Nvidia eyes Korean chip startup Rebellions, valued at $2.3B
What's the deal? Nvidia is in early talks with Korean AI chip designer Rebellions about a possible tie-up, ranging from a technical partnership to an investment or even an acquisition, according to people familiar with the matter. Nvidia chief executive officer Jensen Huang met Rebellions co-founder and CEO Sunghyun Park at the US company's Santa Clara headquarters this week.
The company: Founded in 2020 and based in Bundang, South Korea, Rebellions designs neural processing units for data centers, optimised for AI inference — the higher-volume workload that runs AI services at scale. It is among a group of semiconductor startups building chips to rival market leader Nvidia.
By the numbers: Six-year-old Rebellions has raised roughly $850 million from investors including SK HynixDealroom has a profile for this one. Try Dealroom →, Samsung VenturesDealroom has a profile for this one. Try Dealroom →, and Arm HoldingsDealroom has a profile for this one. Try Dealroom →, plus a direct investment from the Korean government. It was most recently valued at around $2.3 billion.
Why now? Huang has led a far-reaching investment spree as Nvidia works to keep its central role in the AI era. In late 2025, it struck an unconventional deal with US AI firm Groq, securing a nonexclusive licence while absorbing most of its engineering talent — a move Nvidia said was not an acquisition, since Groq keeps operating and running its cloud business independently.
What could go wrong? Deliberations are preliminary and may not lead to a transaction, the people said. Nvidia's dominance in AI training chips already invites antitrust scrutiny, and large deals typically need clearance from bodies such as the US Department of Justice.
The Korea factor: An acquisition-like arrangement could also draw scrutiny in Korea, where semiconductors are viewed as critical strategic assets. Samsung ElectronicsDealroom has a profile for this one. Try Dealroom → and SK HynixDealroom has a profile for this one. Try Dealroom →, leading memory-chip makers, work closely with the government on new investment projects.
The signal: Whichever form a deal takes, Nvidia's interest in an inference specialist signals where the next battle lies — not in training AI models, but in the larger, everyday task of running them.
Read more: Bloomberg
Image credit: Generated with Gemini
