(Image source: SCMP)
AsianFin — The global semiconductor race intensified this week with major developments on both sides of the Pacific, underscoring a deepening rivalry between China and the United States in AI chips, mobile processors, and technological sovereignty.
In China, Xiaomi Group CEO Lei Jun announced on May 15 that the company’s long-awaited, self-developed smartphone chip—the Xuanjie O1 (Xring O1) SoC—will officially launch in late May.
The chip will debut in the new Xiaomi 15S Pro, and according to Group President Lu Weibing, it won’t be limited to smartphones. The Xuanjie O1 will also power other devices in Xiaomi’s ecosystem, signaling broader ambitions in AIoT and edge computing.
The announcement caps a decade-long journey that began with Xiaomi’s 2014 launch of its chip subsidiary Pinecone Electronics, followed by the 2017 debut of the mid-range Surge S1. While that first chip faltered due to baseband limitations, Xiaomi has since restructured and re-entered the field through its subsidiary Xuanjie Technology, now boasting over 1,000 R&D engineers and backing from venture arms Shunwei Capital and Xiaomi Industrial Investment.
In a parallel move, Lenovo has also stepped onto center stage. Its new YOGA Pad Pro 14.5 tablet features China’s first 5nm domestically designed SoC, the SS1101, developed by Lenovo’s chip subsidiary Dingdao Zhixin. The chip marks a key milestone as China races to localize its semiconductor supply chain amid mounting geopolitical pressure.
Meanwhile, in the United States, chip giants Nvidia, AMD, and Qualcomm announced multi-billion-dollar partnerships with HUMAIN, a new Saudi Arabian state-owned AI enterprise, during former President Donald Trump’s recent visit to the kingdom. In return, Saudi Arabia will invest $20 billion into U.S.-based AI data centers, bolstering America’s edge in AI infrastructure.
However, Nvidia’s stance on China is shifting. On May 18, CEO Jensen Huang acknowledged that U.S. government export restrictions on the Hopper-architecture H20 chips are forcing the company to rethink its China strategy. Huang confirmed Nvidia will no longer release new Hopper-based products in China, noting "It won’t be Hopper, because Hopper can no longer be adjusted." The company previously estimated that export controls could result in $5.5 billion in lost revenue.
Huang is expected to outline Nvidia’s forward-looking roadmap—potentially including developments in embodied intelligence software—in his Computex 2025 keynote address on May 19.
The tit-for-tat dynamic in chip development reflects a broader strategic decoupling between the U.S. and China. While Washington tightens AI chip restrictions to constrain China's tech rise, Chinese companies like Xiaomi and Lenovo are forging ahead with homegrown silicon, despite high costs, IP challenges, and talent shortages.
An Economic Daily article published on May 19 called for China to not only invest in foundational chip technologies but also build application ecosystem dominance and establish governance standards in emerging computing fields.
Developing cutting-edge chips is an uphill battle. The cost to develop a 28nm chip is estimated at $40 million; that number balloons to $217 million for 7nm, and $416 million for 5nm—with 3nm development nearing $1 billion.
