Lablup Inc. today announced that its AI infrastructure platform Backend.AI officially supports the Intel Arc Pro B70 workstation GPU, extending the platform's Intel hardware coverage from the existing Gaudi 2 and Gaudi 3 AI accelerators to Intel Arc Pro graphics.

Backend.AI provides a single control plane and a single model serving stack that spans desktop workstations, shared clusters, and full-scale data centers. With Arc Pro B70 support, customers running Intel AI silicon can now manage their entire fleet, Gaudi in the data center and Arc Pro on the desktop, through one platform.

The Arc Pro B70 is built on Intel's Xe2 (Battlemage) architecture and ships with 32 GB of ECC-enabled GDDR6 memory, 608 GB/s memory bandwidth, and 367 TOPS of INT8 compute. As agentic AI workloads place growing pressure on GPU memory capacity, the B70's 32 GB helps delay the point at which memory constraints degrade performance and allows more agents to run concurrently on a single card. Lablup and Intel jointly benchmarked the Arc Pro B70, validating both its price-to-performance ratio and its usability within the Backend.AI platform.

"The Arc Pro B70's large memory makes it well suited for memory-intensive workloads like agentic AI, even in multi-user environments," said Jeongkyu Shin, CEO of Lablup. "We optimized Backend.AI's GPU resource management and model execution environment to take full advantage of that capacity, so users get a production-quality AI development and inference experience right on their desktop."

A solution brief with detailed benchmark results is available on the [Lablup](https://www.backend.ai/resources/backend-ai-intel-arc-b70) and [Intel](https://community.intel.com/t5/Blogs/Tech-Innovation/Artificial-Intelligence-AI/Lablup-adds-Intel-Arc-Pro-B70-support-to-Backend-AI/post/1751014) websites.

![intel_b70_support_announcement-en.jpg](https://cdn.lablup.com/intel_b70_support_announcement_en_6fdd001290.jpg)

Lablup adds Intel Arc Pro B70 support to Backend.AI

- Lablup contributes AI infrastructure and workload orchestration expertise to the Linux Foundation's open-standards body for agentic AI.

![lablup-joined-aaif.jpg](https://cdn.lablup.com/lablup_joined_aaif_0f1f672f33.jpg)

Lablup Inc., the company behind the AI infrastructure platform Backend.AI, today announced that it has joined the Agentic AI Foundation (AAIF) as a Silver Member.
 
AAIF was co-founded in December 2025 by Anthropic, Block, and OpenAI under the Linux Foundation. The foundation develops and maintains open standards and protocols so that AI agents can interoperate across platforms and vendors. Its inaugural projects include Anthropic's Model Context Protocol (MCP), Block's goose, and OpenAI's AGENTS.md. As of April 2026, the foundation counts more than 170 member organizations, among them AWS, Google, Microsoft, Bloomberg, and Cloudflare.
 
"Lablup's roots are in open source," said Jeongkyu Shin, CEO of Lablup Inc. "We want agentic AI to develop on top of open standards rather than inside closed ecosystems, and we plan to contribute infrastructure technology alongside the AAIF community to help make that happen."
 
Through its membership, Lablup plans to adopt the MCP standard, contribute agent infrastructure technology, and help grow an open agent ecosystem in South Korea. The company brings operational expertise built over years of orchestrating heterogeneous GPU and NPU clusters for enterprises, financial institutions, and research organizations of varying scale. Lablup expects to collaborate with the AAIF community specifically at the infrastructure layer of agentic AI workloads, building on that experience.
 
Lablup has been a Silver Member of both CNCF (Cloud Native Computing Foundation) and the Linux Foundation since 2020, became a sponsor of the Python Software Foundation (PSF) in 2026, and now adds AAIF to its growing list of open-source foundation commitments.
 
## About Lablup Inc.
 
Lablup Inc., founded in 2015, builds Backend.AI, a software-defined AI infrastructure platform that orchestrates heterogeneous GPU and NPU clusters at hyperscale with secure multi-tenancy and multi-node workload management. Its container-level GPU virtualization technology further maximizes accelerator utilization across training, inference, and deployment. Headquartered in Seoul with a U.S. subsidiary in San Jose, Lablup manages over 16,000 GPUs across 110+ sites worldwide, and Backend.AI is validated as NVIDIA DGX-Ready Software. Learn more at [lablup.com](https://www.lablup.com).
 
## About the Agentic AI Foundation (AAIF)
 
The Agentic AI Foundation (AAIF) is a Linux Foundation project established to advance open, interoperable agentic AI technology. Centered on its inaugural projects, MCP, goose, and AGENTS.md, the foundation stewards the core standards and protocols that enable AI agents to interoperate across platforms. Learn more at [aaif.io](https://aaif.io).
 
