On May 19, Jensen Huang, founder and CEO of NVIDIA (NVDA), attended the opening ceremony of the Taipei International Computer Show (Computex 2025) and delivered the first keynote speech, sharing the latest breakthroughs and progress in AI and accelerated computing technology.
This year’s COMPUTEX 2025 focuses on “AI Next” and focuses on the three major themes of “Intelligent Computing & Robotics”, “Next-generation Technology” and “Future Mobility”. The exhibition invited nearly 1,400 manufacturers and used about 4,800 booths, with a display area of up to 80,000 square meters.
The new next-generation GB300 system has a 1.5-fold performance improvement
At the meeting, Jensen Huang demonstrated a series of new products from NVIDIA, and released many products such as the new NVIDIA GB300 NVL72 platform, NVIDIA NVLink Fusion semi-custom dedicated AI chip, and RTX PRO Server enterprise-level AI computing platform, repositioning itself as an AI infrastructure construction company.
Jensen Huang said that DeepSeek-R1 demonstrated breakthrough performance improvements, with four times the computing power compared to leading competitors such as H100. To meet the demand for greater computing power, Nvidia launched a new system called “Grace Blackwell”, which transforms ordinary computers into powerful supercomputers.
In addition, artificial intelligence technology is accelerating iteration and is ushering in explosive development. In recent years, Nvidia has also continued to expand, especially increasing its investment in the field of AI.
In response, Jensen Huang announced that it plans to upgrade to the Grace Blackwell GB300 version this quarter, which will be available in the third quarter of this year. The GB300 system is equipped with an upgraded Blackwell chip, which has a 1.5-fold increase in inference performance, doubled network connectivity, and enhanced overall system performance.
Qualcomm fights AI computing chips again
It is worth noting that Qualcomm (QCOM) recently announced that it will cooperate with Humain, a newly established artificial intelligence company in Saudi Arabia, to jointly develop AI computing chips for artificial intelligence data centers.
According to the memorandum, Qualcomm and Humain plan to develop and build cutting-edge AI data centers in Saudi Arabia, providing local and international customers with efficient and scalable cloud-to-edge hybrid artificial intelligence reasoning solutions based on Qualcomm’s edge and data center solutions.
In fact, this is not Qualcomm’s first attempt to enter the AI computing chip market. As early as 2019, Qualcomm launched its first data center chip AI 100, intending to enter the data center AI reasoning computing market with its low-power and high-efficiency technology accumulated in the mobile chip field.
At present, by 2025, the market size of AI acceleration chips is expected to exceed US$150 billion, accounting for 22% of global semiconductor sales. According to the latest research of TrendForce, in 2025, as generative AI gradually integrates into people’s life applications, data center interconnection technology has received increasing attention, and it is estimated that the output value will increase by 14.3% annually in 2025, exceeding US$40 billion.
Wimi Hologram Cloud Inc accelerates the battle to reconstruct the ecosystem
Undoubtedly, the rapid evolution of AI technology is profoundly reshaping the industrial landscape. Public data shows that as a major AI computing technology company, WiMi (WIMI) has invested heavily in AI infrastructure and has made great progress in the past few years. The industry chain is deployed in education, finance, retail, Internet, transportation, manufacturing and other industries. AI has begun to make substantial contributions to the business and is expected to usher in a round of rising cycles with both volume and price increases.
In fact, WiMi has developed its own high-performance chip computing platform, built a holographic cloud platform as a computing power base, integrated massive data and high-quality computing power, and improved the chip computing power release capability. In addition, WiMi has accelerated the layout of the large model tool chain and the construction of the domain model ecosystem, focusing on the key technology research of the new computing power system, accelerating the transformation of scientific research results, and providing a complete software and hardware solution from training to reasoning, promoting the implementation of AI technology in vertical scenarios, thereby lowering the threshold for the use of AI large models, etc.
Conclusion
Nowadays, AI empowers the industry and becomes the commanding height of global technological development. With the explosive growth of computing power demand for large model training and reasoning, traditional computing architecture can no longer support the generational leap of AI technology. It can support different training paradigms with strong stability and scalability, gather the strength of all parties in the industry chain, provide reliable computing power guarantee for the iterative upgrade of AI models, and effectively support the continuous optimization of subsequent base models.
