AI for Automation
Back to AI News
2026-03-19NVIDIAAI chipsChinaJensen HuangAI industry

Beijing just approved Nvidia's AI chips for Chinese buyers

China approved Nvidia H200 chip sales after years of restrictions. Jensen Huang says Blackwell and Rubin chips alone could generate $1 trillion by 2027.


After years of back-and-forth between Washington and Beijing over AI chip exports, China has officially approved Nvidia to sell its H200 AI processor to Chinese customers. CEO Jensen Huang confirmed the company has secured U.S. export licenses and orders from "numerous Chinese customers."

This is a big deal. China once represented 13% of Nvidia's total revenue, and the company had to halt H200 production last year due to regulatory obstacles on both sides. Now that dam has broken.

Which chips and why they matter

The H200 is Nvidia's second-most-powerful AI processor. While it's technically a previous-generation chip, it outperforms every alternative currently available in China. Chinese tech companies have been making do with older, less capable hardware — and some, like Baidu, have started building their own chips to fill the gap.

H200 — Nvidia's second-most powerful AI chip, now approved for Chinese buyers

Groq inference chip — A China-compatible version launching in May, based on technology Nvidia acquired in a $17 billion deal last year

Blackwell & Rubin — Nvidia's next-generation systems that Huang predicts will generate over $1 trillion in revenue by 2027

Who's buying

In January, Beijing granted preliminary import approvals to China's biggest tech companies: ByteDance, Tencent, Alibaba, and DeepSeek. These are the companies behind TikTok, WeChat, and some of the most advanced AI models coming out of China. With H200 access, they'll be able to train and run larger AI models significantly faster.

Nvidia is also building a China-ready version of its Groq inference chip (inference = the process of actually running an AI model after it's been trained). These aren't stripped-down versions — they're designed to be adaptable to different computing systems while meeting export requirements. That chip is expected in May.

The $1 trillion prediction

Jensen Huang used the announcement to make a staggering forecast: Nvidia's Blackwell and Rubin chip systems alone could generate over $1 trillion in revenue by 2027. For context, Nvidia's total revenue in 2025 was around $130 billion. A trillion-dollar target would require the entire AI infrastructure buildout to continue accelerating.

Whether you believe that number or not, the direction is clear: AI chips are becoming as strategically important as oil was in the 20th century. And Nvidia just opened a massive new market.

What this means for the AI industry

More competition from Chinese AI companies — With better hardware, expect faster progress from DeepSeek, Alibaba's Qwen, and other Chinese AI labs. This could drive down AI costs globally.

Easing geopolitical tensions — The approval signals a potential thaw in the U.S.-China tech standoff. Both sides benefit: Nvidia gets revenue, China gets hardware.

Nvidia's dominance deepens — Even with Chinese competitors building their own chips, Nvidia remains years ahead. The H200 approval cements that lead.

Source: The Decoder

Related ContentGet Started with Easy Claude Code | Free Learning Guides | More AI News

Stay updated on AI news

Simple explanations of the latest AI developments