP) and UFIDA Network (a participant in the development of enterprise service big models), and suggest following Hande Information (a partner of Baidu Wenxin big models);

Vertical segmentation of industry tracks: It is recommended to use Beilu Intelligent Control and Longsoft Technology for the intelligent coal industry, and it is recommended to pay attention to Yunding Technology; Intelligent recommendation of central control technology in the process industry (with obvious advantages in scenarios, patterns, and data, looking forward to the integration of products and AI implementation);

In other industrial scenarios, CICC recommends Rongzhi Rixin and Dongtu Technology, and suggests paying attention to Huolaiwo.