In recent years, the semiconductor industry has faced significant challenges, leading to a wave of bankruptcies across various chip companiesLast year alone, over ten thousand chip firms closed down, averaging about 31 failures per dayThis grim situation has continued into the current year, marked by the bankruptcy and liquidation of emerging IP design firms like Huaxia Chip and the financial struggles of Lisan Technology, which burned through 300 million RMB and faced wage disputes in June of the previous yearAs if that wasn't enough, the withdrawal of dollar-based funds has pushed these companies to the brink, nearly collapsing just before their critical production phases.

Lisan Technology, established in 2021, began operations early in 2022 with its subsidiary in Shanghai, which serves as the company's core

The founders, Xuan Yifang, Kong Dehai, and Niu Yixin, bring extensive engineering experience and technical expertise from S3, a GPU company based in Silicon ValleyWith a vision to create competitive domestic GPU chips that mirror mainstream global products, Lisan Technology has positioned itself as a key player in China's estimated 200 billion RMB graphics processing market.

According to the official information from the company, Lisan Technology is committed to developing its own architecture with independent intellectual property rightsIts goal is to create domestic GPU chips that serve a myriad of applications, ranging from education, work, and entertainment to the metaverse, gaming, digital twins, film production, augmented and virtual reality (AR/VR), smart manufacturing, professional design, smart cockpits, and intelligent healthcare.

Fortunately, Dongxin Co., Ltd

• Common Psychological Pitfalls for Investors.
• Tongcheng Targets International Expansion After $8B Valuation
• Can A-shares Reach New Heights with Expected Rate Cuts?
• Global Chip Spending Jumps 31% in Q4
• How Billionaires Navigate the U.S. Tax System

recently announced plans to invest in Lisan Technology, potentially acquiring up to 40% of its stakes with a maximum investment of 200 million RMBThis support reflects the intensifying competition in the GPU sector, highlighting the critical challenges firms like Lisan Technology face in a rapidly evolving market.

The advent of the GPU (Graphics Processing Unit) significantly transformed how graphical data is processed and displayedInitially designed to process images and handle primitive data exclusively, the role of GPUs has expanded dramatically over the years.

Origins of the GPU

Developed as specialized chips, GPUs are inherently suited for intensive data handling compared to CPUs

The architectural design of GPUs has evolved from the original CPU structure, integrating optimizations that allow them to share RAM with the CPUBy employing SIMD (Single Instruction, Multiple Data) architectures, GPUs can process thousands of operations simultaneously—this parallel processing capability is what distinguishes them from traditional CPUs.

GPUs excel at executing multiple tasks concurrently, transforming computational data into the stunning graphics we see on our screensFor computational tasks, GPUs leverage their parallel computing advantages to accelerate processing in areas such as deep learning, scientific computation, and other data-intensive applications.

The foundational work in computer graphics began in 1962 with MIT doctoral student Ivan Sutherland's paper and his pioneering sketchpad program

During the 1962-1984 period, there was no dedicated graphics processing hardware; all graphics tasks were handled by CPUs.

As the need for more sophisticated image processing grew, the narrative of GPU development evolvedIt can be traced back to graphic display controllers like the monochrome display adapter and color graphics adapter integrated into the first personal computer—IBM5150—released in 1981. This innovation opened new doors, leading to the advancement of high-end graphic workstations and the eventual commercialization of GPUs.

By 1995, the release of 3DFX's Voodoo graphics card marked a pivotal moment in GPU history, spurring a competitive landscape with major players like AMD, ATI, and NVIDIA entering the fray

This led to significant changes in CPU’s role in the graphics processing domain, alleviating some tasks from CPUs but indicating that others, such as vertex transformations, still required CPU input.

Founded in 1993, NVIDIA introduced the concept of the GPU with the release of its GeForce 256 graphics chip in 1999. This launch marked the emergence of the GPU as a significant component in graphic rendering technology, introducing advanced functionalities such as hardware T&L (Transformed and Lighting), environment mapping, and vertex blending.

As the industry matured, fixed-function pipelines gave way to programmable architecture, enabling GPUs to perform a myriad of visual effects dynamically

By 2003, both NVIDIA and ATI launched new products featuring programmable vertex processing, enhancing their capabilities.

Entering the 21st century, the advent of gaming and multimedia applications surged demand for enhanced GPU performanceChip manufacturers invested heavily in R&D for next-gen GPUs, changing the internal architecture significantlyThe composition of modern GPUs has diversified to include components like graphics memory controllers, compression units, video management units, and power management units.

With the rise of artificial intelligence and deep learning technologies, the application scope of GPUs broadened even further

NVIDIA's development of CUDA technology transformed GPUs from mere graphic rendering units into tools capable of handling extensive data analytics, deep learning, and machine vision tasks effectivelyThe evolution of GPUs from specialized processors to versatile computing platforms is a significant shift witnessed in this era.

