InnoCloud Tech Unveils EBFlex Private Computing Platform at CCIG 2026 to Optimize university research computing
From May 29 to 31, the 2026 China Conference on Image and Graphics (CCIG 2026) was held in Guangzhou. Hosted by the Chinese Society of Image and Graphics and organized by Sun Yat-sen UniveRSIty, the event was co-organized by the Guangdong Society of Image and Graphics, South China University of Technology, and the Pazhou Laboratory. Over 4,200 experts, scholars, and corporate representatives from academia and industry gathered to exchange insights on cutting-edge fields, including image and graphics, Artificial Intelligence, multimodal generation, and embodied intelligence.
During the conference, Innocloud Tech (Yingbo Shuke) exhibited with its InnoCloud brand, focusing on showcASIng its computing power service capabilities tAIlored for universities, research institutions, and AI innovation teams. On May 29, Qin Weijun, Vice President of InnoCloud Tech, delivered a keynote speech titled "exploration and PrACTice of InnoCloud in Empowering Scientific Research" at the "Frontier technology Forum on Multimodal Generation and Embodied APPlications." During the forum, he officially unveiled the EBFlex private computing management platform for the first time. By combining EBCloud's public cloud Intelligent computing services with the EBFlex private platform, InnoCloud Tech aims to provide university research scenarios with a comprehensive computing service system that covers "local computing governance plus elastic cloud supplementation."
Exhibiting at CCIG 2026: Focusing on University Research Computing Scenarios
CCIG 2026 focused on frontier directions such as image and graphics, AI, multimodal generation, and Embodied Intelligence, attracting extensive participation and exchange from universities, research institutions, industrial enterprises, and technology ecosystem partners.
During the event, InnoCloud Tech highlighted its InnoCloud product capabilities in its exhibition booth, covering scenarios such as scientific computing, large model training and inference, multimodal model R&D, and local GPU resource management. During on-site interactions, university research users expressed strong interest in GPU resource utilization efficiency, unified management of local computing power, rapid deployment of research environments, and elastic supplementation during peak computing demand.
Qin Weijun stated that participating in CCIG 2026 was not only a product showcase but also a vital communication channel with university research users. Through on-site exchanges, the company further gathered practical issues regarding computing supply, platform usage, and resource management from universities and research teams, providing valuable references for future product optimization and industry-academia-research collaborations.
AI for science Drives Growth in Research Computing Demand
In recent years, the integration of AI with scientific innovation, Education, and industrial applications has deepened. AI computing power is transitioning from a single technical resource to a critical infrastructure supporting scientific innovation, talent cultivation, and industrial development.
From a policy perspective, "AI+" initiatives are promoting the deep integration of AI with scientific R&D, education, and industrial applications. Concurrently, the continuous advancement of computing infrastructure is making intelligent computing supply, scheduling, network-computing coordination, and resource utilization efficiency key components of new infrastructure construction.
At the application level, AI for Science is accelerating. Universities and research institutions have a growing demand for AI computing power. Fields such as materials science, drug discovery, life sciences, computational chemistry, and Earth sciences increasingly rely on large-scale model training, simulation computing, and high-throughput inference. AI is no longer just an aUXiliary research tool but is gradually becoming part of the core scientific research infrastructure.
However, the computing needs of universities differ significantly from enterprise scenarios. Universities have numerous research projects with dispersed demands, unstable project cycles, and fRAGmented funding sources. Furthermore,阶段性 (periodic) peaks in computing demand occur during concentrated course teaching, academic competitions, and project closures. Merely increasing GPU procurement cannot fully resolve issues related to computing utilization rates, usage thresholds, and management efficiency.
This signifies that university research computing construction is shifting from simply "purchasing equipment" to "building platforms, manAGIng resources, and improving efficiency." How to uniformly manage dispersed GPU resources and obtain flexible cloud computing supplementation during peak periods has become a practical challenge in building university AI research infrastructure.
Three Major Pain Points in University Research Computing: Supply, Usage, and Management
In his speech, Qin Weijun highlighted three primary pain points currently facing university research computing:
Difficulty in Computing Supply: Common sources of computing power for university research teams include self-procurement by laboratories, university-built infrastructure, and third-party leasing. In practice, these often face resource shortages, delayed equipment upgrades, and an inability to meet periodic peak demands. Especially during intensive courses, competition training, or project closures, computing demand is often released in a highly concentrated short period.
Difficulty in Computing Usage: Many researchers do not have a computer science background. When using GPU resources, they must handle configurations for drivers, CUDA, Frameworks, and dependency environments. Time-consuming environment setup, inconsistent toolchains, and fragmented training and inference workflows all hinder research efficiency.
Difficulty in Computing Management: Local GPU resources in universities are often scattered across different laboratories, research groups, and colleges. There is a lack of unified scheduling, permission management, usage statistics, and cost accounting mechanisms. Clear platform-based management tools are needed to determine whether resources are fully utilized, how to allocate quotas among different teams, and whether funding consumption is reasonable.
Based on these issues, InnoCloud Tech believes that university research computing construction is transitioning from a "resource procurement model" to a "platform service model." Universities need more than just additional GPU devices; they require a comprehensive computing service system that connects resource supply, research usage, and daily management.
