Rebellions raises $400M pre-IPO round at $2.34B valuation, launches two AI infra platforms
What's the deal? South Korean AI chip startup Rebellions has raised $400 million in a pre-IPO funding round, pushing its valuation to approximately $2.34 billion and total funding to $850 million.
The round was led by Mirae Asset Financial GroupDealroom has a profile for this one. Try Dealroom → and South Korea's state-backed National Growth Fund, a public-private vehicle operated by the Financial Services CommissionDealroom has a profile for this one. Try Dealroom → and the Korea Development BankDealroom has a profile for this one. Try Dealroom →, marking its first direct equity investment under the K-Nvidia initiative.
Alongside the raise, the company launched two new products: RebelRack and RebelPOD, fully deployable AI infrastructure platforms.
Founded in 2020 by Sunghyun Park, Rebellions designs AI inference chips and outsources fabrication. Its Rebel100 neural processing unit (NPU) is built for inference — the compute required for AI models to respond to user queries in real time — rather than for training. The company has operations in the US, Japan, Saudi Arabia, and Taiwan.
The raise follows a $250 million Series CDealroom has a profile for this one. Try Dealroom → closed in November 2025, meaning Rebellions has pulled in $650 million in the six months to March 2026 — over 75% of its total capital to date.
Why now? As large language models mature and see widescale commercial deployment, inference is becoming the dominant compute workload — and its economics increasingly favour specialised hardware over general-purpose GPUs. Rebellions is betting that enterprises, cloud providers, and governments will pay a premium for inference-optimised silicon that runs efficiently within real-world power and cost constraints.
The KNGF's involvement is also telling: South Korea is positioning Rebellions as a national champion in the global race for AI infrastructure sovereignty, with state-backed capital amplifying private investment.
What could go wrong? Nvidia's grip on the AI compute stack remains formidable, and hyperscalers including AWSDealroom has a profile for this one. Try Dealroom →, Google, and Meta are all developing proprietary inference chips. Rebellions must convince cloud providers and enterprises to fragment their supply chains — a high bar even with competitive performance-per-watt metrics.
As a fabless chipmaker, Rebellions is also exposed to fabrication bottlenecks and supply chain disruptions it cannot fully control.
The signal: The AI infrastructure investment wave is moving downstream — from models to the silicon and systems that run them. Rebellions' trajectory reflects a broader market conviction that specialised inference hardware will be as strategically important as the models it serves.
The KNGF's backing signals something larger still: nations are treating AI compute as critical infrastructure, and state capital is flowing accordingly.
Sources:
Rebellions
CNBC
Reuters
FinSMEs
TechCrunch
Morningstar
Data Center Dynamics
Silicon Canals
PR Newswire
Image credit:
Rebellions
J.V.
