NVIDIA Dominates AI Chip Market as Chinese Rivals Diverge

As the sun sets on November, the spotlight in Asia's semiconductor scene continuously gravitated towards Nvidia. In a recent visit to Hong Kong, Nvidia's co-founder and CEO, Jensen Huang, received an honorary doctorate in engineering from the Hong Kong University of Science and Technology. On November 25, during a meeting in Beijing with Jay Puri, Nvidia's Executive Vice President, Wang Shouwen, the Minister-level International Trade Negotiators’ Representative of China's Ministry of Commerce, was informed that Nvidia regards China as a critical market.

The wave of artificial intelligence has turned Nvidia (NVDA.O) into a cornerstone of the technology sector. With an explosive surge in demand, the company has emerged as a leader in accelerated computing, driving AI development rapidly forward and reshaping industries around the globe.

However, Nvidia is not the only player in the burgeoning AI landscape. In China, which it has identified as a vital market, a host of companies dubbed “the Chinese Nvidia” have also emerged. Prior to 2019, amidst a backdrop of active capital support, a wave of seasoned engineers from chip giants like Nvidia, AMD, and Huawei's HiSilicon ventured into entrepreneurship. While these founders brought varying backgrounds and experiences, what bound them was a common vision of attempting to carve out their niche in the booming AI market.

Advertisement

Over the years, these domestic contenders began to chart their own paths. Some are now eyeing public listings, while others have faced the harsh realities of the market, resulting in layoffs and even company closures. At the Hong Kong University of Science and Technology commencement, Huang reflected on AI’s profound impact on our era, declaring: “Artificial intelligence is undoubtedly the most important technology of our time; the world has been reset.”

In this intense competition, survival of the fittest is the name of the game. The technological trajectories and operational models of these homegrown chip manufacturers are now poised for their next market test.

Looking back at the inception of these “domestic Nvidias”, we find that when the first Chinese companies ventured into AI chips, Nvidia was merely synonymous with gaming graphics. The story took a notable turn in 2016, with the establishment of Cambrian (688256.SH), which grew from a small academic team of ten researchers at the Institute of Computing Technology, Chinese Academy of Sciences.

CEO Chen Tian-shi reminisced about the early days: “When we first embarked on developing AI chips, the segment wasn’t hot at all—AI itself wasn't in vogue, and targeting AI with dedicated chips was even less so.” But they nevertheless caught the initial AI wave when Google's AlphaGo made headlines in 2016, sparking a global fascination with AI.

By the following year, Cambrian’s chips powered Huawei's flagship Mate 10 smartphone, which featured the Kirin 970 processor embedded with the Cambrian 1A chip. In 2020, Cambrian rang the bell and proudly proclaimed itself as the “first AI chip stock in China.” Meanwhile, during the same period, Nvidia’s data center revenues surged to $1.75 billion, marking a historic pivot away from graphics cards to becoming Nvidia's primary revenue stream.

This frenzy around AI chips led to a flurry of investments and the birth of several companies such as Sunway Technology, Moore Threads, Biren Technology, and Xiangdi Xiang Technology. Sunway Technology's founding story revolves around former AMD colleagues Zhao Lidong and Zhang Yalin, while Biren's founder Zhang Wen transitioned from senior investment roles on Wall Street to leading significant tech ventures, thus enhancing Biren’s fundraising capabilities. Within just 18 months from its inception, Biren raised over 4.7 billion RMB, establishing a record in funding for Chinese semiconductor startups.

Moore Threads stands closest to Nvidia, founded by Zhang Jianzhong, a former Nvidia VP. Under his leadership, the company sought to develop a complete ecosystem for GPUs in China. Each of these start-ups emerging during the AI boom presented chips capable of competing with some of Nvidia's latest offerings—Cambrian's Suyuan 590, Moore Threads' GPU dubbed "Chunxiao", Biren's general-purpose BR100 series, Sunway's AI training chip “Suishi 2.0”, and Xiangdi Xiang’s “Tianjun 1 GPU.”