EBFlex Unveiled: Transforming Dispersed GPUs into Unified Computing Services
The EBFlex platform, officially unveiled during CCIG 2026, is a private computing management platform within the InnoCloud product matrix, primarily targeting the management of local GPU resources in universities.
According to InnoCloud Tech, EBFlex supports integration starting from a single server and can gradually scale to shared resource pools at the research group, laboratory, college, or even university-wide level. Through resource pooling, unified scheduling, and flexible allocation, GPU devices originally scattered across different teams can be transformed into unified computing resources serving specific research tasks.
On the user side, EBFlex provides pre-configured AI Development environments, team spaces, project management, permission configuration, and resource quotas. This helps students and faculty reduce底层 (underlying) environment configuration work, allowing them to enter model training, experimental validation, and R&D processes faster.
On the management side, EBFlex supports resource monitoring, usage statistics, member quotas, and billing displays. This helps administrators grasp GPU resource utilization and facilitates resource allocation, cost accounting, and project auditing for colleges, laboratories, and research groups.
Qin Weijun noted that the core value of EBFlex is reflected in four aspects: improving GPU utilization to reduce idle resources; supporting permission and quota management to adapt to the needs of different users, research groups, and projects; lowering O&M thresholds to help administrators uniformly manage local computing resources; and accelerating research ouTPUt by allowing researchers to dedicate more time to experiments and innovation.
Synergy Between EBCloud and EBFlex: Forming a "Local + Cloud" Computing Service System
In addition to EBFlex, InnoCloud Tech introduced its EBCloud public cloud intelligent computing service. Built on a Kubernetes-native architecture, EBCloud provides research teams with ready-to-use GPU computing power, AI development environments, and model services. It also supports elastic distributed training, high-performance networking, multi-cluster resource scheduling, and MLOps toolchains.
EBCloud primarily addresses the issue of elastic cloud supplementation. When universities face peak periods such as course training, academic competitions, large model training, or periodic research tasks, they can rapidly scale up via Cloud Computing to alleviate the resource contradiction of "idle during normal times, scarce during busy times."
Thus, EBFlex and EBCloud have a clear division of labor: EBFlex handles the unified governance of local GPU resources, while EBCloud manages elastic cloud computing supplementation. Combined, InnoCloud aims to help university research computing achieve "adequate supply, excellent usability, and clear management."
On the supply side, local unified management and cloud supplementation enhance resource elasticity. On the usage side, Standardized environments and consistent experiences lower the usage threshold. On the management side, unified platform capabilities make resource scheduling, permissions, quotas, and cost accounting much clearer.
Validation Across Multiple Universities: Promoting the Platformization of Research Computing
During his speech, Qin Weijun shared practical cases of InnoCloud in university research scenarios. InnoCloud Tech has collaborated with numerous domestic universities and research institutions, covering scientific research projects, course teaching, and academic competitions.
For example, the Gaoling School of artificial intelligence at Renmin University of China utilizes InnoCloud's public cloud services to support research projects in Video Generation, large language model training, and social simulation. Shanghai University of Finance and Economics has adopted a college-level cloud management model to serve interdisciplinary research in computational economics and financial modeling.
In local computing management and private deployment scenarios, InnoCloud Tech has built an intelligent computing cluster Operation platform for relevant teams at Tsinghua University and deployed computing management systems for multiple laboratories at the University of Science and Technology of China (USTC). These cases dEMOnstrate that university research computing is gradually moving from single-point device usage to unified platform management and service-oriented operations.
Public Information shows that InnoCloud Tech is a wholly-owned subsidiary of the listed company Hongbo Co., Ltd. The company has long focused on GPU intelligent computing center construction, AI computing services, and intelligent cloud platforms. Currently, the InnoCloud product service matrix covers public cloud intelligent computing services, private computing management platforms, and AIDC construction and O&M services, providing computing resources, platform management, deployment delivery, and continuous O&M support for universities, research institutions, and enterprise AI teams.
Continued Open Collaboration for the University Research Ecosystem
The official unveiling of EBFlex at CCIG 2026 signifies that InnoCloud Tech is extending its InnoCloud service capabilities from public cloud elastic computing further into the university local computing governance scenario. In the future, InnoCloud Tech will continue to promote the coordinated development of EBCloud and EBFlex, helping universities and research institutions improve computing resource utilization efficiency and reduce usage and management costs.
Concluding his speech, Qin Weijun stated that InnoCloud's goal is to become a "computing partner that truly understands scientific research." Facing frontier research directions such as AI for Science, multimodal generation, and embodied intelligence, InnoCloud Tech will continue to strengthen cooperation with universities, research institutions, and industry partners to promote the deeper integration of computing resources, platform capabilities, and research scenarios.
During CCIG 2026, InnoCloud Tech also launched trial support for InnoCloud products targeted at university research users, including a free single-machine trial version of EBFlex. InnoCloud Tech welcomed university faculty, students, research teams, and industry partners to further communicate and experience the platform, jointly exploring new pathways for AI Research computing services.
Comments & Questions (0)
No comments yet
Be the first to comment!