Within China, a wave of companies dedicated to GPU research has emerged, such as Jingjia Microelectronics and XMOS Microelectronics, among othersChangsha-based Jingjia Microelectronics made history in April 2014 by creating the first domestically-produced high-performance, energy-efficient GPU chip, the JM5400, followed by the successful launch of the JM7200 in 2018.

In 2020, the chip industry experienced an entrepreneurial boom centered around GPUs

With significant investment, a multitude of start-ups entered the GPU domain, aiming to meet the ever-increasing demand for both graphic processing and AI computations—a field now commonly referred to in China as GPGPU (General-Purpose GPU). However, financial exhaustion has led to the downfall of some companies, resulting in bankruptcies amid tightening market conditions.

According to IDC data, the global AI chip market is projected to reach $35.2 billion by 2022, with GPUs commanding a major shareGoldman Sachs anticipates that by 2025, the GPU market share will soar to 57%, particularly as the U.Simposes restrictions on the importation of the latest Nvidia products, thus creating a favorable landscape for domestic GPU manufacturers in China.

A Triumvirate of Dominance in the Global GPU Market

In the global GPU market, a triadic oligopoly has emerged, with NVIDIA as the unchallenged leader among independent GPU manufacturers

Benefiting from its dominance in the laptop and traditional PC sectors, Intel has secured a commanding presence in integrated GPU markets, while NVIDIA and AMD share the independent GPU market.

NVIDIA has solidified its status as the leading developer and seller of standalone graphics cards, offering a lineup that includes the consumer-focused Geforce series and the professional Quadro seriesIts GPUs are renowned for capabilities like CUDA general-purpose computing, PureVideo high-definition video technology, and PhysX physics accelerationIn March of this year, NVIDIA introduced its new generational AI graphics processing architecture, Blackwell, alongside the launch of GPU products B200 and GB200 based on this architecture.

Intel, the world's largest supplier of integrated GPUs, primarily focuses on graphics chips used in its motherboards and laptops

The company released its first dedicated graphics card in 2022 and has since made advancements, recently integrating Battlemage G21's core code into its oneAPI DPC++ compiler documentation, signaling promising developments in its standalone GPU lineup.

AMD, a semiconductor company that has existed since 1969, possesses a unique capability to design high-performance CPUs, high-performance standalone GPUs, and chipset architectureIts integration of ATI in 2006 made it the first semiconductor company capable of creating both CPUs and GPUsAMD’s product ranges include its consumer Radeon series and the professional FireGL series.

In 2023, AMD officially launched next-gen AI and HPC accelerators—Instinct MI300X and MI300A—where MI300A combines both CPU and GPU into a singular device, specifically targeting high-performance computing (HPC) markets, demonstrating newly enhanced architecture and computing capabilities.

Flourishing Landscape of the Chinese GPU Market

Despite its late inception, the Chinese GPU market has recently exploded, thanks to significant investments and a surge in demand driven by artificial intelligence

Consequently, many companies are now vying for a share, leading to a vibrant ecosystem filled with emerging enterprises and remarkable products.

Jingjia Microelectronics stands out as China's foremost domestically developed GPU manufacturerFollowing years of technological exploration and development, they successfully created the JM5400—China's first high-performance GPU chip—overcoming the long-standing proficiency of ATI in military graphics controlSubsequent developments, including the JM7200, further bolster Jingjia's position in the market.

Xindong Technology has ventured into various domains, including semiconductor IP customization and GPU development

The release of the “Fenghua 1” in late 2021 marked a significant advancement in their offerings, particularly as a high-performance 12nm graphics card capable of supporting four simultaneous 4K60 fps outputs, filling crucial market gaps.

Mole Thread, established in 2020 by former NVIDIA global vice president Zhang Jianzhong, aims to lead innovations in GPU intellectual propertyTheir MUSA architecture, launched in the first year of establishment, showcases a potent desktop GPU equipped with advanced features and a robust framework.

The Longxin Zhongke's development of integrated GPUs aligns with their commitment to enhancing performance in personal computing and server applications, exemplified by the optimization-rich design of the 7A2000 processor.

Xintong Semiconductor, founded in December 2019, aims at GPU chip design and has launched the second generation GB2062 after successfully developing architectures suited for various computing applications.

Additionally, several other companies, like Birun Technology and Tienshu Smart Chips, are pushing the boundaries of GPU innovation

However, recent U.Ssanctions have constrained access to advanced manufacturing technologies, greatly impacting growth and development efforts.

Synergy of Domestic Innovation and U.SSanctions Creates Market Opportunities

Due to historical limitations, China has long relied heavily on imitation and imported technologies, with international giants commanding substantial market segmentsHowever, geopolitical tensions have led to a significant transition aimed at self-sufficiency in information technology, collectively known as "Xinchuang" (New Innovation), pushing for a domestic rebound.

In 2022, the U.S