## Media Contact
 
Wooyoung Yoo, DevRel, Lablup Inc.
 
Email: [wyyoo@lablup.com](mailto:wyyoo@lablup.com)
 
Website: [https://www.lablup.com](https://www.lablup.com)

Lablup Joins the Agentic AI Foundation as Silver Member

![Jeongkyu Shin, CEO of Lablup Inc. (sixth from left), receives the Presidential Commendation at the 2026 Science and ICT Day Ceremony for contributions to the advancement of science, technology, and information & communications.](https://cdn.lablup.com/2026_53d65ab335.jpg)

**The recognition honors Backend.AI for advancing AI technology adoption across Korean industry and for contributing to the creation of skilled engineering jobs.**

Lablup Inc., the developer of Backend.AI, announced that it received the Presidential Commendation at the 2026 ICT Merit Awards, held on April 21 at the Korea Science and Technology Center as part of the Science and Information and Communication Day ceremony organized by the Ministry of Science and ICT.

The ICT Merit Awards are presented each year in conjunction with Science and Information and Communication Day to recognize organizations and individuals that have contributed to the growth of Korea's information and communication industry. Lablup was cited for expanding the development and deployment of AI technology, and for generating quality engineering employment, through its independently developed AI infrastructure operating platform.

Since its founding in 2015, Lablup has built Backend.AI, an AI infrastructure orchestration platform now in use at roughly 120 enterprises and institutions worldwide. Backend.AI manages heterogeneous AI accelerators, including NVIDIA, AMD, Intel, and domestic NPUs, from a single platform. On NVIDIA GPUs, it applies container-level GPU virtualization (partitioning a single GPU so that multiple containers can share it concurrently) to raise utilization. The platform also integrates with leading high-performance storage systems so enterprises and research institutions can run their full infrastructure more efficiently.

This year's Presidential Commendation, awarded to Lablup as a corporate entity, follows the 2024 Industrial Service Medal presented to founder and CEO Jeongkyu Shin. Lablup plans to use the recognition as momentum for broadening the technical base of Korea's AI ecosystem and accelerating its entry into the global AI operations platform market.

"This award belongs to the team that has built Backend.AI together over the past 11 years, and to the customers who trusted the platform and put it to work," said Jeongkyu Shin, CEO of Lablup Inc. "In the end, AI transformation runs on the foundational software that ties models to accelerators. Lablup will keep growing Backend.AI into a platform used across more industries, so that anyone can put AI to work and scale it as freely as they need."

**About Lablup Inc.**

Lablup Inc., founded in 2015, builds Backend.AI, a software-defined AI infrastructure platform that orchestrates heterogeneous GPU and NPU clusters at hyperscale with secure multi-tenancy and multi-node workload management. Its container-level GPU virtualization technology for NVIDIA GPUs further maximizes accelerator utilization across training, inference, and deployment. Headquartered in Seoul with a U.S. subsidiary in San Jose, Lablup manages AI infrastructure across 120+ sites worldwide, and Backend.AI is validated as NVIDIA DGX-Ready Software. Learn more at lablup.com.

Lablup Awarded Presidential Commendation at Korea's 2026 ICT Merit Awards

**\- Operated a 60+ node B200 cluster for 73 days, cutting average recovery time by 47%**

**\- Shares real-world operational strategies from the Solar Open 100B large-model training project**

**\- Live demos of Backend.AI Continuum and Backend.AI:GO on NVIDIA DGX Spark systems**

SEOUL, South Korea – March 16, 2026 — Lablup Inc., a leading AI infrastructure company, will share its experience operating a large-scale Sovereign AI training environment powered by 504 NVIDIA B200 GPUs at NVIDIA GTC 2026, held March 16–19 in San Jose, California. During the GTC Theater Session on March 18, CEO Jeongkyu Shin will present insights from 73 days of large-cluster operations, covering Lablup’s fault-tolerant scheduling strategies and fast recovery techniques. The company will also host hands-on demonstrations at its booth, featuring automatic recovery functions in Backend.AI Continuum and Backend.AI:GO running on NVIDIA DGX Spark systems.

**Sovereign AI in Action: Building a 100B-Parameter Model**

In the session titled “Building Sovereign AI: Scaling 100B+ Model Training on NVIDIA Blackwell Infrastructure,” Shin will walk through how Lablup trained a 100-billion-parameter model on a 60+ node B200 cluster (504 GPUs). This project was part of the Korea Ministry of Science and ICT / NIPA initiative to develop homegrown AI foundation models, with Lablup serving as the infrastructure partner for Upstage’s Solar Open 100B training.

