Nvidia to Launch Affordable Blackwell AI Chip for China in June

Nvidia is set to introduce a cost-effective graphics processing unit (GPU) based on its cutting-edge Blackwell architecture for the Chinese market, priced between $6,500 and $8,000, sources familiar with the matter revealed.

This move, slated for mass production as early as June 2025, comes in response to tightened U.S. export controls that banned Nvidia’s pricier H20 chip in April.

Despite a shrinking market share in China, Nvidia aims to stay competitive against domestic rivals like Huawei by offering a lower-spec, budget-friendly chip tailored to comply with U.S. restrictions.

A Strategic Response to U.S. Export Curbs

Nvidia’s new GPU, derived from its RTX Pro 6000D server-class processor, is designed to navigate U.S. export restrictions that limit advanced AI chip capabilities, including memory bandwidth capped at 1.7-1.8 terabytes per second.

Unlike the H20, which retailed for $10,000-$12,000 and boasted 4 terabytes per second, the new chip uses conventional GDDR7 memory instead of high-bandwidth memory (HBM) and forgoes Taiwan Semiconductor Manufacturing Co.’s (TSMC) advanced Chip-on-Wafer-on-Substrate (CoWoS) packaging.

This simpler design slashes production costs, enabling the lower price point, according to industry insiders. Nvidia’s spokesperson noted the company is still evaluating “limited” options pending U.S. government approval, as China’s $50 billion data center market remains critical, accounting for 13% of Nvidia’s sales last year.

ALSO READ | Taiwan’s AI Leap Forward: NVIDIA and Foxconn Power a Groundbreaking Supercomputer

China’s Competitive Landscape and Nvidia’s Challenges

China’s AI chip market has grown fiercely competitive since U.S. export curbs began in 2022, slashing Nvidia’s market share from 95% to 50%, as reported by CEO Jensen Huang in Taipei on May 20, 2025.

Huawei’s Ascend 910B chip has gained traction, with analysts predicting Chinese technologies could match Nvidia’s downgraded chips’ performance within one to two years.

Nvidia’s edge lies in its CUDA platform, widely used by developers for AI model integration, giving it a temporary advantage.

However, the new GPU’s reliance on conventional memory may limit its suitability for intensive AI training and inference tasks, according to Nicolas Gaudois, head of technology research at UBS Asia, in a May 2025 analysis.

Did You Know?
Nvidia’s CUDA platform, launched in 2006, is used by over 4 million developers globally, powering AI applications across industries, per a May 2025 Nvidia report.

Production Plans and Market Impact

Mass production of the new Blackwell-based GPU is expected to begin in June 2025, with a second chip variant potentially entering production by September, sources said.

While the chip’s final name remains undecided, possibly the 6000D or B40, per a May 2025 GF Securities note, its memory bandwidth of approximately 1.7 terabytes per second aligns with U.S. restrictions.

The H20 ban forced Nvidia to write off $5.5 billion in inventory and forgo $15 billion in sales, Huang revealed in a May 2025 podcast. Despite these setbacks, the new chip’s affordability could help Nvidia regain footing in China, where demand for AI infrastructure remains robust, driven by data center growth reported in a May 25, 2025, industry analysis.

Looking Ahead: Nvidia’s Balancing Act

Nvidia’s pivot to a budget-friendly chip underscores its adaptability in a constrained market. However, as U.S. restrictions tighten and Huawei closes the technological gap, Nvidia faces pressure to innovate while complying with export limits.

The company’s ability to leverage its CUDA ecosystem and maintain developer loyalty will be critical, analysts say. With China’s AI market evolving rapidly, Nvidia’s latest move could either solidify its position or signal further challenges ahead.