A Low-Key PR Move

Cited by Investor Network | Wu Wei

On January 24, 2026, at the Lujiazui Jinde Vegetable Market in Shanghai, a seemingly “ordinary customer” made an appearance at a fruit stall. With a practiced hand, he picked out several ponkan oranges and then handed the stall owner a red packet with his signature on it. This scene quickly spread like wildfire on social media. For Jensen Huang, the CEO of NVIDIA (NASDAQ: NVDA), this moment that appeared relaxed and “down-to-earth” was, in fact, a high-stakes public relations (PR) maneuver and a strategic inspection under intense pressure. It is reported that Huang arrived in Shanghai on January 23, marking his first visit to China in 2026. Unlike his high-profile engagements two years prior, this time around, he eschewed public speeches and media interviews. Instead, he focused on internal communications at NVIDIA's new Zhangjiang office and held closed-door meetings with key clients.

Behind the Silence: A Fierce Market Battle

Behind this apparent silence lies a cutthroat market competition. On one hand, the U.S. government has conditionally approved the entry of NVIDIA's H200 chip into China. However, this comes at a high cost, including substantial tariff expenses and a complex end-user review process. On the other hand, domestic alternatives, led by Huawei's Ascend series, have rapidly gained market share over the past year. Huang's visit is aimed at “rallying the troops” for the Chinese team and safeguarding NVIDIA's last stronghold in the Chinese market before the domestic AI ecosystem becomes fully self-sufficient.

The Silent 'Ice-Breaking': H200's China Entry and Compliance Challenges

While the vegetable market visit garnered significant attention, the capital market is likely more interested in Huang's internal talk at NVIDIA's new Shanghai office, located at 600 Naxian Road, Zhangjiang Hi-Tech Park.

This marked Huang's first visit to the office building since its inauguration. According to sources close to NVIDIA, during his interactions with employees, Huang deliberately steered clear of sensitive topics related to technology exports and instead emphasized the company's long-term vision. This “beating around the bush” approach reflects the structural contradictions that NVIDIA's business in China is currently facing.

Since 2024, NVIDIA's revenue share in China has plummeted due to geopolitical tensions. To reverse this trend, NVIDIA has been actively promoting the entry of its higher-performance H200 chip into the Chinese market.

However, the “return” of the H200 has been far from smooth, as it encounters dual constraints. The first is the erosion of competitiveness due to compliance costs. Although the H200 has received approval for entry into China, the imposition of high tariffs and stringent “End-User Review” mechanisms has significantly driven up procurement costs for Chinese clients. Moreover, delivery times have stretched from a few weeks to several months, creating uncertainty that directly undermines the long-term reliance of internet giants on NVIDIA.

The second constraint is the forced transformation of market strategies. During the two-year supply gap, companies such as Baidu (NASDAQ: BIDU), Alibaba (09988.HK), and Tencent (00700.HK) have diversified their computing power sources and are no longer solely reliant on NVIDIA. Huang's low-key meetings with these key clients are aimed at offsetting the disadvantages of hardware delivery delays through software optimization services and system integration solutions. For Huang, the challenge is to convince the Chinese market that, even under heavy restrictions, NVIDIA's ecological value still justifies the high “premium” paid by clients.

Under Siege: Domestic Substitution and Ecological Competition

The high level of attention that Huang and NVIDIA are paying to the Chinese market stems not only from the fact that it once contributed nearly a quarter of NVIDIA's global revenue but also from the fact that China is emerging as the only market globally capable of forging an “alternative path” in AI computing power.

Industry research indicates that domestic substitution in China is no longer a simple hardware replacement but a systemic reconstruction that spans from underlying chips to software ecosystems. This reconstruction impacts NVIDIA primarily in two ways.

In critical infrastructure sectors such as government, enterprises, and finance, the localization rate has surged due to stringent information security requirements. Domestic computing power companies like Hygon Information (688041.SH) and Cambricon (688256.SH) have effectively covered NVIDIA's mid-to-low-end inference products through deep collaboration with telecom operators.

Within the realm of internet giants, “hybrid architectures” have become the norm. It is estimated that, in terms of inference, domestic chips have captured nearly 50% of the workload in some large model businesses, thanks to their higher cost-effectiveness and stable supply.

In the past, NVIDIA's core asset was the CUDA ecosystem. However, under the weight of sanctions, Huawei's CANN software stack and compatibility tools provided by startups like Moore Threads are rapidly lowering the migration threshold for large models. A notable trend is that new-generation domestic large model manufacturers, exemplified by DeepSeek, are beginning to optimize for domestic hardware from the early stages of development. This “reverse adaptation” initiated from the application side is transforming NVIDIA from a “must-have” to an “option” in the Chinese market.

Huang is acutely aware that if Chinese tech giants become accustomed to innovating under non-CUDA ecosystems, NVIDIA risks permanently losing its foothold in the world's largest growth market. Therefore, Huang must maintain frequent on-site visits to sustain ties with China's developer community.

The Shanghai Model: AI's 'Independent Evolution' Fueled by Policies

Just a few kilometers away from the new Zhangjiang office building that Huang inspected, Shanghai's AI industry is evolving at a breakneck pace, following a distinct logic. Shanghai has become a vantage point for observing domestic substitution by constructing a closed-loop industrial chain that spans from chip design and manufacturing to model applications.

The Shanghai municipal government has adopted a highly pragmatic approach to promote AI autonomy and controllability. In July 2025, Shanghai launched an AI industry support program worth approximately RMB 10 billion. The most direct measure is the implementation of “AI Computing Power Vouchers,” which provide financial subsidies of up to 20%-30% to large model companies renting domestic computing power. This “real money” incentive directly offsets the cost disadvantages of domestic chips in their early iterative stages and promotes the large-scale adoption of domestic computing power.

In addition to financial support, Shanghai has deployed high-density intelligent computing centers in Zhangjiang and Lingang. With SMIC (688981.SH) serving as the manufacturing foundation, Shanghai has assembled the country's top chip design cluster. At the same time, Shanghai has emerged as a hub for large models, with companies like SenseTime (00020.HK), MiniMax, and Step AI contributing to the country's leading number of model registrations.

This industrial density, where “supply and demand are resolved within the same city,” has significantly accelerated the iterative cycle of domestic computing power. Domestic cards can identify issues during local large model testing and quickly resolve them at local chip design factories, creating an efficient feedback loop that overseas manufacturers struggle to match.

According to Shanghai's development plan, the city's total intelligent computing power will exceed 30 EFLOPS by 2025. This means Shanghai is not only building a consumer market but also an independent, self-sustaining AI infrastructure. The successful implementation of this “Shanghai Model” signifies that China's AI industry has shifted from mere “technology following” to “ecological self-improvement.”

Conclusion: A Strategic Crossroads

Huang's 24 hours in Shanghai epitomize the struggle of commercial giants to find a balance in a complex geopolitical environment.

The red packet he handed to the vegetable stall owner was a gesture of goodwill, but what he left for the Zhangjiang employees and clients was a dilemma of “choice.” In today's world, where computing power is synonymous with national strength, NVIDIA still possesses cutting-edge technology. However, the Chinese market is no longer content with being a mere purchaser.

As domestic computing power transitions from “usable” to “user-friendly,” NVIDIA's moat in the Chinese market is no longer impregnable. Huang's “extreme attention” to the Chinese market is essentially a strategic alertness to an emerging competitive system. For China's AI industry, by continuously accumulating data and optimizing algorithms through pioneering zones like Shanghai, the path to computing freedom is becoming increasingly visible in the night of Zhangjiang. (Produced by Thinker Finance) ■

Source: Investor Network