Apr 14 (East Asia Forum) - As Meta fine-tunes its Llama artificial intelligence (AI) models on global data and OpenAI deepens its enterprise integration, Asia’s middle powers face mounting pressure to shape their own technological futures.
Regionally influential countries like Australia, South Korea, India, Indonesia and Singapore — that possess important AI capabilities but lack the ability to shape AI system development and deployment at the scale of the United States or China — risk becoming permanent renters of external infrastructure, confined to application-layer innovation and vulnerable to shifting geopolitical winds.
In February 2025, French President Emmanuel Macron floated the idea of an ‘AI third way’ — a multilateral path for middle powers to cooperate on AI development and infrastructure, rather than defaulting to systems built in Silicon Valley or Shenzhen. His appeal echoes a growing recognition across Asia that the global AI landscape is consolidating around two poles, leaving limited room for strategic autonomy or ensuring public redistribution AI’s benefits.
Breakthroughs in open-source technologies by AI companies like China’s DeepSeek and France’s Mistral highlight the technical potential for alternatives. But these successes depend on preexisting open-source ecosystems and face limitations in long-term product scaling and distribution without supportive infrastructure.
A viable AI third way in Asia will require strategic collaboration across national AI efforts to produce not only AI models but to integrate those models with competitive AI products.
National AI strategies are advancing across the region, but most cannot compete at the frontier of product development. India is actively integrating AI into its digital public infrastructure stack through focused efforts like BharatGPT, a conversational AI platform tailored to India’s linguistic diversity. In February 2025, Indonesia announced a sovereign wealth fund for mineral processing, AI and renewable energy, but the country lacks domestic compute capacity and remains reliant on foreign AI platforms and applications. Singapore has emerged as a regional leader in public–private AI partnerships and Australia is advancing promising data governance and safety frameworks.
These national efforts cannot individually match the scale of compute, talent or high-quality data of US or Chinese AI ecosystems. Worse, they risk reinforcing a dependency cycle in which public sector investments in open infrastructure are captured by private hyperscalers with the capacity to productise and monetise faster.
To avoid this trajectory, middle powers need a common technical and policy framework to ensure interoperability between sovereign AI assets. This should include cross-border data sharing protocols, approaches to sharing model weights and, crucially, joint product development.
In an era of open-source systems, easy distillation and rapidly advancing capabilities, it is not only the models themselves, but also the iterative data generated through deployment of applications, that matter. Whoever controls these feedback loops holds the leverage.
One emerging proposal gaining traction is to create an ‘Airbus for AI’, modelled on the European response to US aviation dominance.
First proposed in 2024 by the Barcelona Supercomputing Center, the idea is to create a public–private consortium of middle powers’ national AI laboratories to build scalable AI products and sell them under a public utility model. The proposal is a go-to-market strategy for middle powers’ existing public AI investments, designed to serve public needs while anchoring technological sovereignty. Just as Airbus gave Europe a seat at the table in commercial aviation, an AI consortium could offer Asian nations an opportunity to establish industrial capacity and strategic autonomy.
The consortium would aggregate compute from participating national labs and codevelop applications built on shared open-source or jointly trained foundation models. A dedicated team sourced from participating nations — combining engineering, policy and product expertise — would scale these applications across jurisdictions, ensuring commercial viability, responsiveness to local needs and alignment with public interests.
Initially the focus would be government services — ID verification, document processing and healthcare triage — where national data assets are rich and deployment needs are urgent. Over time, this could expand into digital utilities for civil society, small enterprises and country-specific industries, providing a counterbalance to proprietary offerings from US and Chinese hyperscalers. These deployments would help establish the critical feedback loops that many public AI initiatives currently lack and demonstrate that public value and performance can go hand in hand.
Beyond the need to pool compute, the consortium’s success will rely on three enablers — operational coherence, a coordinated regional data policy framework and talent.
First, international coordination requires more than aligned policy declarations — it also demands operational coherence. Lessons from initiatives like intergovernmental physics laboratory CERN, chip manufacturing consortium SEMATECH, EU global navigation satellite system Galileo, US Operation Warp Speed and Airbus underscore the importance of clear incentives, product focus and intellectual property sharing mechanisms.
Second, a coordinated regional data policy framework will be essential to ensure sufficient high-quality training data while upholding privacy and security standards. Countries’ digital public infrastructure systems — including ID systems, payment platforms and data exchanges — can form the foundation of equitable data ecosystems. Cross-border interoperability will rely on the development of trusted governance and shared protocols.
Third, securing top AI talent remains a global challenge. The consortium could serve as a case for piloting more open regional labor and migration rules. It will also need to offer competitive compensation and highlight opportunities to work on mission-driven applications as well as the prestige of building a regional alternative to AI superpowers.
A consortium can build on existing strengths across the region. For example, Australia’s National AI Centre could anchor safety protocols and Singapore’s AI Verify Foundation offers a testbed for model auditing. The India Enterprise Architecture offers a framework for connecting between digital public infrastructure and AI-driven public service applications and South Korea’s semiconductor strategy could be leveraged for chip-aligned model development. ASEAN’s Digital Economy Framework and APEC’s 2025 focus on regional integration of national AI efforts serve as key fora for regional policy coordination.
Airbus succeeded because Europe had the collective vision to build strategic capability in a field dominated by others. If Asia’s middle powers are serious about forging a third path in AI, they must move beyond national scale strategies and take regional action through joint engineering, shared infrastructure and coordinated product development.
Jacob Taylor is Fellow at the Brookings Institution’s Center for Sustainable Development.
Joshua Tan is Co-founder and Research Director at Metagov.
Source: East Asia Forum
