LABLUP
Seoul, KR · 2 known investors
Find your way into LABLUP
115 people in our graph share verified history with the LABLUP team — schools, employers, funds. One of them is your warm intro.
Lablup builds software that manages AI infrastructure, absorbing the complexity of running GPU clusters so teams can build and serve AI regardless of scale. It serves researchers, companies, and institutions, from small teams with a single GPU to large clusters, and offers GPU-as-a-Service and related infrastructure solutions.
Also known as Backend.AI · Lablup Inc.
Founders & leadership
Investors · 2
Company profile
researched Aug 2026Lablup Inc. develops software for managing AI infrastructure, positioning its flagship product Backend.AI as an "operating system for AI infrastructure" that spans the workflow from data preparation to model serving. The platform provides unified management of heterogeneous accelerators, including CUDA GPUs, ROCm GPUs, Gaudi NPUs, Google TPUs, GraphCore IPUs and other NPUs, and supports both AI training and inference on a single platform. Alongside the core Backend.AI cluster platform, the company offers Backend.AI FastTrack 3 for pipeline and workflow orchestration, and AI:GO, described as a desktop AI product.
The product line is built around a container-based computing cluster platform whose components are published as open source, including the Backend.AI core (Python), the Backend.AI Web UI desktop and web client (TypeScript), container recipes for Backend.AI kernels, and supporting libraries such as Callosum (an asyncio RPC transport library), raftify (a Raft framework in Rust) and an etcd client Python binding. More recent open-source work includes mlxcel, an LLM/VLM inference runtime and server for Apple Silicon and NVIDIA CUDA devices, bssh, a parallel SSH execution tool for cluster management, and all-smi, a command-line GPU monitoring utility.
The company is organized into research divisions (GenAI Lab, Model & Data Lab, Next-gen Accelerated Computing Lab), DevOps and engineering groups (cloud native engineering, frontend, core development), an AI platform and consulting group providing deployment consulting and technical support, sales and marketing including developer relations, product, and operations functions.
Business model
Lablup sells and supports its Backend.AI AI infrastructure platform to organizations, with subscription licensing described as the basis for recurring revenue, complemented by deployment consulting and technical support delivered by a dedicated AI platform and consulting team. Core platform components are released as open source under LGPL-3.0 and Apache-2.0 licenses.
Subscription licensing of the Backend.AI platform, described by the company as driving sustainable growth, along with consulting and technical support services for customer deployments.
Traction
Lablup states it operates GPU clusters of more than 2,000 GPUs at a single site and serves leading organizations. Its open-source Backend.AI repository has 672 GitHub stars and 181 forks, the Web UI repository 133 stars, and the mlxcel inference runtime 383 stars, across 151 public repositories.
Latest developments
In July 2026 Lablup received KOSDAQ preliminary listing approval, and in August 2026 filed a securities registration statement to advance the IPO. In August 2026 the Upstage consortium, in which Lablup is the infrastructure partner, advanced to phase three of Korea's sovereign AI foundation model project, and the company announced a presentation on confidential computing security for AI agents and heterogeneous accelerator testbed operations at 2026 OCP Korea Tech Day. Its lab|up /conf/6 conference, themed "Make AI Composable", was scheduled for September 29, 2026 at COEX.
▸Full profile — market position, technology, go-to-market, geography, history
Market position
Lablup presents itself as an AI infrastructure software provider serving leading organizations and acting as infrastructure partner within Korea's sovereign AI foundation model program, and is pursuing a KOSDAQ listing.
The company emphasizes resource-efficiency technology for AI infrastructure, unified management of GPUs and NPUs from multiple vendors on one platform, coverage of the full path from data preparation to serving, and demonstrated operation of GPU clusters exceeding 2,000 GPUs at a single site.
Technology
Backend.AI is a container-based computing cluster platform that hosts common computing and machine learning frameworks and multiple programming languages, with pluggable heterogeneous accelerator support covering CUDA GPUs, ROCm GPUs, Gaudi NPUs, Google TPUs, GraphCore IPUs and other NPUs. It includes a web and desktop UI offering dashboards and statistics in addition to CLI-equivalent operations, a FastTrack pipeline product, and an engine registry for Backend.AI:GO. Supporting technologies developed in-house include an asyncio RPC transport library, a Rust Raft framework, a Rust inference runtime for Apple Silicon and CUDA (mlxcel), a parallel SSH cluster tool (bssh), and a GPU monitoring utility (all-smi). The company also works on confidential computing security for AI agents and operates a heterogeneous accelerator testbed.
Go-to-market
Direct engagement through a sales, go-to-market and developer relations organization, an inbound contact channel for AI cluster planning and production scaling, open-source distribution of Backend.AI components on GitHub, developer content via the Lablup and Backend.AI technical blogs, a partner ecosystem spanning hardware, storage, cloud and service providers, and an annual company conference (lab|up /conf).
Research organizations, enterprises and public-sector institutions that run AI training and inference workloads on GPU and NPU clusters, ranging up to single-site deployments of more than 2,000 GPUs.
Geography
Headquarters in Gangnam-gu, Seoul, Republic of Korea, with a US office in San Jose, California.
History
Lablup Inc. develops the Backend.AI platform and related open-source projects, maintains offices in Seoul and San Jose, and in 2026 entered the KOSDAQ listing process, receiving preliminary listing approval in July 2026 and filing a securities registration statement in August 2026.
Compiled by commissioned research from 4 cited public sources — announcements, filings, and press listed under research sources below.
Key figures
latest reportedCompany-reported or press-reported figures, each dated to when it was claimed — not independently audited.
Timeline · 5
launches, deals, and filingsLablup scheduled its sixth company conference, themed 'Make AI Composable', for September 29, 2026 at COEX in Seoul.
Lablup filed a securities registration statement to advance its KOSDAQ IPO, citing unified GPU and NPU management, AI training and inference on a single platform, subscription licensing, and operation of GPU clusters of more than 2,000 GPUs at a single site.
Lablup announced it would present on confidential computing security for AI agents and operation of a heterogeneous accelerator testbed at 2026 OCP Korea Tech Day.
Lablup serves as infrastructure partner in the Upstage-led consortium, which advanced to the third phase of Korea's sovereign AI foundation model project.
Lablup announced it received preliminary listing approval for the KOSDAQ market, which the company describes as the shortest approval time of the year.
Dated company events from announcements, filings, and press; legal rows summarize public dockets and regulator releases.
▸Research sources · 4
primary sources listed
- LABLUPlablup.com · web
4 public sources were cited for this profile; the first-party ones are listed here.
Frequently asked questions
- What does LABLUP do?
- Korean company building Backend.AI, an operating system for AI infrastructure that manages GPU and NPU clusters.
- Who founded LABLUP?
- LABLUP was founded by Jeongkyu Shin, Joongi Kim.
- Who are LABLUP's investors?
- LABLUP's investors include Kakao Ventures, LB Investment.
- Where is LABLUP headquartered?
- LABLUP is headquartered in Seoul, KR.