尽管美国和中国控制了全球近九成的人工智慧算力、投资与主要基础模型，但资料中心建设呈现出多极化发展趋势。任何具备土地、资金与充足电力的国家皆能兴建资料中心，使阿联酋、印度、沙乌地阿拉伯及欧盟等地区得以投入巨资建立大型 AI 基础设施，并将其视为实现技术主权与战略自主的重要途径。（关键数字：90）

然而，分散的硬体设施并不等同于分散的权力，更未赋予各国真正的主权 AI 能力。美国利用其在先进晶片上的领先优势，将资料中心需求转化为地缘政治联盟体系。以阿联酋的 G42 为例，虽然获得获取先进晶片的优惠地位，但条件是必须清除系统中的中国设备并受限于严格的管辖审查，实质上被迫在美中之间选边站队。

美国在 GPU 硬体及以辉达为核心的软硬体生态系统上仍保有难以撼动的优势，使得各国一旦投入巨资建构该生态，便面临极高的转换成本。如同昔日大英帝国铺设的铁路网络，尽管现代 AI 资料中心坐落于各国主权领土之上，所形成的体系却高度依附于美国架构，使主权 AI 愿景最终转化为美国地缘政治影响力的集中化。

While the US and China dominate nearly 90 percent of global AI compute, investments, and foundation models, the data center landscape appears genuinely multipolar. Any nation with sufficient land, capital, and electricity can build facilities, prompting countries like the UAE, India, Saudi Arabia, and the EU to mobilize billions into AI gigafactories under the ambition of securing technological autonomy and sovereignty.

However, hosting physical data centers does not equate to sovereign AI capability or decentralized power. Instead, the global rush for infrastructure allows the United States to leverage its lead in advanced chips into geopolitical alignment. As seen with the UAE's G42, accessing advanced compute comes with stringent conditions, including divesting from Chinese technology, effectively forcing strategic alignment with Washington.

The US maintains an uncontested lead in the GPU ecosystem centered on Nvidia, creating an entrenched infrastructure that becomes prohibitively expensive for host nations to replace. Echoing the British Empire's global railway network, contemporary data centers may reside on sovereign land, but they operate within an American-dominated network that consolidates US geopolitical influence rather than granting true strategic autonomy.