As generative AI took off, coupled with international trade tensions and technological embargoes, demand for Nvidia's GPUs skyrocketed. This situation endowed these domestic companies with the moniker "Chinese Nvidia", encapsulating the burgeoning aspiration within local semiconductor manufacturing.

Yet today, some of these self-styled "Chinese Nvidias" are finding their fortunes diverging significantly from the Nvidia legacy.

In late August, when Xiangdi Xiang Technology announced large-scale layoffs, Sunway Technology was executing a listing mentor agreement with CITIC Securities, commencing an IPO process to become the “second AI chip stock.” Despite Xiangdi Xiang’s subsequent clarification that it was adapting its workforce due to pressures in the Chinese GPU market, the negative perception lingered.

The adjustments made at Xiangdi Xiang were understood to be efforts to optimize team structure and reduce operational costs, while ensuring core R&D teams remained intact to drive continued innovation in the GPU sector.

Concerns about funding also surfaced, with indications that the company was actively engaging potential investors to ease their financing challenges. The alarm bells may have sounded even earlier. In April, upon being named a “unicorn” for 2024, the company’s CEO mentioned converting unissued bonuses into company stock options. Yet there were employees who reported not receiving any such documentation, casting doubts on leadership’s promises.

Bearing the brunt of corporate decisions, employees encountered abrupt layoffs, often without room for negotiation. Current and former employees shared critical insights into the harsh realities of the layoffs while also hinting at deeper financial troubles that hindered investor attraction, concerns reflected in internal meetings and discussions.

Xiangdi Xiang had completed various funding rounds since its founding in 2020, yet they had struggled to secure sufficient investment as valuations soured. Reports indicated unfulfilled contractual obligations related to a financing agreement, contributing to the company’s squeeze. Meanwhile, industry experts highlighted that the mounting issues were symptomatic of broader cash flow challenges exacerbated by an inadequate R&D budget for the competitive landscape.

In a contrasting narrative, other companies have started to find their footing. Sunway has progressed towards an A-share IPO, while Biren recently disclosed plans for an initial public offering. Following closely, Moore Threads activated its own IPO process in November, further marking a trend of burgeoning public interest in domestically produced AI chips. Though Cambrian remains unprofitable, its stock has recently captured investors’ attention, with significant valuations aimed at shoring up its market standing.

Despite the contrasting fortunes, one salient truth remains apparent: the AI chip sector inherently demands extraordinary financial investments. For instance, Nvidia’s R&D expenditures exceeded 61.6 billion RMB in 2023, which rose sharply to 65.4 billion RMB within just nine months, marking a staggering 47% increase from the previous year. Cambrian’s numbers tell a similar story, with their projected R&D expenditures in the early quarters of 2024 surpassing their entire revenue—2.12 billion RMB in R&D versus 1.21 billion RMB in revenue.

The grim financial reality underscores a critical point: the road to success in the AI chip space is anything but smooth. Continuous cash flow, fundraising, and relentless innovation become each firm's lifeline until they achieve stable and self-sustaining operations. For example, Moore Threads has successfully completed five rounds of financing, drawing investment from a mix of state funds and major tech institutions such as ByteDance and Tencent. Meanwhile, Tencent has participated extensively in Sunway’s multiple funding rounds, positioning itself as a major stakeholder.

Importantly, even as companies strive to replicate Nvidia’s success, leveraging GPU markets, they also explore differentiated strategies. Many interviewees noted that edge AI represents a promising avenue. Several players, including Moore Threads, Biren, and Sunway, have escalated their focus on edge AI inference projects. Yet, building a successful AI ecosystem is contingent on solid product development, industry partnerships, and evolving market comprehension.

As Huang succinctly put it, “Artificial intelligence has ushered in a new era of computing—impacting every scientific domain and industry.” The era of AI signifies a profound transformation on multiple fronts; firms must navigate unprecedented challenges, surmount external obstacles, and ultimately adapt to a rapidly changing technological landscape.

In conclusion, while the race among domestic chip manufacturers to emulate the success of Nvidia is on, the complexities and intricacies involved suggest that the path ahead will be laden with trials and uncertainties. Embracing flexibility and proactive innovation may well be the only strategy that helps these firms not only survive but thrive amidst new global challenges.