Lablup at GTC 2026

Get sovereign AI configs ready from sovereign successes with Backend.AI.

Get sovereign AI configs ready from sovereign successes with Backend.AI.

Booth #243

Backend.AI has powered Korea's national sovereign AI program — scaling a 102B-parameter model across 504 NVIDIA B200 GPUs with 47% faster failure recovery. 

From air-gapped government clusters to financial-grade deployment, these are configs that have already passed the hardest tests.

Meet Lablup, a proud silver sponsor of NVIDIA GTC 2026.

Visit us and get Lablup merch

Meet Backend.AI Keyboard switch keychain at scale.

Complete a simple mission to unlock our exclusive merch.

More details coming soon on our social media channels.

Keyboard switch keychain image

Meet Backend.AI:GO

A desktop AI platform that runs models locally on your PC - or connects to Cloud APIs seamlessly.

A desktop AI platform that runs models locally on your PC - or connects to Cloud APIs seamlessly.

Backend.AI:GO, short for "Generative On-device," is an on-device AI solution that leverages Backend.AI infrastructure across diverse local hardware for local inference, personal AI agent management, and P2P resource clustering.

Backend.AI - AI OS for your AI Infrastructure

AI Infrastructure operating system that transforms your GPU complexity into operational simplicity

Certified as NVIDIA DGX™-Ready Software, Backend.AI is optimized to deliver the full performance of DGX systems.

AI Infrastructure operating system that transforms your GPU complexity into operational simplicity

Explore our resources

Learn more with our use cases, soultion briefs, and whitepapers. 

Various preview image of Lablup's papers

Lablup featured sessions

Building Sovereign AI: Scaling 100B+ Model Training on NVIDIA Blackwell Infrastructure

EX82047

Monday, March 18  |  16:20 -16:35 PT

Explore real-world challenges training 100B+ parameter models from scratch on NVIDIA Blackwell GPUs. From a national AI initiative on 500+ B200 GPUs, we overcame GPU errors, NCCL timeouts, and storage bottlenecks. Built resilient stack with auto-recovery, fixed 10x NFS slowdowns, stabilized MXFP8, and optimized RoCE/InfiniBand for reliable large-scale runs.

Sessions with Partners

Expanding the Power of Vultr Cloud GPUs with Backend.AI

Vultr, Booth #1631

Tuesday, March 17  |  15:30 -15:45 PT 

Discover how Backend.AI maximizes the efficiency of Vultr's infrastructure powered by the NVIDIA Blackwell architecture. This presentation demonstrates intelligent GPU resource management and advanced orchestration, enabling organizations to scale AI workloads efficiently across environments. Learn practical strategies for optimizing GPU utilization and deployment on Vultr's state-of-the-art platform.

See you at GTC 2026!

Our team can solve your business challenges.

Schedule a meeting and find solutions to your infrastructure problems.

We're here for you!

Complete the form and we'll be in touch soon

Contact Us

Headquarter & HPC Lab

KR Office: 8F, 577, Seolleung-ro, Gangnam-gu, Seoul, Republic of Korea US Office: 3003 N First st, Suite 221, San Jose, CA 95134

© Lablup Inc. All rights reserved.

We value your privacy

We use cookies to enhance your browsing experience, analyze site traffic, and understand where our visitors are coming from. By clicking "Accept All", you consent to our use of cookies. Learn more