DeepX raises $30M: South Korea produces its own AI inference chip for the edge

South Korean startup DeepX raises $30M for its AI chip dedicated to edge inference. Following datacenter GPUs, Korea bets on a specialized chip for on-device deployments—free from Nvidia dependency.

In plain terms. DeepX, a South Korean semiconductor startup, raises $30M to accelerate its edge AI inference chip. This isn’t a chip for data centers—it’s for devices, robots, and industrial equipment that need to infer locally, without cloud access.

Our take. The edge AI market follows a different logic than data center GPUs: strict power constraints, critical latency, and large-scale deployment in environments with unreliable connectivity. DeepX is targeting this niche with a chip optimized for on-device inference—outside Nvidia’s radar, which remains focused on data centers. The funding aligns with South Korea’s broader momentum: SK Hynix (HBM), Samsung (memory, foundry), and now startups like DeepX aiming for the logic layer. The Korean AI supercycle isn’t just about memory components—it’s an ambition for a complete semiconductor stack.

To watch. DeepX’s early industrial partnerships (robotics, automotive, manufacturing) and energy efficiency benchmarks against Qualcomm and MediaTek’s edge chips.