Behind Jensen Huang’s Tang Suit & Mandarin: Nvidia’s $17B Game in China & the Battle for AI Dominance

Tang suit, Mandarin, applause—behind Jensen Huang’s Beijing debut lies a $17 billion fight for survival.

"I grew up in the U.S., but I learned a lot of Chinese." At the opening of the 3rd China International Supply Chain Expo (CISCE), Nvidia CEO Jensen Huang began his speech in slightly accented yet earnest Mandarin. Dressed in a dark Tang suit, the AI chip titan stood at the center of the Beijing stage, paying high tribute to China’s supply chain:

"China’s supply chain is a miracle—extremely complex and advanced, weaving together countless suppliers, capabilities, and technologies."

Just 24 hours before the speech, Huang revealed a bombshell to Chinese media in Beijing: The U.S. government had officially approved Nvidia to resume sales of its H20 AI chips to China.

Facing cameras, he made no effort to hide his excitement: "This is very good news. We’ve received massive orders." Behind this brief statement lay three grueling quarters of waiting for Nvidia in the Chinese market.

01 Tang Suit & Mandarin: A Carefully Staged China Debut

On July 16 in Beijing, Huang chose an unprecedented way to appear on a Chinese stage. A dark Tang suit replaced his iconic leather jacket, and halting Mandarin took the place of fluent English—the tech giant’s leader presented a "localized" image for his high-profile expo debut.

"This is my first time speaking in Chinese. I’m a bit nervous," Huang admitted at the start. That nervousness masked meticulous planning for his third China trip this year.

Two days earlier, a photo of Huang standing shoulder-to-shoulder with Xiaomi founder Lei Jun, both beaming, went viral on social media. Taken under Beijing’s 35°C heat, Huang was still wearing his signature black leather jacket.

Sources revealed this was part of an unannounced client visit, accidentally captured by a passerby.

Huang’s CISCE message was precise and deliberate. He highlighted that over 1.5 million Chinese developers are innovating on Nvidia’s platform, creating world-class models like DeepSeek, Alibaba’s Tongyi, and Tencent’s Hunyuan.

"China’s open-source AI is a catalyst for global progress," he declared. When the chip magnate described a future where "AI will command robots to collaborate with humans" in Mandarin, the hall erupted in applause.

02 The Sanctions Game: H20 Chips and the U.S.-China Tech Tug-of-War

On July 15, Huang broke major news to Chinese media in Beijing: The U.S. government had greenlit H20 exports to China. "I’m looking forward to shipping very soon. I’m very happy—this is really great news," he told reporters, visibly thrilled.

Behind this "good news" lay months of tense maneuvering. In April, Washington suddenly banned Nvidia from selling the China-specific H20 chip, citing "national security." Though its performance is just 15%-30% of its flagship H100, this "castrated" chip accounted for 80% of Nvidia’s China revenue.

The impact was immediate: Nvidia took a $5.5 billion hit, and its stock plunged 6%. Worse, export controls cost the company over $10 billion in lost sales across two quarters.

How crucial is China? Data shows that in the fiscal year ending January 2025, China contributed $17 billion—13% of Nvidia’s global revenue.

Huang repeatedly warned: "Losing China would weaken U.S. tech competitiveness." This became his key bargaining chip with Washington.

The approval came amid layered motives: U.S. Senator Marco Rubio hinted the "H20 deal was tied to rare earth agreements," pointing to mineral trade-offs. Meanwhile, Chinese chips like Huawei’s Ascend 910B now outperform the H20, and China’s AI chip self-sufficiency is projected to hit 40% by 2025—undercutting U.S. containment.

03 Three Trips to China: Huang’s Defense and Offense

With 2025 not yet half over, Huang has visited China three times. Each trip sent distinct signals, weaving Nvidia’s defensive net.

In January, Huang toured Shenzhen, Beijing, and Shanghai, joining Nvidia’s China team celebrations. He handed out bonuses up to $5,500 per employee while stressing China’s strategic role—even as Nvidia’s market cap hit $4 trillion, clouds gathered over its China business.

