EN / 中文

Jensen Huang’s Resignation Accelerates China’s Domestic GPU Substitution Wave

by juchaoWAVE·November 19, 2026

By Cheng Du, Edited by Yang Xuran

Jensen Huang, a leading figure in the global Chinese tech business community who has led NVIDIA for 33 years and propelled it to become the world's most valuable company, resigned from his positions as President, CEO, and Director due to unexpected circumstances.

On NVIDIA's official website, Jensen Huang has been removed from the executive leadership list and placed alongside Chris Malachowsky under the "Founders" section. Even on platform X, his verified blue checkmark and CEO title have disappeared.

The era of Jensen Huang steering this global AI computing powerhouse has effectively ended. Even if he were to return for certain reasons, NVIDIA's strategic approach to global computing capacity would inevitably undergo significant changes.

On November 12, following the announcement, NVIDIA's stock price dropped by more than 13% during intraday trading, placing pressure on the A-share NVIDIA supply chain. However, on the other side of the coin, NVIDIA's long-time rival AMD, along with domestic GPU companies such as Cambricon, Enflame Technology, and MetaX, began to attract greater market attention.

Jensen Huang, as a global leader among Chinese tech entrepreneurs, managed one of the most critical companies in the AI era, yet he was forced to resign from all management positions within such a short timeframe. This coincided with the conclusion of the US midterm elections, indicating that the underlying political and economic maneuvering is extremely complex.

The departure of a founder will not cause NVIDIA's technology moat, accumulated over many years, to vanish. In other words, there will be no obvious issues with its production and supply. The problems lie with customers who have clear demands for computing power and are caught in international trade disputes—namely, Chinese AI companies.

Jensen Huang's personal exit poses stricter challenges to China's domestic GPU industry, as supply risks have been fully exposed. The Chinese AI industry will increasingly recognize clearly that while achieving "autonomous controllability" is a long and difficult road, it has become a top-priority project that must be addressed.

This article is an in-depth value piece from the Juchao WAVE content team. We welcome your attention across multiple platforms.

Reversal

GPUs are the underlying hardware for large model training and inference, as well as for constructing computing clusters, effectively serving as the "heart" of the AI industry.

Since October 2022, the US directly included high-end chips such as NVIDIA's A100, H100, and AMD's MI300 on the control list. Subsequently, NVIDIA's "special edition" chips, ranging from the A800 and H800 to the H20 and H200, have also been restricted from export to the Chinese market.

By the first half of this year, the contribution of the Chinese market to NVIDIA significantly contracted, and its market share in China had fallen to 37%, being overtaken by domestic peers—although this does not represent NVIDIA's share of all available computing power used by Chinese AI enterprises, as institutions and equipment outside the Chinese market also provide computing power for Chinese companies.

Currently, domestic GPUs have passed the stage of "whether products can be made" and have entered a period of scaled delivery. In the Q3 2026 reports, core manufacturers generally achieved several-fold growth in revenue, and profitability quality has shown clear improvement.

Cambricon and Hygon Information were the first two companies to enter a positive profit cycle with leading scale.

In particular, Cambricon's net profit margin for the first three quarters approached 39%, with net earnings of CNY 3.7 billion, indicating that it has survived the critical life-or-death juncture and possesses real profitability. Hygon Information has the largest scale, with revenue for the first three quarters already approaching CNY 15 billion and net profits of CNY 2.8 billion.

Moore Threads, the "first domestic GPU stock," saw rapid revenue expansion, crossing the break-even point into profitability. Its Q3 revenue hit a single-quarter high of CNY 1.552 billion, and its non-GAAP net profit turned positive for the first time at CNY 20.07 million. However, its cumulative non-GAAP loss for the first three quarters remained at CNY 131 million, with operating cash flow showing a net outflow of CNY 3.309 billion.

Moore Threads follows a full-function GPU route, with products covering AI computing, graphics rendering, scientific computing, and video encoding/decoding. While the potential is vast, R&D and ecosystem construction similarly consume cash.

MetaX reported a net profit attributable to parent company shareholders of CNY 639 million for the first three quarters. After deducting significant fair value changes, its non-GAAP net profit was CNY 259 million. Moreover, the company has achieved non-GAAP profitability for two consecutive quarters, indicating that revenue is beginning to cover costs such as R&D.

