Currently, advancements in artificial intelligence (AI) technology are driving the evolution of AI from iterative improvements in algorithms to significant breakthroughs in computational infrastructure. During this round of technological evolution, a massive demand for AI inference computing is emerging, setting new benchmarks for the architecture and cost-efficiency of computing power.
Compared with general-purpose GPU architectures, NPU chips designed specifically for AI inference scenarios have become the foundation supporting the large-scale commercialization of AI industries due to the advantages such as high cost-effectiveness, energy efficiency and customization. These chips are gradually becoming one of the mainstream development directions for AI chips, accelerating the industry’s transition from an era centered on GPUs for large model training to an era focused on NPUs for AI inference computing. In this race, the innovation capabilities at the foundational architecture level and extensive experience in large-scale application deployment are progressively contributing to a company’s overall competitiveness and sustainability.
Among these players, Shenzhen Intellifusion Technologies Co., Ltd. (“Intellifusion”) is the first company to commercialize domestic high computing power AI inference chips in China. As a pioneer in NPU research and design, Intellifusion integrates multiple generations of architectural technical knowledge with mature commercialization experience to create high-performance, cost-effective and highly versatile inference chips. Intellifusion has achieved large-scale deployments across enterprise, consumer and industry-grade scenarios, establishing a leading position in China’s AI inference chip-related products and services market in terms of revenue in 2024. On July 30, Intellifusion submitted its listing application to the Hong Kong Stock Exchange.
Led by a “Hardcore” management team: providing strong traction for strategic implementation
In the deep tech sector of AI chips where technology, capital and talent concerntration converge, the management team serves not only as decision-maker for corporate strategy but also as “navigators” guiding Intellifusion through technological and industrial cycles. Intellifusion’s ability to gain a firm foothold in the highly complex and fiercely competitive AI inference chip market is inseparable from its core team, which boasts strategic foresight, technical prowess and hands-on industry experience.
Dr. Chen Ning, founder, executive director and general manager of Intellifusion, is a worldwide rare expert with deep experience in semiconductor and possesses both overseas and domestic industry experience. Standing at the forefront of technology, Dr. Chen deeply understands the path from theoretical breakthroughs to product transformation, and not only possesses visionary technical foresight, but also demonstrates strong execution capability in industrialization and commercialization, which enable him to grasp Intellifusion’s long-term strategic trajectory from the cutting edge of technological development with commercialization capability, and drive deep innovation along the “algorithm + chip” integration path, providing long-term traction for Intellifusion’s sustained growth.
As one of the key pioneers in China’s domestically-developed AI inference chips, Dr. Chen Ning has spearheaded the design and development of homegrown AI inference chips, achieving end-to-end innovation from algorithmic breakthroughs to chip-based implementation. He can rightly be considered one of the founding figures in the industrialization of China’s AI inference chips. His recognition as the prominent individuals in innovation and entrepreneurship and outstanding role models at the 40th Anniversary of Shenzhen Special Economic Zone, and in July of this year, he was selected alongside entrepreneurs such as Xingxing Wang, founder of Unitree Robotics, Weiliang Chen, founder of MetaX, and Peng Zhang, CEO of Zhipu AI, as one of CCTV’s ‘AI Leaders of the Year’,.further solidifies his strategic leadership position in China’s AI industry.
Under the leadership of Dr. Chen Ning, Intellifusion has assembled a elite Research & Development (“R&D”) team with core members averaging over 25 years of experience in the AI industry. These veterans have long served at leading domestic technology companies, covering the entire critical technology chain from chip architecture design to algorithm development and optimization. As of the Latest Practicable Date, Intellifusion boasts 489 R&D personnel, including more than 50 engineers with over a decade of chip design experience and multiple national-level leading technical experts. This composition has established a synergistic system for chip design and algorithm development guided by seasoned industry specialists, providing robust support for rapid product innovations and continuous technological breakthroughs.
“Algorithm + chip” integrated strategy builds high technological barriers
Unlike traditional “algorithm companies” or “chip companies” that focus on singular breakthroughs, Intellifusion has consistently adhered to the advanced concept of co-designing algorithms and chips since its inception. This approach has forged an integrated software-hardware technical capability that creates a self-reinforcing virtuous cycle, where insights from real-world deployment continuously refine algorithms, algorithmic innovations drive chip architecture upgrades, and enhanced chip performance unlocks new application scenarios. This creates a compounding flywheel effect, progressively elevating R&D efficiency and commercial scalability.