In April, he returned to meet China Council for Trade Promotion head Ren Hongbin. "China is a vital market," he stated publicly, seeking deeper ties. This came as U.S. ban rumors swirled; one month later, H20 sales halted.

For his July visit, Huang prepared thoroughly. Before departing, he met Donald Trump at the White House. When Chinese reporters asked about it, he downplayed: "He knew I was coming to China and speaking at CISCE. He wished me a pleasant trip."

Strategically, Nvidia is pursuing a dual track: restarting H20 sales for immediate needs while launching RTX Pro GPUs for industrial digital twins to open new fronts. Huang emphasized in his speech: "The next wave of AI will be robotic systems" that understand, reason, and act in the physical world.

04 Rise of Domestic Chips: Nvidia’s China Challenge

Behind Huang’s charm offensive lies a harsh reality: Chinese AI chips have made critical breakthroughs. Huawei’s Ascend 910B delivers twice the inference efficiency of H20 at 20% lower cost—and has won orders from the Middle East. Cambricon’s Siyuan 590 matches H20 in inference tasks, with Q1 revenue soaring 4,230% YoY.

Top Chinese tech firms are adopting "hybrid tactics": using H20 for daily inference while training on Ascend chips. This mix cuts costs by 60%, sidesteps supply risks, and accelerates local ecosystem adoption.

Deeper shifts are reshaping the tech landscape. China aims to lift domestic AI chip market share above 40% by 2027, using efficiency standards to curb imports. State-owned telecom giants already source over 50% locally.

"H20 relies on Nvidia’s CUDA ecosystem—software remains a weakness for Chinese chips," noted an industry analyst. But that gap is closing fast: While 90% of China’s AI code runs on CUDA, migration costs to Huawei’s MindSpore have dropped to ~$400K per firm.

05 Future Battlefield: AI Ecosystems vs. Robotic Systems

At CISCE, Huang painted a visionary future: "Within a decade, factories will run on robotic teams directed by software and AI, building smart products alongside humans." This wasn’t just prophecy—it’s Nvidia’s roadmap.

Today, hundreds of Chinese projects use Nvidia’s Omniverse for digital twins—optimizing factories and training autonomous vehicles.

Huang spotlighted how "AI powers China’s iconic platforms": WeChat, Taobao, Douyin; Xiaomi’s smart cars and phones; Meituan’s delivery network. These cases showcased Nvidia’s embedded reach.

Facing China’s chip rise, Nvidia plans a September launch of its Blackwell-based B30 chip, tweaked for export rules. Huawei’s Ascend 920 (rivaling H100) will enter mass production by year-end—signaling intensifying high-end chip wars.

"China’s innovation pace exceeds all expectations. We must keep evolving to stay relevant," Huang admitted in an interview. This confession reveals the core of Nvidia’s China strategy: finding a path to survive in the narrow space between tech blockade and open market.

After the H20 approval, U.S. Treasury Secretary Beth Steel admitted the chip was used as a "negotiating chip" in London talks. Chinese Foreign Ministry spokesperson Lin Jian fired back: "China opposes politicizing tech and trade."

Markets reacted instantly: Nvidia shares jumped nearly 4% pre-market, while China’s computing power and CPO stocks surged. Yet analysts see a new reality—during the ban, local chips grabbed 25% market share, and top firms’ hybrid setups slash costs by 60%.

"H20 is candy given by others—the real roots grow from your own trees," commented Pan Helin, an MIIT expert, capturing the battle’s essence. As Huang spoke at CISCE of AI-run factories, engineers at Huawei labs were fine-tuning the Ascend 910B—a chip twice as powerful as H20.

In this war for AI supremacy, victory won’t hinge on chip specs alone. It will be decided by 1.5 million Chinese developers voting with their feet—choosing the ecosystem that shapes tomorrow.

--The above article is purely AI-generated. If there is any infringement, please contact Cloud Engine/Cloud GPU Rental Platform for deletion!