Enflame Technology, which listed later, saw its revenue surge by 3.3 times in the first three quarters but remains in a high-investment phase. Its single-quarter loss in Q3 narrowed to CNY 122 million, though sustained profitability is still distant. However, due to consecutive sharp rises in its stock price, its market capitalization has surpassed that of Moore Threads and MetaX, which hold more advanced positions in the industry.

Biren Technology, listed on the HKEX, saw its H1 revenue grow nearly20 times, with gross margins rising to 42.7% and losses narrowing significantly. However, its stock price has retraced nearly 70% from its peak, and the company needs to overcome further hurdles such as customer repurchases, software adaptation, and delivery to achieve sustained profitability.

Another HKEX-listed company, Iluvatar CoreX, achieved profitability in H1 with revenue of CNY 946 million, nearly triple that of the same period last year. Its adjusted net profit was CNY 246 million, marking its first turnaround from loss to profit since listing. However, this profit primarily stemmed from CNY 760 million in equity investment income; excluding this, its core business still incurred losses.

It can be seen that domestic GPUs have entered a period of rapid volume expansion, with domestic players seeing substantial revenue growth and improving profitability. Some leading companies have even entered a phase of sustained profitability, meaning the operational status of domestic AI chip companies has improved significantly compared to before.

Game Theory

Jensen Huang's departure occurred during the sensitive period following the US midterm elections, further disrupting the stability of the global AI supply chain.

Over the past several years, based on NVIDIA's commercial interests, Jensen Huang long advocated for relaxing chip export controls against China. After the A100 and H100 were restricted from export to China, NVIDIA immediately launched the "downgraded versions" A800 and H800. Once the A800 and H800 were added to the prohibited sales list, the H20 quickly emerged.

He also leveraged his personal reputation, status, and influence to mediate between Congress and the Department of Commerce. In discussions regarding US chip export controls against China, Jensen Huang was one of the most explicit and influential opponents.

Dario Amodei, founder of Anthropic, represents Silicon Valley's "containment faction." He has long maintained a hardline stance advocating for strict restrictions on China's access to advanced US AI chips, stating that selling H200 chips to China would be a "huge mistake," comparable to "selling nuclear weapons to North Korea."

Subsequently, Jensen Huang responded in a podcast episode, calling the analogy of comparing AI chips to nuclear weapons "crazy," describing the logic of export controls as "extremely stupid" and reflecting a "loser mentality."

In some sense, Jensen Huang served as a "buffer pad" in the Sino-US chip game. His departure effectively silenced the most influential voice opposing US chip export controls, removing a heavyweight counterbalance from internal US debates.

After Jensen Huang's resignation, US House Speaker Johnson stated that consensus within Congress regarding restrictions on semiconductor exports to China is "expanding." Meanwhile, the US Department of Commerce is researching new chip export control rules, intending to include Southeast Asian countries and others under stricter export license management to block transshipment routes.

Some reports indicate that during Jensen Huang's tenure, NVIDIA retained some ambiguous operational space when selling to China, allowing certain Chinese data center projects to obtain specific NVIDIA chips. These projects largely involved contracts signed years ago, with some already under construction.

After Jensen Huang's departure, once control measures tighten again, these originally operable "gray areas" will be completely interrupted, causing projects to either stall or force redesigns at higher costs.

The drastic change in the external environment has directly blocked the strategy where many domestic large models treated NVIDIA chips as "high-performance" options with domestic GPUs as supplements, forcing the entire industry to shift toward endogenous substitution solutions.

If the use of NVIDIA chips faces more restrictions, domestic chips will default to becoming the "primary choice," accelerating the arrival of the "substitution window" for domestic computing power.

For example, DeepSeek's next-generation large model V5 will rely natively on Huawei Ascend for pre-training for the first time. Previously, Ascend only handled inference or post-training, while DeepSeek V3 and V4 completed training on NVIDIA chips and ecosystems.

If successful, this will mark a shift of domestic chips from the inference side to the more challenging training side, demonstrating the capability to support the full-process development of top-tier large models.

Overtaking

At the Huawei Connect conference on September 17, Huawei Rotating Chairman Xu Zhijun mentioned: "It is difficult to statistically measure NVIDIA's market share in China, but based on the data we can track, Ascend should have already surpassed NVIDIA."

