In 2025, Huawei Cloud faced losses last year and was plagued by internal management chaos and plagiarism scandals this year. The company recently announced a major restructuring focusing on the AI field, with reports suggesting that dozens of departments will be reorganized and integrated, potentially affecting thousands of employees. Experts believe that Huawei Cloud’s integration this time may be influenced by the strategic needs of the Chinese Communist Party, as Chinese cloud products may not compete with those of Europe and America, but could flow into developing countries, raising concerns about security risks.
On August 25th, according to reports from “The Paper,” in response to rumors of internal department and personnel adjustments, Huawei Cloud stated that it would “allocate more resources to the AI and computing industries through software-hardware collaboration and architectural innovation.”
It is reported that Huawei Cloud will invest more resources in Ascend Cloud, PanGu large models, digital intelligence integration, and provide more support for the development of Ascend CANN ecosystem.
Earlier, on August 23rd, “Sina Technology” reported that Huawei Cloud CEO, Zhang Ping’an, announced on August 22nd the restructuring of multiple departments to focus on the AI field. The company was in a loss-making state in 2024 and aims to achieve profitability this year, hence the decision to concentrate on the AI field. The structural adjustments may involve layoffs, but the specific proportion is unknown.
On August 24th, WeChat public account “Niheliujian” revealed that the restructuring at Huawei Cloud may involve core teams such as the product department, public cloud services department, and research and development department, impacting dozens of subordinate departments and organizations, potentially affecting thousands of people.
However, the revelation that the PanGu large model department was reorganized may not be accurate.
Since June, Huawei Cloud’s official website has issued announcements for the discontinuation or sales suspension of more than twenty products. Over the past few years, Huawei Cloud has attempted to cover enterprise mailboxes, domain name registration, cloud monitoring, and other basic services with a “wide-net” approach, but these businesses have mostly been unprofitable, with a lack of strong cooperation with the group’s Ascend computing power.
“Niheliujian” stated that recent high-level personnel changes point to one goal – to stop bleeding quickly under the pressure of profitability and slowing growth.
American economist David Huang analyzed that Huawei Cloud’s current restructuring and focus on AI are due to the slowdown in China’s cloud market growth and the fierce price competition in the general IaaS/PaaS market. Huawei Cloud must concentrate resources on AI native cloud and industry models with higher profit margins and greater growth potential. By removing and combining departments, funds and manpower can be concentrated on the PanGu large model, industry suites, and the Ascend computing power cluster.
However, from a military perspective, as Huawei is a key node in the Chinese Communist Party’s military intelligence system, the cluster and models of AI native cloud will serve scenarios such as satellite communications, battlefield image recognition, and electronic warfare data fusion in the so-called “civil-military integration.” By divesting non-core businesses and focusing on AI computing power and models, there may be considerations of Chinese Communist Party’s national strategy.
The United States currently imposes many restrictions on technology exports to China and Huawei.
Huang noted that the essence of U.S. export restrictions is to prevent computing power from being used for military intelligence and weapon simulations, primarily tightening the export of high-end GPUs, restricting China’s access to NVIDIA’s A100/H100 series. Although Huawei has its Ascend 910B/910C chips and MindSpore framework, their technology and efficiency lag behind, and can only substitute through multi-core integration and system optimization in some scenarios.
He pointed out that Huawei Cloud is gradually becoming the computing base of the Chinese intelligent warfare and intelligence system, but its ability to make breakthroughs in a blocked technical system relies on system design and policy support.
Earlier, the PanGu large model-related department had been embroiled in controversy over “shell plagiarism” and former employees exposing management chaos.
According to a Reuters report, Huawei launched the first generation of PanGu models in 2021, entering the field of large models early, but has been perceived as lagging behind competitors since then.
Four months ago, Huawei Cloud CEO Zhang Ping’an stated: resolutely build the Ascend AI computing base, resolutely reshape the industry with the PanGu large model, and resolutely advance ecosystem construction. Subsequently, the CloudMatrix 384 super node cluster based on the new high-speed bus architecture was brought online in the Wuhu data center, becoming the technical support center for the “Ascend + PanGu” combination.
On June 30th, Huawei officially released the open-source PanGu 7B dense model and the PanGu Pro MoE 72B hybrid expert model, and disclosed its model inference technology based on Ascend chips. However, the PanGu large model was soon caught in a plagiarism storm.
The research team HonestAGI accused Huawei’s PanGu large model on July 4th, claiming that its structure highly resembles Alibaba’s TongYi Qianwen-2.5 14B model, with a correlation coefficient as high as 0.927, far exceeding the benchmark in the industry of 0.7, suggesting that Huawei’s model may have been “upcycled.” Huawei refuted the plagiarism accusations on July 5th, but admitted that some basic component code was “referencing industry open-source practices.”
On July 6th, another person claiming to be a Noah’s Ark Lab employee at Huawei anonymously disclosed on GitHub that the PanGu large model had been “shelled” multiple times from Alibaba’s Qianwen and DeepSeek AI products, which were competitors.
Public information shows that in China’s cloud service market, Alibaba Cloud holds a 33% share, Huawei Cloud only holds 18%, Tencent Cloud holds 10%, and DeepSeek’s role in China’s domestic AI industry consolidation is drawing attention.
Huang stated that currently, Chinese cloud service providers, Alibaba is inclined towards e-commerce and public cloud, Tencent focuses on social and content, while Huawei focuses on government enterprises and AI native cloud. In the military sector, Alibaba Cloud assumes the commercial/financial side, Huawei Cloud takes on government affairs, energy, and military sectors, Tencent Cloud focuses on civilian data and content security, and DeepSeek provides tools for “cost consolidation” on a large scale.
He noted that due to export controls, Chinese companies face challenges competing directly with AWS/Azure/GCP in high-end markets in Europe and America. However, in the Middle East, Africa, and Southeast Asia, Huawei may enter the market with an “AI+security+military cooperation package,” challenging international competitors.
Huang warned that foreign users using Huawei/Chinese-funded AI cloud services need to be cautious about security, with data sovereignty and intelligence risks being top priorities, “because the Chinese ‘National Intelligence Law’ explicitly requires companies to cooperate with national intelligence work, which means commercial cloud data may be accessed by military and intelligence systems.”
Other security issues include cross-border data compliance, as Chinese authorities strictly scrutinize data, and foreign users must ensure that data is not defined as “critical data.” Furthermore, China’s AI models and computing platforms could potentially be repurposed for military use, such as using image recognition models for commercial security as well as battlefield monitoring.
Huang emphasized that China’s Ascend/CloudMatrix ecosystem is still unstable and may be subject to sanctions or stalled updates.