AMD, FuriosaAI and MangoBoost have signed an agreement to explore mixed-chip AI infrastructure, ChosunBiz reported from an AMD event in Seoul on November 4.
The big change
- The plan combines AMD CPUs and GPUs with FuriosaAI neural processors and MangoBoost data processors and software.
- A Seoul reference cluster is planned for spring 2027 to test performance and interoperability.
- Lower inference costs and power use are goals, not demonstrated results.
Why mix chips?
Inference runs a trained AI model to produce an output. Systems serving those requests also coordinate applications and move data. In a general explanation, AMD describes CPUs handling application and coordination work, GPUs accelerating AI workloads, and data processing units taking on networking or storage tasks.
FuriosaAI describes its RNGD processor as an NPU designed for deep-learning inference. That vendor description is not a measurement of the proposed combination.
What remains to be tested
ChosunBiz says the companies will review inference infrastructure designs and pursue a reference system using MangoBoost's Seoul infrastructure. The spring target is a plan, not evidence that a cluster is running.
Separately, South Korea's science ministry said it held the first steering committee for its earlier AMD agreement, covering AI infrastructure, a research centre and training. That government process is distinct from the three-company memorandum.