As NVIDIA falls into a dual squeeze of management vacuum and congressional investigation, domestic GPUs are growing freely at the juncture of exploding AI demand and obstructed exports by international giants.

Domestic GPUs have currently formed a landscape of "one superpower (Huawei Ascend) + multiple strong contenders (T-Head, Cambricon, Kunlunxin, etc.)." According to IDC statistics, in the first half of this year, NVIDIA shipped 678,000 units in China, a sharp year-on-year drop of 44%, while Huawei's shipments surged to 572,000 units, with the difference in market share between the two sides being only 5.3%.

Alibaba's T-Head shipped 133,000 units, ranking third, followed by Cambricon, Baidu Kunlunxin, Hygon Information, and Moore Threads, with shipment volumes in the range of 50,000 to 100,000 units.

Furthermore, some institutional analyses suggest that Huawei Ascend's share in the Chinese market has reached half, while NVIDIA may only retain around 8%. Compared to the situation in 2022, where NVIDIA held 95% alone and Ascend just 3%, a massive reversal has occurred over four years.

Morgan Stanley estimates that the self-sufficiency rate of Chinese AI chips will be 35% in 2025, reach 45% in 2026, jump to 68% in 2027, reach 77% in 2028, and hit 95% in 2030, essentially achieving complete autonomous substitution.

The recent turmoil in NVIDIA's management has further accelerated the localization process. Just five days after Jensen Huang's departure, NVIDIA established a new National Security and Export Control Oversight Committee, responsible for board-level supervision of the company's national security policies, export control compliance, and global trade.

The US side also quickly began investigating Singapore-based Megaspeed, alleging illegal transshipment of NVIDIA chips to China. Foreign media reports suggested that former executives of this Singaporean company might have personal relationships with Jensen Huang.

From various signs, the window for purchasing high-end overseas chips is shrinking, making further localization of GPUs imperative.

Compared to NVIDIA, domestic manufacturers still lag in advanced process nodes, single-card performance, HBM supply, stability in ultra-large-scale training, and system ecosystems.

In particular, CUDA has accumulated over a long period, resulting in a thriving ecological rainforest, imposing huge migration costs on domestic chips. NVIDIA's status as the global AI chip hegemon rests fundamentally on the CUDA moat. Developers are accustomed to CUDA, which remains the biggest barrier facing new chips like Ascend, whereas the gap in hardware itself is not the primary issue.

However, domestic manufacturers are already seeking new solutions.

Hua Wei, Chief Scientist of Huawei Semiconductor, frankly admitted that the first obstacle encountered by Ascend is not hardware, but mainly ecological barriers. Now, with the emergence of new compiler languages, the software barriers of CUDA are gradually dissolving, and frontier AI labs no longer prioritize CUDA as they did two years ago.

In terms of performance, since single cards cannot compete, thousands or even tens of thousands of cards are assembled into a supercomputer to enhance performance through cluster formation. This is the "Super Node" route currently bet upon by Huawei and Alibaba.

As Xu Zhijun pointed out, Huawei is achieving global leadership in overall computing power through systemic innovation via "Super Nodes + Clusters."

Huawei has released the world's first Super Node Ascend 960 based on NPO technology, scalable to 4,096 cards; Alibaba launched its first SNPO optical interconnect Super Node Panjiu AL64 at the Yunqi Conference. Sugon, Tsinghua Unigroup, MetaX, and others have also joined in succession.

Just as DeepSeek took a different path, this innovation focused on engineering and systems offers Chinese chips the possibility of catching up with NVIDIA in performance.

Final Words

"The leather jacket remains, but the crown is lost."

Jensen Huang was once the "buffer zone" for NVIDIA's trade with China, quotas, and even gray transshipment channels. He coordinated the Sino-US computing power contradictions in a relatively pragmatic, somewhat gray manner.

But with the change in leadership, controls by NVIDIA and likely more US tech companies against China will probably continue to tighten. This will inevitably push the Chinese AI industry chain to completely abandon the fantasy of "obtaining cards via detours" and fully bet on the autonomous route.

From a longer time perspective, NVIDIA's leadership change is not the starting point but an accelerator. As Jensen Huang repeatedly emphasized, restricting US chip exports can only force China to achieve more independent innovation.

Now, the consensus in China's AI chip industry has strengthened further, and comprehensive localization is merely a matter of time.