The team built a fault-tolerant scheduling system capable of automatically detecting and recovering from common distributed training failures such as GPU errors and NCCL timeouts. The system reduced the average recovery time by 47% compared to previous runs, restarting failed processes in under three seconds. Lablup will also share its experience troubleshooting issues like a single NFS driver misconfiguration that degraded performance by nearly tenfold, and its optimizations for MXFP8 training stability and NCCL tuning across both RoCE and InfiniBand networks adapted for the Blackwell architecture.

**Hands-On Demos: Backend.AI Continuum and Backend.AI:GO**

Visitors to Lablup’s booth (#243) can take part in live demonstrations showcasing Backend.AI Continuum’s model routing and resilience features. For example, when a network cable is unplugged to simulate a fault, inference requests seamlessly reroute through alternate paths in real time. Even if a cloud connection drops, Continuum maintains API calls by switching instantly to local resources, allowing attendees to physically experience its fault tolerance.

Lablup will also present Backend.AI:GO, a local AI runtime that runs on devices ranging from personal laptops to high-end systems like the NVIDIA DGX Spark with 128GB unified memory. Backend.AI:GO provides a personal, high-performance AI environment that fully utilizes local hardware without depending on remote cloud access.

“Running a 504-GPU B200 cluster continuously for 73 days taught us what truly breaks in large-scale distributed training and how to rebuild it,” said Lablup CEO Jeongkyu Shin. “At GTC 2026, we’re excited to share those lessons and our vision for building sovereign AI infrastructure that enables nations and industries to operate AI on their own terms.”

**About Lablup Inc.**

Lablup Inc., founded in 2015, builds Backend.AI, a software-defined AI infrastructure platform that orchestrates heterogeneous GPU and NPU clusters at hyperscale with secure multi-tenancy and multi-node workload management. Its container-level GPU virtualization technology further maximizes accelerator utilization across training, inference, and deployment. Headquartered in Seoul with a U.S. subsidiary in San Jose, Lablup manages over 16,000 GPUs across 110+ sites worldwide, and Backend.AI is validated as NVIDIA DGX-Ready Software. Learn more at lablup.com.

Lablup to Showcase Sovereign AI Training Infrastructure at NVIDIA GTC 2026 Powered by 500+ B200 GPUs

*   Lablup and UK-based HPC specialist Boston Limited enter a strategic distribution partnership to bring the Backend.AI platform to the EMEA market.
*   Lablup Partners with Boston Limited to Distribute Backend.AI Across EMEA

**SEOUL, SOUTH KOREA & MWC 2026 | BARCELONA, SPAIN – March 2, 2026** – Lablup Inc., the AI infrastructure operating system company, and Boston Limited, the UK-based high-performance computing specialist, today announced a strategic partnership to distribute and sell the Backend.AI platform across the EMEA region.

Under the agreement, Boston Limited will offer Backend.AI cloud services (SaaS) and on-premises software licenses directly to end customers in EMEA, along with installation, configuration, and consulting services. The two companies plan to build local sales and technical support operations through Boston Limited's established customer network and HPC infrastructure distribution capabilities across the region.

Backend.AI is a software-defined AI infrastructure platform that manages heterogeneous accelerators, including chips from NVIDIA, AMD, Intel, and domestic NPU vendors, through a single control plane. Organizations can operate an entire cluster of mixed GPU and NPU hardware as one unified resource pool. Container-level GPU virtualization and the Sokovan orchestrator, both developed in-house, give EMEA customers:

*   A single platform to manage heterogeneous GPU and NPU hardware from multiple vendors
*   Local installation, configuration, and consulting support through Boston Limited's established regional network
*   Integrated operation across existing HPC infrastructure, reducing the cost and operational complexity of deploying AI at enterprise scale

Boston Limited has supplied HPC infrastructure throughout the UK and internationally for over 30 years. The company delivers custom high-performance computing solutions (servers, storage, GPU clusters) to research institutions, universities, enterprises, and public-sector organizations, and maintains close partnerships with Supermicro, NVIDIA, Intel, and AMD.

Lablup established a U.S. subsidiary in San Jose in 2025 to enter the North American market. With this new partnership, Backend.AI's commercial distribution now spans Asia, North America, and the EMEA region.

"Expanding Backend.AI's reach into EMEA required a partner with deep knowledge of the local market," said Jeongmook Kim, CRO at Lablup Inc. "Boston Limited's three decades of HPC infrastructure distribution and broad customer network across the region made them the right fit. Lablup is now working closely with Boston Limited to deliver strong technical support for customers on the ground."

