Korean AI Chip Startup Rebellions Raises $400M to Challenge Nvidia

Seoul: Rebellions has raised $400 million in a pre-IPO funding round, valuing the South Korean AI chip startup at $2.34 billion. The round was led by Mirae Asset Financial Group and the Korea National Growth Fund. Rebellions is the first company backed under South Korea’s K-Nvidia initiative to develop a domestic AI chip champion.

Participating investors include Samsung Catalyst Fund, SK Hynix, Aramco’s Wa’ed Ventures, and KT. The raise brings total funding to $850 million. More than 75% of that capital arrived in the past six months alone.

Rebellions builds inference chips, silicon designed specifically for running AI models in production, not training them. This new direction increasingly seen across startups building model-specific AI inference chips to optimize performance and cost.

Its flagship Rebel100 NPU powers two newly launched systems: RebelRack, a production-ready inference compute unit, and RebelPOD, which clusters multiple RebelRacks for larger deployments. Both are available now.

The pitch is performance-per-watt. Rebellions targets the inefficiency gap in AI infrastructure, the cost and energy burden of running inference workloads at scale. Its chiplet-based architecture is optimised for that specific problem rather than retrofitted from training silicon.

The software stack is fully open source. It integrates natively with vLLM, PyTorch, Triton, Hugging Face, and OpenShift, a deliberate choice to reduce developer friction and avoid proprietary lock-in.

That software-first approach positions Rebellions as a systems company, not just a chipmaker. At the same time, innovation is also accelerating at the design layer, with companies exploring AI-driven chip design automation to shorten development cycles and improve efficiency.

Rebellions competes directly with Nvidia on inference, alongside a growing field of challengers. Cerebras and Groq are pushing their own architectures. Cisco has also entered the race with its own AI chip targeting Nvidia and Broadcom. Nvidia still dominates through ecosystem and developer adoption, but the pressure from multiple directions is building.

The competitive landscape is also expanding through startups working on AI chip architecture and design innovation, pushing beyond traditional GPU-based approaches.

Rebellions is betting that energy efficiency and open standards will matter more as inference workloads scale globally. The $400M will fund US market expansion, scaled production of the Rebel100 platform, and IPO preparation. An IPO is targeted for later this year.