With the rapidly increasing demand for AI applications worldwide, the GPU subscription and rental services market is experiencing a significant surge. Mr. Hans Huang, the General Manager of Cloud Business Division at eASPNet, has identified various factors that are contributing to this trend. First, as enterprises actively develop AI applications, they require significant GPU computing power. Secondly, the global rush to embrace AI has led to intense competition for computing resources, making GPUs scarce and expensive, with long waiting times for procurement. This has driven up setup costs and entry barriers, prompting enterprises to adopt a rental model. Additionally, GPU servers impose stricter requirements on data center infrastructure—such as cooling systems, power supply, floor load capacity, and network transmission—posing further challenges compared to traditional servers.
To address these issues, many businesses have shifted to GPU rental services to ease the pressure of upfront investments. Responding to this demand, eASPNet has launched its GWS AI Cloud virtual GPU subscription platform, tailored to meet the needs of small and medium-sized enterprises (SMEs).
Expertise in Cloud and Data Centers: Laying the Groundwork for Efficient GPU Services
Mr. Huang emphasizes that eASPNet’s extensive experience in managing data centers and cloud services enables it to deliver high-performing and stable platforms. GPU computing power is a natural extension of their expertise. Furthermore, eASPNet’s data center hosts one of Asia’s largest network exchange hubs, offering abundant network and bandwidth resources to ensure stable support for GPU computing and AI applications.
In addition to GPU services, eASPNet has developed diverse cloud offerings for government agencies, educational institutions, and various industries. This allows GWS AI Cloud users to access GPU computing power while also benefiting from complementary solutions such as cybersecurity protection.
Scalable GPU Rentals: From 0.25 to Multiple GPUs
GWS AI Cloud leverages NVIDIA H100 GPUs, meeting the robust requirements of large language models (LLM) and machine learning (ML) training. The platform provides an intuitive graphical management interface, replacing the traditionally complex command-line operations often associated with AI/ML training tools. Preconfigured training and inference environments further reduce the setup and usage barriers. For specialized needs, the platform offers customizable tools and environments.
A key highlight is the platform’s flexibility in GPU allocation. Mr. Huang notes that most market offerings rent GPUs by full card, which can be cost-prohibitive. In contrast, GWS AI Cloud allows rentals starting from 0.25 GPUs, with options scaling to 0.5, 1, 2 GPUs, or more, depending on project requirements. For example, users renting 0.5 GPUs for six months can flexibly adjust usage during the term. If additional computing power is needed, users can scale up at any time.
Expanding MSP Role: Tailored Solutions for Complex Needs
eASPNet also extends its capabilities to design customized security architectures. A recent case involved assisting the Ministry of Digital Affairs in building a dedicated GPU compute pool at eASPNet’s data center. This project incorporated multiple security features, including zero-trust architecture, DDoS protection, and two-factor authentication. eASPNet not only fulfilled all requirements but also ensured the system’s efficiency and stability.
In summary, eASPNet’s deep-rooted expertise in cloud services, combined with the capabilities of GWS AI Cloud, enables users to seamlessly perform AI analytics and modeling while accessing additional services such as data backup, cybersecurity protection, and disaster recovery. Acting as an MSP, eASPNet assists government agencies, academic institutions, and enterprises in designing, planning, validating, testing, deploying, and managing private GPU pools. These solutions can be hosted in eASPNet’s data center or other suitable locations, offering unparalleled flexibility for diverse needs.