"As AI workloads grow, our customers increasingly need software platforms that let them manage GPU resources efficiently," said Manoj Nayee, Managing Director at Boston Limited. "Adding Backend.AI to the HPC and AI infrastructure we already supply gives EMEA customers a complete solution from hardware to software management. For organizations running GPUs from multiple vendors, this integration can meaningfully improve the return on their infrastructure investment."

![MWC_2026_BOSTON_PARTNERSHIP_VERTICAL.jpg](https://cdn.lablup.com/MWC_2026_BOSTON_PARTNERSHIP_VERTICAL_7377c1942c.jpg)

The partnership was signed on March 2 at MWC 2026, held at Fira Gran Via in Barcelona, Spain.

---

### About Lablup Inc.

Lablup Inc., founded in 2015, builds Backend.AI, a software-defined AI infrastructure platform that orchestrates heterogeneous GPU and NPU clusters at hyperscale with secure multi-tenancy and multi-node workload management. Its container-level GPU virtualization technology further maximizes accelerator utilization across training, inference, and deployment. Headquartered in Seoul with a U.S. subsidiary in San Jose, Lablup manages over 16,000 GPUs across 110+ sites worldwide, and Backend.AI is validated as NVIDIA DGX-Ready Software. Learn more at lablup.com.

### About Boston Limited

Trusted for over 30 years and operating in more than 40 countries, Boston Limited is a global provider of co-designed, custom and certified IT systems powering next-generation intelligent infrastructure. Its portfolio spans edge devices, desktops, servers, storage, clusters and reference architectures, delivered through an engineering-led, customer-centric engagement model and available via boston.co.uk.

Over three decades, Boston has continually evolved to meet the changing demands of the technology landscape - from component distribution to value-added system integration, and today as a leader in data, high-performance computing (HPC) and AI solutions. Boston Limited is part of US-based SourceCode, LLC.

Lablup Partners with Boston Limited for EMEA Distribution of Backend.AI

_**Backend.AI integrates with VAST AI Operating System to deliver high-performance data access for AI workload orchestration across training, deployment, and inference**_

**SEOUL, SOUTH KOREA & VAST Forward | SALT LAKE CITY, UT – February 25, 2026** – Lablup Inc., the AI infrastructure operating system company, today announced it has joined VAST Cosmos as a Technology Alliance Partner, collaborating with VAST Data, the AI Operating System company, to help customers accelerate AI training and inference workloads through a validated integration between Backend.AI and the VAST AI Operating System.

As enterprises race to operationalize AI, proven partner integrations have become one of the fastest ways to deliver validated architectures, reduce deployment risk, and accelerate time-to-value. Through Cosmos, VAST and Lablup are able to build, validate, deliver, and scale AI-ready solutions across the data center, cloud, and at the edge.

"Lablup is committed to democratizing AI infrastructure through accessible, efficient, and scalable technology," said Jeongkyu Shin, CEO at Lablup Inc. "By integrating Backend.AI with the VAST AI OS, developers and enterprises can now handle large-scale datasets and complex workloads while maintaining the reliability and high performance that modern AI demands. Storage infrastructure matters just as much as the orchestrator managing AI workloads, and Lablup is dedicated to delivering a unified environment that supports both seamlessly."

With this collaboration, Lablup integrates Backend.AI with the VAST AI OS to enable customers to orchestrate end-to-end AI workloads, from training to deployment and inference, with high-throughput access to training datasets and model artifacts through RDMA and GPU-direct data paths powered by VAST DataStore. This integration streamlines data access, reduces latency, and maximizes GPU utilization across heterogeneous compute environments.

By delivering a validated solution combining Backend.AI's GPU orchestration and workload management capabilities with VAST's Disaggregated Shared-Everything (DASE) architecture and unified data services, organizations can achieve:

\- Faster deployment and simplified integration for AI pipelines through validated architectures that reduce implementation complexity

\- Improved performance and predictable outcomes under sustained workloads with high-throughput data access and GPU-direct connectivity

\- Consolidated infrastructure management that reduces operational overhead while scaling AI training, inference, and deployment environments

"Enterprises are looking for a simpler, more consistent way to deploy production AI, backed by validated integrations they can trust," said John Mao, Vice President, Global Technology Alliances at VAST Data. "We're proud to welcome Lablup into VAST Cosmos, expanding the ecosystem of industry-leading solutions built on the VAST AI Operating System and helping customers reduce integration risk while accelerating AI-ready deployments from the data path to the application layer."

Lablup and VAST Data are collaborating on infrastructure for South Korea's Sovereign AI foundation model initiatives, and this Cosmos partnership further strengthens their joint commitment to delivering production-ready AI infrastructure solutions.

