DYXnet Launches AI Computing Solution - Accelerating Enterprises’ Private DeepSeek Deployment

Responding to the vigorous demand for tailoring AI with private LLM, DYXnet — a wholly-owned subsidiary of VNET Group (NASDAQ: VNET) and a leading enterprise network service provider in Greater China — has introduced "DeepSeek-in-a-Box", a one-stop DeepSeek AI computing solution. This offering integrates model deployment, computing power, storage, and network resources, addressing the needs of enterprises for model training and AI application deployment. Supported by DYXnet's specialized AI technical team, this solution enables businesses to utilize DeepSeek and achieve innovative advancements effectively.

Comprehensive "DeepSeek-in-a-Box" Powered by Dedicated NVIDIA GPU
DeepSeek has gained significant attention globally following the release of its DeepSeek-R1 model earlier this year. The model's "Deep-Think" feature excels at tasks that require logical inference, with performance comparable to OpenAI's ChatGPT-4o. Its cost-effective MoE architecture has lowered the threshold for enterprises developing dedicated AI systems. Additionally, the model's open-source nature allows for creating private LLM, enabling developers to foster innovation by tailoring the model to their specific needs while enhancing security, controllability, and compliance. According to a report from CICC Investment Bank, the total sales of China's DeepSeek AI server market are projected to reach 54 billion yuan by 2025. Approximately 5% of this demand is expected to come from sectors that handle sensitive data, such as government, finance, public services, and healthcare.

Due to the elevated demand for private DeepSeek deployments, DYXnet offers enterprise customers "DeepSeek-in-a-Box". This AI computing solution is delivered as Infrastructure-as-a-Service (IaaS), integrating DeepSeek model deployment, computing resources, storage, and network capabilities in one package, which can be delivered in as short as 48 hours. The solution not only helps enterprises avoid the upfront costs of purchasing hardware but also guarantees complete AI technical support.

DYXnet's DeepSeek AI computing solution is designed to tackle enterprises' main challenges when privatizing DeepSeek models, including model deployment, fine-tuning, and application. By aligning various DeepSeek versions with the appropriate GPU resources and optimizing them through fine-tuning, ensuring optimal performance and a ready-to-use experience. DYXnet's specialized AI team also provides professional technical support in model selection, hardware configuration, operational updates, and services such as model fine-tuning, evaluation, deployment, data management, knowledge base development, and system integration. The solution effectively meets the diverse needs of customers across different industries.

"DeepSeek-in-a-Box" is available in three packages: "Basic," "Advanced," and "Full Power," corresponding to the DeepSeek-R1-Distill-Qwen-32B, DeepSeek-R1-Distill-Llama-70B, and DeepSeek-R1 models, respectively. All utilize NVIDIA RTX 4090 graphics cards or H20 chips, allowing enterprises to leverage the performance of dedicated NVIDIA GPU.

Enhance Data Transmission Quality and Security with Dedicated Line for AI
DYXnet's parent company, VNET Group, is a leading player in the domestic data center industry and has actively transformed toward the development of AIDC (Artificial Intelligence Data Center) in recent years to meet the substantial demand from corporate clients for AI digital transformation. Leveraging VNET's robust infrastructure resources, along with DYXnet's 25 years of expertise in enterprise networking and its experience in integrating AI with cloud, network, and cybersecurity services, the DeepSeek AI computing solution further enriches its AI full-chain layout and responds to the urgent demand for AI innovation from enterprise customers.

DYXnet is accelerating the construction of AINet, a reliable and high-speed dedicated line for AI. With exceptional network services and a robust backbone, DYXnet enables businesses to transmit data seamlessly. It also creates a secure private computing environment for customers, maximizing the value of their data and computing resources while ensuring data security.

DYXnet is committed to working hand in hand with customers, helping enterprises seize opportunities in a rapidly changing market and achieve business innovation and growth. In the future, DYXnet will continue to optimize and upgrade its solutions to meet the evolving market demands and promote the widespread application of AI technology.

To learn more about the benefits of private DeepSeek deployment and DYXnet's AI computing solution, please visit: https://www.dyxnet.com/hk/deepseek-in-a-box/
Hashtag: #DYXnet #AI #DeepSeek


https://www.dyxnet.com
https://hk.linkedin.com/company/dyxnet

The issuer is solely responsible for the content of this announcement.

DYXnet

DYXnet, established in 1999 in Hong Kong, became a wholly-owned subsidiary of VNET Group (NASDAQ: VNET) in 2014. It is a leading ICT service provider in Hong Kong and the Asia Pacific region, delivering innovative technologies and solutions to enterprises of all sizes and helping them fully embrace digital transformation. DYXnet offers a comprehensive portfolio of services, including enterprise networks (MPLS and SD-WAN), cloud solutions, data center services, cybersecurity, and AI solutions, empowering businesses to realize their full potential.

As a pioneer in the ICT field, DYXnet operates over 200 PoPs worldwide, covering more than 700 cities and serving over 30,000 client sites. DYXnet was among the first ICT service providers in the region to obtain prestigious ISO certifications, including ISO/IEC 27001, ISO/IEC 20000-1, and ISO 9001. Additionally, DYXnet is one of the inaugural official members of the China Cross-border Data Telecommunications Industry Alliance and has played a crucial role as one of the initial drafting units for SD-WAN service standards.

To learn more, please visit DYXnet's website at: https://www.dyxnet.com.