Nvidia to use Groq license for China-focused AI chips under export controls

Nvidia is moving to use Groq’s technology license to develop specialized AI chips for the Chinese market while complying with U.S. export controls, according to Reuters. The company disclosed at its GTC 2026 developer conference in San Jose that a Groq-powered chip line tailored for China is in the works.

The move follows a roughly $17 billion license deal Nvidia struck with Groq late last year to access Groq’s technology. Nvidia’s announcements at GTC 2026 indicate the Groq-based designs will be integrated into a new product lineup, aimed at preserving its share of the Chinese AI inference market.

NVIDIA sign outside their headquarters office campus on Scott Boulevard in Santa Clara, California, located at 2800 & 2806 Scott Boulevard, Santa Clara, California 95050. NVIDIA is best known for making Graphical Processing Units, or GPUs, but also creates System-on-a-Chip, or SoCs, such as the NVIDIA Tegra, used in a variety of mobile applications. NVIDIA owns Arm Holdings, which designs ARM CPUs, and Mellanox Technologies, maker of next-generation networking devices.
Representative image for context; not directly related to the specific event in this article. License: CC BY 2.0. Source: Wikimedia Commons.

The strategy also aligns with Nvidia’s broader effort to resume production of the older generation H200 chips, a path pursued after receiving an export license during the Trump administration that allowed Chinese orders. Nvidia says it will adapt its designs to keep Chinese demand flowing while staying within regulatory constraints.

A key part of the plan is the use of Groq’s language processing unit, or LPU, technology. Nvidia argues that the LPU can significantly speed up AI inference compared with traditional GPUs, and the company intends to employ Groq’s approach to maintain strong performance in China without breaching export rules for its next-generation Vera Rubin chip, which cannot be shipped to China.

Groq’s LPU is positioned as a core element in Nvidia’s China-focused strategy, providing a path to high-speed inference in a regulatory-compliant package. Nvidia described the Groq-enabled designs as crucial for real-time AI responses in China’s growing AI deployment landscape.

ASUS NVIDIA GeForce 210 silent graphics card with HDMI
Representative image for context; not directly related to the specific event in this article. License: CC BY-SA 3.0. Source: Wikimedia Commons.

The development matters beyond Korea because it highlights how major U.S.-based AI hardware firms navigate export controls while pursuing market presence in China. The approach has implications for global supply chains, pricing, and competition in the AI accelerator sector, and it illustrates how policy decisions shape technology development and international markets.

For U.S. readers, the situation is relevant to broader tech competition and national security considerations. Chinese demand for advanced AI chips can influence global supply dynamics, supplier ecosystems, and strategic policy debates in Washington about how to balance promoting domestic innovation with managing technology competition and security risks.

Subscribe to Journal of Korea

Don’t miss out on the latest issues. Sign up now to get access to the library of members-only issues.
jamie@example.com
Subscribe