**About Lablup**

Lablup Inc., founded in 2015, builds Backend.AI, a software-defined AI infrastructure platform that orchestrates heterogeneous GPU and NPU clusters at hyperscale with secure multi-tenancy and multi-node workload management. Its container-level GPU virtualization technology further maximizes accelerator utilization across training, inference, and deployment. Headquartered in Seoul with a U.S. subsidiary in San Jose, Lablup manages over 16,000 GPUs across 110+ sites worldwide, and Backend.AI is validated as NVIDIA DGX-Ready Software. Learn more at lablup.com.

Lablup Joins VAST Cosmos to Accelerate Integrated AI Training and Inference

**The free desktop application runs small language models entirely offline on personal PCs and supports multi-device distributed inference through Mesh Mode. Lablup will also sign a partnership with UK-based AI infrastructure firm Boston Limited during the event.**

**SEOUL —** Lablup Inc. today announced it will showcase Backend.AI:GO, its free personal AI platform, at MWC Barcelona 2026, held March 2 through 5 at Fira Gran Via in Barcelona, Spain. The company will also hold a partnership signing ceremony with Boston Limited, a UK-based AI infrastructure company, as it formally enters the European market.

Backend.AI:GO, launched at CES 2026 in January, is a free desktop application that runs small language models (SLMs) directly on a personal PC with no cloud connection required. Once a model is downloaded, the application operates entirely offline, handling document analysis, image understanding, and code review without transmitting data externally. It supports macOS, Windows, and Linux, with native Apple Silicon optimization through Lablup's own MLxcel inference engine.

When a task exceeds local hardware capacity, Backend.AI:GO can automatically offload it to a Backend.AI Core cluster within the organization or to cloud APIs from providers such as OpenAI and Google. Users can also link multiple GO devices together through Mesh Mode, pooling computing resources for distributed inference on models too large for a single workstation.

Since launch, Backend.AI:GO has added capabilities at a rapid pace. Version 1.1, released in early February, introduced Stable Diffusion-based image generation, document translation with OCR support for TXT, MD, DOCX, PDF, and image files, and a headless server mode for standalone deployment in server environments. Version 1.2 adds an agent marketplace where users can browse and download agent profiles, along with a Cowork mode that lets multiple agents collaborate under shared folder permissions and automated approval rules.

**From Desktop to Mobile**

Lablup is developing the next phase of Backend.AI:GO: the ability to remotely access and control AI models running on a desktop, laptop, server, or NVIDIA DGX Spark from a mobile device. Any machine running GO becomes a personal AI server that a smartphone can connect to for conversations or task requests. The orchestration capabilities that Lablup has built for enterprise AI infrastructure now extend to individual users and their personal devices.

"Backend.AI:GO was the first step in bringing AI workloads that once required a data center onto personal devices, and now Lablup is extending that reach to mobile," said Jeongkyu Shin, CEO of Lablup Inc. "There is no reason anyone should be tied to a specific location or a specific machine to use AI. That has been the direction Backend.AI has pursued from the beginning."

Lablup will be exhibiting at MWC Barcelona 2026 in Hall 7, Booth #7A62-01.

**About Lablup Inc.**

Lablup Inc., founded in 2015, builds Backend.AI, a software-defined AI infrastructure platform that orchestrates heterogeneous GPU and NPU clusters at hyperscale with secure multi-tenancy and multi-node workload management. Its container-level GPU virtualization technology further maximizes accelerator utilization across training, inference, and deployment. Headquartered in Seoul with a U.S. subsidiary in San Jose, Lablup manages over 16,000 GPUs across 110+ sites worldwide, and Backend.AI is validated as NVIDIA DGX-Ready Software. Learn more at lablup.com.

Lablup Brings Personal AI Platform Backend.AI:GO to Europe at MWC Barcelona 2026

_\- Enterprise AI infrastructure now extends to personal PCs while maintaining centralized management_  
_\- Cloud version Backend.AI:DOL also launches, completing full-stack coverage from AI PCs to hyperscale clusters_

![(Lablup) CES 2026.jpg](https://cdn.lablup.com/Lablup_CES_2026_491d15e4a7.jpg)

**LAS VEGAS, January 6, 2026** – Lablup, an AI infrastructure platform company, today announced the official launch of Backend.AI:GO at CES 2026. The desktop application enables users to run Small Language Models (SLMs) directly on personal computers without cloud servers. This marks Lablup's fourth consecutive year exhibiting at CES.

