Advanced Micro Devices is executing a dual-track strategy that is as ambitious as it is geopolitically delicate. Within the span of a single week, AMD CEO Lisa Su visited Shanghai to personally court Chinese AI developers, then announced a sweeping $10 billion investment in Taiwan’s semiconductor ecosystem. Together, these moves position AMD as the most aggressive Western chip company in the race to capture the AI hardware market, both in the countries that make the chips and in the country that is most desperate to buy them.
The $10 Billion Commitment to Taiwan’s Ecosystem
The Taiwan investment, reported by CNBC and confirmed by AMD, is structured around three pillars. First, AMD is deepening its engagement with TSMC to secure capacity on the foundry’s cutting-edge 2nm process node for the upcoming EPYC Venice server chips. Second, it is expanding its advanced packaging partnerships with ASE and SPIL, two of Taiwan’s leading packaging and testing firms, to support the production of its next-generation AI accelerators. Third, the investment supports the rollout of the AMD Helios rack-scale platform, a complete AI infrastructure solution designed to compete directly with Nvidia’s flagship GB200 NVL72 systems.
The scale of this commitment reflects the strategic importance of Taiwan’s semiconductor ecosystem to AMD’s global ambitions. By locking in advanced node capacity and packaging volume, AMD is ensuring it can meet the surging global demand for AI compute—a critical bottleneck that has constrained the entire industry and given Nvidia an outsized advantage due to its earlier and deeper relationships with TSMC.
Lisa Su’s Shanghai Gambit
However, the Taiwan investment is only half of the story. Four days before the announcement, on May 19, Lisa Su was in Shanghai hosting AMD’s annual “Advancing AI” developer event, the first time the event has ever been held in China. The event drew more than 2,000 Chinese developers, a turnout that reflects the enormous pent-up demand within China for engagement with Western AI hardware companies.
Su’s message at the event was striking in its directness. “China is at the core of AMD’s roadmap,” she declared, a statement that stands in sharp contrast to the cautious, hedged language that most U.S. tech executives use when discussing their China strategies. The visit followed an earlier Shanghai appearance in which Su had predicted that there would be 5 billion daily AI users by 2030 and that computing demand would increase by 100x, as EastFrontier reported.
Filling the Void Left by Nvidia
AMD’s aggressive China posture is a direct response to the competitive vacuum created by U.S. export controls. Nvidia has largely conceded the Chinese AI chip market to domestic players like Huawei, with its most powerful systems blocked from export and its market share in China effectively reduced to zero for AI training applications. This has left a significant portion of the Chinese AI developer community—those who are not locked into Huawei’s CANN ecosystem and who prefer open-source, CUDA-compatible alternatives—without a clear hardware supplier.
AMD’s MI308 chip has received partial U.S. export licenses for China, and it has been reported that Alibaba is weighing a major order of up to 50,000 units. By promoting its ROCm software stack as an open and flexible alternative to both CUDA and CANN, AMD is positioning itself as the natural choice for this underserved segment. The company’s open-source credentials are particularly appealing to Chinese developers who are wary of proprietary lock-in, having already experienced the consequences of over-dependence on a single foreign supplier.
Navigating the Geopolitical Tightrope
AMD’s strategy is not without significant risks. The company is walking a tightrope between the U.S. government’s export control regime and its commercial ambitions in China. Any perception that AMD is helping China’s military or surveillance apparatus could trigger regulatory action that would undermine the entire China strategy. The company will need to demonstrate that its sales in China are going to legitimate commercial AI applications and that it is complying rigorously with all applicable export regulations.
Furthermore, the relationship between Taiwan and China adds another layer of complexity. AMD’s massive investment in Taiwan’s semiconductor ecosystem is taking place amid ongoing cross-strait tensions, and any deterioration in that relationship could threaten the supply chain that underpins AMD’s entire global business. For now, however, AMD appears to be betting that the commercial opportunity in China is worth the geopolitical risk, and that its dual investment in Taiwan’s manufacturing and China’s developer ecosystem will position it as an indispensable player in the global AI hardware market.
The contrast with Nvidia’s posture is instructive. While Jensen Huang has publicly acknowledged that Nvidia now holds zero percent market share in China’s AI chip market and has largely accepted this as a consequence of export policy, Lisa Su is actively pursuing the opposite strategy, seeking to maintain a commercial presence in China even as the regulatory environment tightens. Whether AMD’s more aggressive approach ultimately proves to be a competitive advantage or a regulatory liability will be one of the defining stories of the AI hardware market in the second half of 2026.