Friday, September 11, 2026
AI 인프라 · 뉴스 & 분석
정책리포트
정책 · 리포트

글로벌 AI 개발이 냉전 시대의 양상으로 분열되는 것을 경계하며, 아시아 국가들이 개방형 기술 생태계를 유지할 것을 촉구하는 논평이 나왔다.

국경 간 반도체 공급망 및 국제 데이터센터 자금 조달을 방해할 수 있는 규제 및 수출 통제 위험을 강조합니다.
업계 전문지Slicast · September 5, 2026 · 미국 · 출처: ndtvprofit.com
중요도 65

The global landscape continues to fracture, and artificial intelligence is now dividing along geopolitical lines long before it can generate sustainable profits. The fault line is straightforward: China dominates resources, while the West controls technology. Beijing processes the majority of the rare earths, metals, and minerals essential for AI infrastructure, leaving few viable alternatives. Conversely, Washington and its allies lead in semiconductor chips, lithography equipment, and foundational AI models.

A second divide exists within the AI models themselves. Leading American systems from Google, OpenAI, and Anthropic remain proprietary, with developers withholding details on training data and internal architectures. Nvidia’s Nemotron stands as a notable exception. In stark contrast, nearly all leading Chinese models are open-source. This creates a striking asymmetry: China seeks to monopolize raw materials and supply chains while distributing its models openly, whereas the West controls advanced technology but restricts access to its models. This divergence has become a central battleground in the emerging AI conflict.

China is consolidating its grip on critical minerals through tightened export controls and enhanced screening of dual-use goods. Exporters now face stricter approval requirements before shipping certain materials abroad. Indian companies have experienced these restrictions firsthand. Chinese suppliers have frequently withheld key technologies and components from Indian firms, causing delays for electric vehicle and electronics manufacturers. Additionally, Chinese experts have been barred from training workers at Foxconn facilities in India. Beijing is leveraging its material dominance to protect its strategic lead, mirroring Western tactics further down the supply chain.

The United States and its allies are actively curbing technology transfers to China. Sales and servicing of ASML’s lithography machines—essential for semiconductor manufacturing—are prohibited in China. Washington has also banned exports of high-end AI chips, including Nvidia’s A100 and H100, while restricting American investment in sensitive Chinese sectors. However, a structural challenge remains: the West cannot rebuild its supply chains as rapidly or cost-effectively as China did. To bridge this gap, Washington has launched several strategic initiatives.

Project Vault aims to stockpile critical minerals vital for AI development. The initiative plans to secure reserves of approximately 60 minerals, functioning similarly to the Strategic Petroleum Reserve for oil. US manufacturers would gain access to these stocks during supply disruptions. The minerals will be housed in domestic facilities managed by VaultCo, a public-private partnership, with up to $10 billion in financing backed by the US Export-Import Bank. The plan faces criticism, however. Veronique de Rugy of the Mercatus Centre warns it could devolve into government-backed risk management for a select few corporations. Skeptics also question whether mere stockpiling sufficiently reduces America’s reliance on foreign supply chains.

While Project Vault provides a short-term buffer, expanding mining and processing capacity requires sustained coordination. The Forum on Resource Geostrategic Engagement (FORGE) addresses this need. Its mandate includes identifying priority mining and processing projects and aligning policies, standards, export controls, and pricing frameworks across participating nations. FORGE also promotes long-term purchase agreements to mitigate financial risks. Such projects demand substantial upfront capital, and minor price fluctuations or sudden project halts often deter private lenders. Public finance has therefore been deployed to de-risk these early-stage investments.

MINVEST, or the Minerals Investment Network for Vital Energy Security and Transition, complements FORGE by structuring the financing for identified projects. It convenes private investors, mining firms, manufacturers, export-credit agencies, and governments to fund critical mineral ventures. Fifty-five countries were represented at its inaugural Critical Minerals Ministerial. India has participated in the broader ministerial dialogue but has not formally joined FORGE.

Pax Silica seeks to unify the entire AI supply chain under a US-led coalition. The framework encompasses everything from critical minerals and advanced materials to semiconductors, data centers, energy, logistics, AI models, and finance. Its primary objective is reducing dependence on China by pooling member strengths: US chip design, Dutch lithography, Japanese chemicals, Korean memory chips, and Australian minerals. Through coordinated investment and shared infrastructure, members can access capabilities they lack domestically. The coalition currently comprises 24 member nations, including India. Together, these initiatives address distinct vulnerabilities: Vault establishes a strategic buffer, FORGE coordinates resource development, MINVEST secures financing, and Pax Silica links upstream materials to downstream AI technologies. In essence, the West is constructing a mine-to-model alliance designed for mutual resilience.

China is pursuing a parallel strategy by assembling an international coalition centered on Chinese AI models, infrastructure, and standards, particularly targeting the Global South. Known as the World Artificial Intelligence Cooperation Organisation (WAICO), the group includes 29 nations, such as Russia, Brazil, Kazakhstan, Indonesia, Malaysia, Pakistan, and several African and Central Asian states. No G7 country has joined. Many governments prioritize affordability and access to computing power, infrastructure, skilled talent, and localized language data. Advanced US models carry steep costs, whereas Chinese platforms like Kimi, Qwen, and DeepSeek offer relatively affordable deployment options. Beijing has also pledged approximately 5,000 AI training opportunities, establishing an alternative channel for influence. WAICO does not explicitly exclude countries aligned with the US or its allies, and Chinese commentators insist it should not be framed as an anti-American bloc. The sustainability of this inclusive posture remains uncertain.

Southeast Asia has emerged as the clearest test of whether the world will fracture into two rigid camps. Thus far, regional nations have resisted alignment. Five ASEAN members—Indonesia, Malaysia, Cambodia, Laos, and Myanmar—are founding WAICO participants, while Singapore and the Philippines are engaged with Pax Silica. Membership has not yet forced policy shifts. Indonesia, despite joining WAICO, maintains that national interests will always take precedence. Malaysia relies on Chinese raw materials and supply chains while maintaining deep ties to Western cloud and semiconductor ecosystems; it is simultaneously developing sovereign AI cloud infrastructure. Singapore has translated its traditional geopolitical neutrality into an interoperability approach to AI, supported by its own governance framework. Regional governments recognize the necessity of engaging both blocs but prioritize policy autonomy to navigate competing demands.

Kazakhstan exemplifies the challenges of navigating both coalitions. As the sole nation to join both WAICO and Pax Silica, it leverages its substantial reserves of uranium, tungsten, copper, and rare earth elements. WAICO provides access to Chinese AI cooperation and open-weight models, while Pax Silica offers integration into an advanced technological network—a seemingly optimal strategy. However, Reuters reported that the US may caution against dual participation, which could complicate Astana’s position. Should pressure mount, Kazakhstan will face a difficult choice between competing alliances.

India’s stance presents a different set of complexities. Geopolitically, India's position is non-aligned or

원문 보기