Backend.AI:GO (where "GO" stands for "Generative On-device") accelerates AI inference using local computing resources in on-premises environments. Once users download a model, the application operates entirely offline without requiring an internet connection. User data is never transmitted to external servers. Users can securely download models to their personal computers and freely utilize them for internal document analysis, image understanding and generation, code review, and more. The application also allows users to adjust hyperparameters and resource allocation to customize model response characteristics and resource usage.

The key differentiator of Backend.AI:GO lies in its integration with Lablup's AI infrastructure management platform, Backend.AI. By connecting multiple Backend.AI:GO instances or integrating with existing Backend.AI environments and cloud AI services such as OpenAI and Google, users can leverage both large-scale AI models and local PC resources in a unified manner. Organizations can extend the same management approach used for thousands of GPU clusters to AI PCs, enabling centralized model deployment, version control, and usage monitoring from a single console.

Lablup will also officially launch Backend.AI:DOL at CES 2026. Previously introduced as a beta version at AI Infra Summit, Backend.AI:DOL (short for "Deployable Omnimedia Lab") is a cloud-based solution that allows users to access various open models directly through a web browser. With Backend.AI:GO handling on-device workloads and Backend.AI:DOL managing cloud deployments, Lablup now offers a complete full-stack AI infrastructure solution spanning from AI PCs to hyperscale GPU clusters.  
"As the AI PC era accelerates, enterprise demand for running LLMs directly in on-device environments is growing rapidly," said Jungkyu Shin, CEO of Lablup. "With Backend.AI:GO and DOL, we will provide an environment where organizations can operate AI infrastructure with a consistent experience, regardless of scale, from a single PC to thousands of GPU clusters."

Shin added, "Lablup has expanded its presence in the global AI infrastructure market through four consecutive years of CES participation. At this year's CES, we are introducing a full-stack AI infrastructure solution that spans on-device and cloud, setting a new standard for AI infrastructure operations."

Lablup is exhibiting at CES 2026 in LVCC North Hall, Booth #9529, from January 6–9, 2026.

**About Lablup**  
Lablup Inc. is redefining how modern AI systems are powered by delivering the operating model for scalable, high-performance GPU computing. Its flagship platform, Backend.AI, combines software-defined GPU virtualization, ultra-fast containerized environments, and intelligent workload orchestration to maximize performance, utilization, and efficiency across heterogeneous clusters.

Founded in 2015, Lablup works with global technology leaders, research institutions, and top universities to advance AI infrastructure and accelerate data-driven innovation worldwide. Learn more at lablup.com and follow Lablup on LinkedIn.

Lablup Unveils Backend.AI:GO at CES 2026: Run LLMs on a Personal Computer Without the Cloud

**SAN FRANCISCO, CA** – **November 14, 2025** – Lablup Inc., a leader in AI infrastructure solutions, continues to revolutionize the AI education and research landscape through its Backend.AI platform. At SC25 in St. Louis, Missouri, Lablup will showcase significant case studies demonstrating the transformative impact of AI infrastructure across diverse fields including higher education, national research projects, and international trade.

![SC25_LABLUP_BOOTH_3102.jpg](https://cdn.lablup.com/SC_25_LABLUP_BOOTH_3102_87eaf8008c.jpg)

### Pioneering AI Solutions Across Organizations

Lablup’s commitment to making AI accessible and manageable is evident through successful implementations in various high-profile projects:

- **Korea's National AI Initiative**: In collaboration with Upstage, Lablup contributed to the 'Sovereign AI Foundation Model Project,' enabling large-scale language model development and strengthening Korea’s AI capabilities through resilient GPU infrastructure.
- **University of Southern California (USC)**: By utilizing Backend.AI’s GPU virtualization, USC supported over 80 students simultaneously in machine learning exercises, significantly enhancing AI education without the need for extensive technical staff.
- **teamreboott**: Specializing in trade automation, teamreboott fine-tuned large language models using Backend.AI FastTrack 3, streamlining trade operations with reduced human oversight.
- **Korea Institute of Science and Technology (KIST)**: KIST accelerated innovative Cas9 research using Backend.AI Cloud for dynamic molecular simulations, demonstrating the platform’s ability to support complex scientific endeavors.
- Kookmin University (KMU): KMU efficiently scaled its AI education program to accommodate hundreds of students using Backend.AI to virtualize GPU resources, demonstrating significant flexibility and broad access to technology.
- Sungkyunkwan University (SKKU): As Korea's first university to manage an independent supercomputing center, SKKU uses Backend.AI to centralize its HPC resources, optimizing research productivity and operational efficiency.

### Enhancements for Scalability and Accessibility
"Backend.AI’s self-repairing infrastructure ensures our large-scale projects remain resilient and uninterrupted,” said Jeongkyu Shin, CEO of Lablup. “Our partnerships across diverse sectors underscore our mission to eliminate AI management complexities, making advanced computing accessible to all."

