Nvidia is in discussions with South Korean AI chip startup Rebellions about possible collaborations.

Conversations have focused on a potential technical partnership, investment, or even an acquisition, , Bloomberg reports.

Talks between Nvidia CEO Jensen Huang and Rebellions co-founder and CEO Sunghyun Park were held at Nvidia’s headquarters in Santa Clara, California, this week, the sources added. However, the report noted that the deliberations between the two companies were preliminary and were not guaranteed to lead to a transaction.

A representative for Rebellions declined to comment to Bloomberg.

Founded in 2020, Rebellions specializes in AI inference accelerators that offer energy efficiency and low-latency performance.

The company began mass-producing its Atom and Atom Max NPUs to run large language models (LLMs) in the data center in 2023. The chips have been deployed by customers in Japan, Saudi Arabia, and the US, the company said, adding that it is looking to expand its presence in the US, Europe, and Asia-Pacific to support sovereign AI infrastructure initiatives.

In June 2024, the company agreed to a merger with Sapeon Korea, a semiconductor company that was spun out of SK Telecom in 2016, and in July of the following year, partnered with Marvell to offer high-performance, energy-efficient AI systems that support AI deployments across APAC and the Middle East.

In March 2026, Rebellions raised $400 million in a pre-IPO funding round and unveiled two new AI infrastructure platforms, dubbed RebelRack and RebelPod. No target date for the IPO has been disclosed.

Nvidia denies China-specific LPU reports

The news of a potential collaboration between the two chip companies comes in the same week The Information reported that Nvidia has developed a China-specific version of its language processing unit (LPU) and was planning to start shipping small batches of the chip before the end of the year, with several customers having already placed orders.

However, Nvidia has since denied the report, with a spokesperson for the chip company stating: "The reporting in The Information on Nvidia's LPU is incorrect. We have no LPU sales in the China market today, and no China-specific ​LPU product in our roadmap.”

Following the publication of its Q1 2027 financial results in May, Nvidia said it was not recording any anticipated compute revenue from China in its second-quarter projections.

Nvidia unveiled its forthcoming LPU, dubbed the Nvidia Groq 3, at its annual GTC conference in March 2026, two months after announcing its $20 billion license agreement with Groq.

“The LPU... is optimized strictly for that extreme low-latency token generation, offering token rates in the thousands of tokens per second,” said Nvidia data center head Ian Buck back in March. “The trade-off, of course, is that you need many chips in order to get that kind of performance. And the economics, or the tokens per second per chip, is actually quite low.”

Developed using technology from Groq, Nvidia’s LPUs are expected to start shipping to customers in Q3 2026. At GTC, the chip giant said that in addition to continuing with its annual GPU release cadence, it is now on track to introduce a new LPU architecture each year too.

The Nvidia Groq 3 LPU is set for release in the second half of 2026 and will be available in liquid-cooled LPX racks, which feature 256 LPUs with 128GB of on-chip SRAM and 640 TBps of scale-up bandwidth. The rack is focused on low-latency AI inference workloads.