At the foundational technology level, Intellifusion has established its proprietary “IFIC” platform through its algorithm-to-silicon integration capability. This end-to-end system encompasses the entire AI inference chip R&D process: algorithm analysis – instruction set definition – chip architecture design – toolchain optimization. This enables Intellifusion to develop successive generations of NPUs and AI inference chips, ensuring optimal efficiency and scenario-specific adaptability.
For architectural design, Intellifusion launched “AI Computing Blocks” by employing fully home-grown advanced domestic manufacturing processes, secure supply chain with advanced semiconductor processes, featuring standardized compute units that can be stacked like building blocks to flexibly configure chips with variable computing power and enable multi-chip interconnect for scalable systems, realizing single tape-out enables multiple packaging variants, efficiently producing chips of diverse specifications while balancing flexibility and cost. To address the data transmission bottleneck of traditional computer chips, Intellifusion has pioneered an innovative Near-memory Hyper-converged Architecture, which realizes vertical integration by seamlessly combining memory and computing units in a stacked configuration, enabling direct data flow through the 3D structure, eliminating bandwidth limitations of traditional packaging interconnections, achieving on-chip bandwidth and sub-nanosecond latency, dramatically lowering the energy consumption of data transportation, and perfectly adapting to the needs of real-time loading of large-model parameters.
In the chip domain, Intellifusion has developed its DeepEdge Series of inference chips based on the “IFIC” Infrastructure, utilizing the innovative ” AI Computing Blocks” Architecture. Currently, DeepEdge10 Series covers a wide range from 8T to 128T of computing power per chip and supports Transformer-based models for efficient AI inference, which can realize balanced performance & efficiency, low power consumption, and flexible deployment.
In terms of technology update, Intellifusion has completed development of its 4th-generation NPU and is now advancing R&D for the next-generation high-performance NPU, Nova 500.
This integrated R&D system centered on “algorithms + chips” enables Intellifusion to not only ensure leading-edge performance and rapid iteration in chip design but also continuously strengthen its technological moat through product adaptability, deployment scalability, and ecosystem expansion. Notably, Intellifusion has been awarded the highest award for intelligent science and technology in China. – the “Wu Wenjun AI Science & Technology Award ” – three times, making it the only company in China to have won top prizes in all three categories: algorithms, chips, and applications. This recognition underscores its end-to-end innovation capability, from core technology development to industrial implementation.
Riding the Global Inference Wave: Securing Long-Term Value in AI Chips
As the proportion of inference computing in the total lifecycle cost of large models continues to rise, surging application demand will dramatically expand the scale of inference computing needs, creating a historic opportunity for the AI industry. According to the CIC Report, the market size of AI inference chip-related products and services industry in China is expected to reach RMB1,383.0 billion by 2029, at a CAGR of 53.4% from 2024 to 2029. In particular, the NPU-powered market size is projected to climb to RMB395.4 billion by 2029, at a CAGR of 72.3% from 2024 to 2029, significantly outpacing the overall market.
Under the guideline of the advanced algorithm-to-silicon chip design methodology, Intellifusion, a pioneer in the high-potential market, has established an integrated software-hardware technical capability that creates a self-reinforcing virtuous cycle. Its products and services have been validated across multiple critical application scenarios, positioning Intellifusion as an industry frontrunner. According to the CIC Report, Intellifusion is a global top-three leader in full-scenario AI inference chip-related products and services in China, based on the relevant revenue in 2024. Intellifusion is also a top-two provider of NPU-powered AI inference chip-related products and services in China, based on the relevant revenue in 2024.
Leveraging its IFIC Infrastructure, R&D capabilities in AI inference chips, and deep market insight, Intellifusion can respond quickly to changes in industrial demands and grasp the opportunity in the upcoming boom in AI inference chip-related products and services industry. Supported by policy tailwinds, growing application maturity, and semiconductor supply chain localization trends, Intellifusion is strategically positioned to capitalize on China’s golden window for domestic AI chips, emerging as both a key beneficiary and driver in the global recomposition of AI compute power.
In summary, Intellifusion has not only achieved full autonomy from self-developed architectures to ecosystem development but also made breakthroughs across market expansion, technological innovation, and product commercialization, establishing a difficult-to-replicate end-to-end advantage. Should its Hong Kong IPO proceed smoothly, at the time of listing, Intellifusion could potentially become China’s first and only artificial intelligence company with a dual A+H listing (Shanghai + Hong Kong). This milestone would fully unlock cross-border M&A opportunities, allowing Intellifusion to aggregate global resources, attract top-tier AI talent worldwide, and further solidify its leadership in China’s AI industry, particularly in the inference chip segment.
