AMD, South Korean AI chipmaker FuriosaAI and AI infrastructure company MangoBoost have signed a memorandum of understanding to develop an open, heterogeneous artificial intelligence infrastructure system, with a reference deployment planned for Seoul in 2027.
The companies announced the agreement during AMD Advancing AI Seoul 2026 on November 4. The planned system will combine AMD central processing units and graphics processing units, FuriosaAI neural processing units and MangoBoost data processing units and system software. The parties said they will jointly review designs and validate how the components work together across AI inference workloads.
MangoBoost said the reference deployment will be hosted at its Seoul facility. A company-distributed announcement said the first hybrid reference infrastructure is targeted to begin operating in 2027. The companies did not disclose its planned computing capacity, budget or prospective customers.
Heterogeneous computing assigns different parts of a workload to processors suited to those tasks. In the planned configuration, AMD will provide the CPU and GPU layer, FuriosaAI will supply its AI inference NPU, and MangoBoost will contribute DPU-based network acceleration and software intended to coordinate the infrastructure. The companies said the validation work will assess performance and interoperability, but they did not announce test results or commercial terms.
The project builds on AMD’s existing work with MangoBoost. In an October company blog, AMD described MangoBoost’s use of Instinct GPUs and ROCm software for AI inference services and its operation of computing infrastructure in Seoul. MangoBoost said its planned Mango AI Factory will use AMD Instinct GPUs alongside its own networking, server, storage and inference software.
FuriosaAI, founded in South Korea in 2017, develops inference accelerators. Its participation gives the reference project a Korean NPU component alongside AMD’s processors and MangoBoost’s infrastructure stack. In February, FuriosaAI signed a technology supply agreement with Nota AI to optimize AI models for its RNGD accelerator in data center deployments.
The Seoul project is intended to test whether the three companies’ processors and software can operate as a unified system. MangoBoost said the partners also aim to support a broader Korean AI infrastructure ecosystem and create a route for Korean technology into global supply chains. Any expansion beyond the reference deployment will depend on the validation work and subsequent commercial decisions.