NVIDIA is accelerating the development of the next-generation AI chip 'Rubin' - Taiwan Economic Daily News reported.
On the 4th, Taiwan's Economic Daily reported that NVIDIA $NVDA is accelerating the development schedule of the next-generation AI chip "Rubin platform." According to the report, the product launch originally planned for 2026 will be brought forward to the second half of 2025.
1️⃣ differentiation through the adoption of advanced technology
According to the Economic Daily News, semiconductor analyst from Morgan Stanley, Mr. Chan Chia-Hung, has analyzed the Rubin platform as the next-generation AI chip that adopts a 3-nanometer manufacturing process, with features such as Common Photonics Interface (CPO) and HBM4 memory technology. The size of this chip is double that of the current Blackwell platform, characterized by a design that integrates four computing chips.
2️⃣ deployment of supply chains by various companies
According to reports, TSMC $TSM is advancing plans to expand the monthly production capacity of the advanced packaging technology CoWoS (Chip on Wafer on Substrate) to approximately 0.08 million units by the fourth quarter of 2025.
Chairman Liu Yangwei of Hon Hai (Foxconn) stated that the AI server business is expected to account for over 50% of the company's total server revenue in 2025, aiming for a market share of over 40% globally. Deputy General Manager Yang Qiling of Quanta Computer mentioned that the GB200 server will enter mass production by the end of the year and the shipment volume will expand from the first quarter of next year. He predicts that in the next generation AI servers, the product structure will become more complex, and the average selling price (ASP) will increase 2 to 3 times.
📍 Future Outlook
The advancement of NVIDIA's next-generation AI chip "Rubin platform" is likely to be a significant turning point in the semiconductor industry. The decision to bring forward the originally scheduled product launch in 2026 to the second half of 2025 reflects an anticipation of increased AI demand exceeding expectations and a sense of crisis regarding competition from rival companies.
Of particular note is the technical challenge of integrating the 3nm process, Common Photonic Interface (CPO), and HBM4 memory technology. This innovative approach is reflected in the ambitious design of a Blackwell platform with twice the chip size and the integration of four calculation chips.
Such technological innovations will not only accelerate the momentum of TSMC $TSM, but also the HPC forces in Taiwan. TSMC's CoWoS technology with a monthly production capacity of 0.08 million units, Foxconn's AI server world market share target of 40%, and Quanta's doubling forecast for server ASP, vividly demonstrate this change.
From an investment perspective, attention is focused on companies involved in semiconductor manufacturing equipment and advanced packaging technologies. However, it is necessary to be mindful of the geopolitical risks in the Taiwan Strait, yield issues with the 3nm process, and fluctuations in the AI investment cycle.
1️⃣ differentiation through the adoption of advanced technology
According to the Economic Daily News, semiconductor analyst from Morgan Stanley, Mr. Chan Chia-Hung, has analyzed the Rubin platform as the next-generation AI chip that adopts a 3-nanometer manufacturing process, with features such as Common Photonics Interface (CPO) and HBM4 memory technology. The size of this chip is double that of the current Blackwell platform, characterized by a design that integrates four computing chips.
2️⃣ deployment of supply chains by various companies
According to reports, TSMC $TSM is advancing plans to expand the monthly production capacity of the advanced packaging technology CoWoS (Chip on Wafer on Substrate) to approximately 0.08 million units by the fourth quarter of 2025.
Chairman Liu Yangwei of Hon Hai (Foxconn) stated that the AI server business is expected to account for over 50% of the company's total server revenue in 2025, aiming for a market share of over 40% globally. Deputy General Manager Yang Qiling of Quanta Computer mentioned that the GB200 server will enter mass production by the end of the year and the shipment volume will expand from the first quarter of next year. He predicts that in the next generation AI servers, the product structure will become more complex, and the average selling price (ASP) will increase 2 to 3 times.
📍 Future Outlook
The advancement of NVIDIA's next-generation AI chip "Rubin platform" is likely to be a significant turning point in the semiconductor industry. The decision to bring forward the originally scheduled product launch in 2026 to the second half of 2025 reflects an anticipation of increased AI demand exceeding expectations and a sense of crisis regarding competition from rival companies.
Of particular note is the technical challenge of integrating the 3nm process, Common Photonic Interface (CPO), and HBM4 memory technology. This innovative approach is reflected in the ambitious design of a Blackwell platform with twice the chip size and the integration of four calculation chips.
Such technological innovations will not only accelerate the momentum of TSMC $TSM, but also the HPC forces in Taiwan. TSMC's CoWoS technology with a monthly production capacity of 0.08 million units, Foxconn's AI server world market share target of 40%, and Quanta's doubling forecast for server ASP, vividly demonstrate this change.
From an investment perspective, attention is focused on companies involved in semiconductor manufacturing equipment and advanced packaging technologies. However, it is necessary to be mindful of the geopolitical risks in the Taiwan Strait, yield issues with the 3nm process, and fluctuations in the AI investment cycle.
参考
Disclaimer: Community is offered by Moomoo Technologies Inc. and is for educational purposes only.
Read more
Comment
Sign in to post a comment
Jamaica no problem : Mr. klmihiko,
. I am excited about nvidia's unparalleled dominance and look forward to their continued innovation. It's amazing how they are evolving at lightning speed 
despite just introducing Blackwell. I will hold on to not be left behind 
.
Thank you for the prompt and valuable information