Key Takeaways

  • AMD CEO Lisa Su increased on-the-ground engagement with Chinese officials and developers during a mid-May visit
  • NVIDIA CEO Jensen Huang drew larger public crowds, but his company faces sharper constraints from U.S. export rules
  • China’s AI chip market remains strategically significant despite shifting vendor access and maturing local alternatives

AMD CEO Lisa Su’s recent trip to China serves as an important marker in how U.S. semiconductor companies are recalibrating their presence in the world’s second-largest AI hardware market. The visit took place just days after NVIDIA CEO Jensen Huang wrapped up his own China stops, resulting in an unplanned contrast in style, tone, and reception.

Su kept a modest profile compared with Huang’s high-visibility walkabouts in central Beijing. That comparison has appeared repeatedly in regional coverage, although much of it focuses on optics rather than market data. Her meetings spanned an AMD developer gathering, customer discussions, and a publicly announced session with Chinese Vice Premier He Lifeng. According to reporting by the South China Morning Post, Su’s engagement with senior officials raised optimism among local companies about the continued flow of permitted U.S. AI chips. The SCMP report placed particular emphasis on Beijing’s remarks about mutually beneficial cooperation.

The geopolitical climate remains unpredictable. NVIDIA has taken the more outspoken position on U.S. export controls and has repeatedly argued that curbs may accelerate Chinese reliance on domestic AI processors. Huang highlighted at the 2025 Computex event that NVIDIA’s China market share had dropped from 95% to 50% due to earlier rounds of restrictions, a figure he recently suggested has fallen further to near zero. For AMD, the landscape looks slightly different. IDC estimates that the company holds roughly 4% of China’s AI chip market, a smaller share but one supported by a broader hardware product range.

AMD sells CPUs, consumer GPUs, AI chipsets, and FPGAs into China, giving the company a wider set of entry points as local firms build out infrastructure for generative AI. Many Chinese developers are exploring alternatives to NVIDIA’s CUDA software stack, partly because the most advanced hardware is now difficult to obtain under the U.S. Commerce Department’s BIS export control framework. Su signaled during her Shanghai developer event that AMD intends to position ROCm, its open-source suite, as a viable path forward.

Su’s trip coincided with new investment announcements in Taiwan reported by Domino Theory. Both Su and Huang are Taiwanese-born and have publicly described themselves as distant relatives. Each CEO arrived in Taipei before the upcoming Computex event, underscoring how cross-border semiconductor dynamics intertwine with regional politics.

Not every conversation has been flattering for AMD. A person familiar with its China operations told Reuters that Alibaba purchased an undisclosed number of AI chips from the company last year but required substantial engineering resources to adapt them because AMD’s software ecosystem was less mature than NVIDIA's. Alibaba did not comment, although the observation lines up with what many developers acknowledge: building up a full alternative software stack takes time and requires significant codebase adaptation.

Still, Su’s visit offered a clear signal of commitment. She noted that China accounts for about 20% of AMD’s revenue and remains an essential region for its PC, gaming, and select data center businesses. External reporting, including a recent AIWeekly note, highlighted her pledge to expand AMD’s presence in Shanghai. With AI workloads beginning to shift from large-scale model training toward broader enterprise deployments, AMD’s mix of architectures appeals to organizations seeking flexibility across compute types.

China’s generative AI market is scaling quickly, even as Beijing accelerates domestic chip development. NVIDIA currently faces a narrower channel due to its dependence on high-end accelerators. AMD’s diversified strategy creates more potential entry points, but it does not eliminate the constraints. The BIS rules continue to define the ceiling on performance levels that can be sold into China, meaning all U.S. chipmakers are navigating the exact same regulatory map.

Su’s lower-key approach aligns with current political sensitivities, according to advisers quoted in multiple reports. Executive visibility can attract unwanted scrutiny, especially as public sentiment shifts and U.S.-China relations move through alternating cycles of tension and recalibration.

In the background, China’s domestic AI chipmakers are working steadily to claim any market share left unserved by U.S. firms. As Huawei and other domestic manufacturers continue that movement, both AMD and NVIDIA are adapting their product strategies to maintain relevance while strictly complying with the export rules.