LG Uplus said on Sunday it will join hands with domestic AI semiconductor company FuriosaAI to develop a “sovereign AI appliance” that can be used securely inside companies.

LG Uplus signed a memorandum of understanding with FuriosaAI for cooperation in AI infrastructure at MWC26 in Barcelona on March 4 local time. Sang-yeop Lee (이상엽), LG Uplus chief technology officer, and Jun-ho Baek (백준호), FuriosaAI CEO, attended the signing ceremony.

The two companies will cooperate to develop the sovereign AI appliance, an integrated AI device in which data is not sent to an external cloud and is processed only in an on-premise environment. The sovereign AI appliance can be used immediately by connecting only power and a network, without complex server configurations or facilities. It is a finished product that can be used as soon as it is installed, without building AI infrastructure separately.

The sovereign AI appliance will integrate LG Uplus’ enterprise AI platform, LG AI Research’s AI model EXAONE 4.0 and FuriosaAI’s neural processing unit (NPU).

LG Uplus will design an operating structure that enables in-house document search and summarisation, task agentisation and workflow automation based on its enterprise AI platform and EXAONE 4.0. It plans to build a system that can quickly search data and respond even in a closed-network environment, providing an operating model suited to corporate settings.

FuriosaAI will use Renegade, its second-generation NPU that has recently begun mass production, to raise EXAONE 4.0 inference performance. It plans to design an infrastructure structure that maintains stable processing performance even when many users use AI at the same time, while lowering power consumption and costs.

The two companies expect the sovereign AI appliance will enable stable AI operations in environments where use of external clouds is restricted, including public, defence, medical, finance, manufacturing and research institutions. They explained it can also secure a structure capable of 24 hours a day, 365 days a year operation, even at sites where AI system downtime has a major impact on work.

The two companies also plan to expand cooperation starting with this partnership into areas including AI data center-based NPU as a service (NPUaaS) and physical AI. In AI data centers, they will push to advance NPU-based inference infrastructure design and operating structures. In industrial sites requiring robot or equipment control, they will jointly research ultra-low-latency inference technology and operating models.

Jun-ho Baek (백준호), FuriosaAI CEO, said the cooperation is meaningful in that it combines AI models, platforms and semiconductor technology into a single structure and implements it in a form that can be used immediately in industrial sites. He said FuriosaAI will work with LG Uplus to build an AI infrastructure model optimised for internal corporate operations.

Sang-yeop Lee (이상엽), LG Uplus CTO, said for AI to be applied in actual work environments it must have security and operational stability as well as performance. He said LG Uplus will prepare hybrid AI infrastructure that companies can trust and use, starting with development of the sovereign AI appliance.