### Interactive Experience at SC25
At booth #3102, Lablup will offer live demonstrations of Backend.AI running on Intel Gaudi 3 AI accelerators. Attendees can participate in interactive sessions, gain hands-on experience, and earn incentives through involvement in the Innovation Pass Program with Intel.

### About Lablup
Lablup Inc. is redefining how modern AI systems are powered by delivering the operating model for scalable, high-performance GPU computing. Its flagship platform, Backend.AI, combines software-defined GPU virtualization, ultra-fast containerized environments, and intelligent workload orchestration to maximize performance, utilization, and efficiency across heterogeneous clusters.
Founded in 2015, Lablup works with global technology leaders, research institutions, and top universities to advance AI infrastructure and accelerate data-driven innovation worldwide. Learn more at [lablup.com](https://www.lablup.com) and follow Lablup on [LinkedIn](https://www.linkedin.com/company/lablup).









Lablup to Participate in SC25, the World's Largest HPC Conference

- Lablup and VAST Data Power Korea’s Sovereign AI Initiative
- Collaboration resolves performance bottlenecks and enables large-scale model training across 500+ GPUs, advancing Korea’s national AI infrastructure.

Lablup Inc., the developer of the open, scalable AI computing platform Backend.AI and a key participant in South Korea’s government-backed Sovereign AI Project today announced a strategic collaboration with VAST Data, the AI Operating System company. Together, the companies have built a high-performance, scalable data foundation that enables Korea’s national AI consortium, which aims for technological self-reliance, to train and deploy large-language models entirely on domestic infrastructure.

Backend.AI provides a unified environment for AI model training, deployment, and inference across GPU and CPU resources. By integrating the VAST AI Operating System, Lablup and its customers can manage training data, model checkpoints, and artifacts with unmatched throughput and resilience, ensuring smooth data flows from model development to real-time serving.

As Lablup’s AI workloads grew, traditional infrastructure storage systems struggled to keep pace with the throughput and concurrency required for foundation-model training. The team turned to VAST Data to modernize its data layer, eliminating performance bottlenecks while maintaining reliability and control for multi-tenant research environments.

Running on SK Telecom’s Sovereign AI cluster, the integrated Backend.AI and VAST AI Operating System now orchestrates and accelerates model training across more than 500 NVIDIA GPUs, giving Korean researchers a unified environment for both training and, in the future, inference.

“Our mission has always been to democratize AI infrastructure through open, efficient, and scalable technology,” said Jeongkyu Shin, Founder and CEO of Lablup Inc. “By integrating VAST’s AI Operating System into Backend.AI environments, we’re enabling developers and enterprises to handle massive datasets and complex workloads with the reliability and performance that modern AI demands. It’s a technological synergy that really unlocks the full potential of our AI compute while ensuring sovereignty and data security.”

Through direct integration with the VAST AI Operating System, Backend.AI achieves seamless, high-throughput access to data for massive training workloads.Using VAST’s Disaggregated, Shared-Everything (DASE) architecture, Lablup can scale compute and storage independently while maintaining a single global namespace, VAST DataSpace, enabling researchers to access checkpoints, datasets, and model outputs instantly across projects.

Powering Lablup’s role in Korea’s Sovereign AI initiative, the Backend.AI and VAST AI Operating System deliver everything from large-scale model training to real-time orchestration, including:
- Sovereign Data Control: AI sovereignty begins and ends with data infrastructure. Built on VAST DataSpace and the AI Operating System’s Zero Trust architecture, Lablup enforces data residency, access control, and compliance across Korea’s Sovereign AI environments. For government, research, and regulated enterprise participants in the national consortium, this ensures not just where data lives, but how it’s protected, governed, and audited throughout the AI lifecycle.
- High-Performance Data Infrastructure: Backend.AI integrates with the VAST DataStore, delivering high-throughput access to training datasets and model artifacts via NVIDIA GPUDirect Storage (GDS) and RDMA. This direct GPU-to-data pipeline removes I/O bottlenecks and ensures consistent performance across more than 500 NVIDIA GPUs on SK Telecom’s Sovereign AI cluster, accelerating model pre-training and fine-tuning at national scale.
- Unified Global Namespace for Seamless Collaboration: Backend.AI, built on VAST’s DASE architecture, provides researchers and developers with a single, unified namespace that spans every workload; from model training and validation to deployment. With VAST DataSpace, teams can instantly access checkpoints, datasets, and outputs across distributed environments without duplication or downtime, simplifying collaboration across the consortium.
- Secure, Multi-Tenant Scalability: Lablup leverages VAST DataEngine and DataSpace to unify data and compute management across shared GPU clusters. Tight integration between VAST’s quota model and Backend.AI’s user-based access controls ensures secure isolation between tenants while maintaining predictable performance for every user. This architecture allows Lablup to scale compute and storage independently while providing consistent service to government, academic, and enterprise researchers alike.
- Future-Ready for Inference and Agentic AI: As the Sovereign AI initiative transitions from development to deployment, Backend.AI’s integration with the VAST AI Operating System provides a foundation for inference and future agentic AI workloads. With VAST’s DataEngine enabling real-time orchestration and reasoning across massive datasets, Lablup’s customers can move beyond static model training toward dynamic, intelligent systems that continuously learn, adapt, and serve.

