Huawei is rapidly rolling out a new generation of semiconductor products tailored for artificial intelligence (AI), signaling a bold push to fill the void left by tightening US export restrictions on advanced chips. The company is set to release its Ascend 910C chip in May, followed by the introduction of the Ascend 920 in the latter half of 2025. This development occurs against the backdrop of US export restrictions, which were implemented during the Trump administration and have limited Nvidia's capacity to provide high-performance chips to China.
Ascend 910C: A domestic challenger to Nvidia's H100
Reuters reports that the Ascend 910C, which is scheduled for mass production and shipment in May, offers twice the computing power and memory capacity of its predecessor. Its performance is said to rival that of Nvidia's restricted H100 AI chip. Nvidia's China-specific H20 chip, which has reduced performance capabilities, has also been subject to US export bans, fueling market speculation that Huawei's products could challenge Nvidia's market dominance.
Nvidia's regulatory filings indicate an expected loss of US$5.5 billion in the company's fiscal first quarter of 2026, ending April 27, as a result of these restrictions. During a recent visit to China, Nvidia CEO Jensen Huang reiterated the company's commitment to the Chinese market, though analysts caution that increasing US policy constraints and competition from Chinese firms such as Huawei present growing obstacles.
Nvidia's strategic pivot to Japan
Confronted with challenges in the Chinese market, Nvidia is shifting its strategic attention to Japan, supported by the country's strong research and development capabilities and government support. On April 21, 2025, Huang met with Japanese Prime Minister Shigeru Ishiba, advocating for increased power generation to address the rapidly growing energy demands of artificial intelligence technologies.
According to the International Energy Agency, global electricity consumption by data centers, AI, and cryptocurrencies is projected to reach 1,050 terawatt-hours by 2026, roughly equivalent to Germany's yearly electricity usage, with AI applications expected to drive the most significant increase. Huang highlighted the necessity of a reliable energy infrastructure to maintain leadership in AI, suggesting the implementation of smart grids and AI-driven energy distribution systems to effectively manage peak loads in data centers.
Huang's discussions with Ishiba, as reported by The Nikkei, framed AI as a transformative force for Japanese industries, from healthcare and manufacturing to education and agriculture. He predicted that AI robotics would "revolutionize" Japanese enterprises, creating jobs and opportunities through energy-centric infrastructure investments. In a symbolic gesture, Huang gifted Ishiba a DGX Spark AI supercomputer, underscoring Nvidia's support for Japan's AI aspirations. Ishiba pledged to align with tech giants to position Japan as a global AI hub.
Credit: Prime Minister's Office of Japan
Strategic alliances and infrastructure expansion
Nvidia's Asian strategy hinges on Japan, where it is collaborating with SoftBank Group to build a world-class AI supercomputer powered by Nvidia's cutting-edge Blackwell chips. SoftBank, in partnership with OpenAI, is also developing a major data center in Japan to fuel AI-driven enterprise solutions. Huang's visit included planned talks with SoftBank CEO Masayoshi Son to advance these initiatives.
Additionally, Nvidia's B300/GB300 AI servers, showcased at the March GTC conference, are set for large-scale shipments by September, catalyzing growth for Japanese and Taiwanese supply chains.
Credit: Prime Minister's Office of Japan
Japan's broader AI ambitions
Japan's AI push gained momentum with Prime Minister Ishiba's April 17, 2025, meeting with AMD CEO Lisa Su, where he articulated a vision for Japan as the world's most AI-friendly nation. By fostering AI as a cornerstone industry, Japan aims to leverage partnerships with global tech leaders like Nvidia and AMD to build cutting-edge infrastructure and applications, from autonomous driving to medical imaging.
Article edited by Jerry Chen