Within Lablup’s own infrastructure, the VAST AI Operating System supports AI infrastructure development, optimization testing, and model service orchestration. This integration allows engineers to test and deploy new AI models more efficiently, switching model weights in real time without downtime. Through joint engineering efforts, VAST and Lablup resolved early latency challenges, resulting in a stable, high-throughput system that scales seamlessly for both training and research operations.

Within the Korean Foundation Model Consortium, led by Upstage and supported by other national AI research partners, Backend.AI powered by the VAST AI Operating System processes massive-scale corpora for large-model pre-training and fine-tuning, advancing the development of frontier-grade, sovereign AI capabilities for Korea.

“Our work with Lablup shows what modern AI infrastructure actually looks like when it’s built for scale,” said Alon Horev, Co-Founder and CTO, VAST Data. “Legacy stacks that were never designed for GPU-dense or sovereign environments simply can’t keep up. By bringing compute orchestration and data intelligence together inside the VAST AI Operating System, we’ve created a single fabric that removes those bottlenecks and gives nations like Korea the ability to own, control, and accelerate their own AI innovation. It lets teams like Lablup train bigger, move faster, and stay in control of their data. With Backend.AI and VAST Data together, enterprises and researchers are establishing a new blueprint for high-performance AI infrastructure; one that fuses flexibility, efficiency, and trust for the agentic computing era.”

Led by the Ministry of Science and ICT, Korea’s Sovereign AI initiative aims to develop globally competitive foundation models built entirely with domestic technology, reducing dependence on foreign cloud and software providers. Together, VAST and Lablup provide the data and compute fabric that makes this possible, empowering start-ups, enterprises, and research institutions to train and scale AI securely within national borders. This collaboration reflects the growing technical synergy between Backend.AI’s distributed container orchestration and the VAST AI Operating System, a unified foundation for the agentic AI era.

**About Lablup**
Lablup Inc. is redefining how modern AI systems are powered by delivering the operating model for scalable, high-performance GPU computing. Its flagship platform, Backend.AI, combines software-defined GPU virtualization, ultra-fast containerized environments, and intelligent workload orchestration to maximize performance, utilization, and efficiency across heterogeneous clusters.

Founded in 2015, Lablup works with global technology leaders, research institutions, and top universities to advance AI infrastructure and accelerate data-driven innovation worldwide. Learn more at lablup.com and follow Lablup on [LinkedIn](www.linkedin.com/company/lablup).

**About VAST Data **
VAST Data is the AI Operating System company – powering the next generation of intelligent systems with a unified software infrastructure stack that was purpose-built to unlock the full potential of AI. The VAST AI OS consolidates foundational data and compute services and agentic execution into one scalable platform, enabling organizations to deploy and facilitate communication between AI agents, reason over real-time data, and automate complex workflows at global scale. Built on VAST’s breakthrough DASE architecture – the world’s first true parallel distributed system architecture that eliminates tradeoffs between performance, scale, simplicity, and resilience – VAST has transformed its modern infrastructure into a global fabric for reasoning AI. Learn more at vastdata.com and follow VAST Data on [LinkedIn](https://www.linkedin.com/company/vast-data/), [YouTube](https://www.youtube.com/vastdata) and [X](https://x.com/VAST_Data).

Lablup — Make AI infrastructure accessible

Making AI possible
where it wasn’t

Trusted by leading organizations

Latest news

Lablup adds Intel Arc Pro B70 support to Backend.AI

Better, together

Catch up on the latest

Need a hand with your AI infrastructure?

Lablup — Make AI infrastructure accessible

Making AI accessiblepossiblepossiblewhere it wasn’t

Trusted by leading organizations

Latest news

Lablup adds Intel Arc Pro B70 support to Backend.AI

Better, together

Catch up on the latest

Need a hand with your AI infrastructure?

Making AI possible
where it wasn’t