China’s AI ambitions face a capital constraint

In China’s 15th Five-Year Plan, artificial intelligence (AI) is the most frequently mentioned strategic priority, while ‘involution’ and weak domestic demand also feature prominently. Yet financial market liberalisation — an issue underlying each of these priorities — receives far less attention.

China is not liberalising its financial markets. Rather, Beijing is selectively internationalising the renminbi (RMB) to allow for international circulation where this is strategically useful, while preserving capital controls. This creates a system in which firms can raise capital at home but face political and regulatory barriers when attempting to scale abroad, despite this typically being central to competitiveness in emerging industries. Beijing’s strict control over capital flows is increasingly constraining Chinese AI firms to a lower-return domestic market, where investment and competition are growing faster than profitable opportunities.

Recent internationalisation initiatives are not steps towards opening the capital account. HKEX, the Hong Kong stock exchange operator, has expanded its HKD–RMB Dual Counter Model, which allows investors to trade selected Hong Kong-listed securities in either Hong Kong dollars or RMB. There has also been the use of RMB for toll payments through the Strait of Hormuz and a surge in ‘panda bonds’, RMB-denominated bonds sold in China but issued by offshore entities.

But these initiatives are not traditional liberalising mechanisms. Each expands transactional use of the RMB internationally while preserving state control. Because the RMB still does not function as a fully convertible investment currency, its role in market-driven global capital allocation remains limited.

While selective RMB internationalisation is driven by multiple objectives — facilitating cross-border transactions with China, preserving state control over the financial system and increasing financial resilience to sanctions — the resulting constraints on outward capital mobility may limit Chinese AI firms’ ability to access and deploy capital globally. For investors, this asymmetry introduces uncertainty about entry, exit, pricing and timing.

State-backed investors, including new venture capital organisations, have stepped in to support domestic-led growth of technology companies as foreign investment in China’s domestic economy diminishes, in part due to unclear future expectations about regulation and economic performance. AI firms are dependent on large, patient, globally mobile capital and compute access, both of which are harder to obtain inside China’s controlled financial system.

But attempts to bypass Beijing’s capital constraints have drawn swift regulatory scrutiny, as in the case of Manus AI — which also received state-backed venture capital funding. After relocating operations to Singapore and agreeing to a US$2 billion acquisition by Meta Platforms in late 2025, its founders were reportedly barred from leaving China during a regulatory review. By January 2026, China’s Ministry of Commerce had issued a vague statement launching an investigation into the organisation’s outbound economic activity, offering no specific justification.

Then, in late April 2026, Chinese regulators reportedly ordered the acquisition to be unwound, signalling that offshore restructuring may not shield strategically sensitive AI firms from Beijing’s jurisdiction. For founders and investors, concerns about access to capital and regulatory caprice alter their incentives at a fundamental level. If acquisitions, offshore listings or cross-border restructuring can be blocked retroactively on political grounds, investors face greater uncertainty about possible exit pathways and timing. In sectors such as AI that depend on large amounts of long-term capital, such uncertainty can reduce the expected value of investment.

Compounding these constraints, US export controls have limited Chinese firms’ access to advanced chips and semiconductor manufacturing equipment. This pushes these firms towards compute efficiency and lower-cost deployment rather than simply scaling frontier training runs. That adaptation aligns with a broader Chinese AI strategy focused on developing AI systems that are cheaper to operate and easier to roll out, particularly across emerging markets and the Global South, where Beijing has explicitly sought to expand AI capacity.

But capital constraints and regulatory pressures may also push Chinese AI firms towards applications and deployment strategies that fit domestic political and economic conditions rather than those optimised for global scale. Over time, this may create growing divergence not only in market access, but in the commercial ecosystems, technical standards and acquisition pathways needed to establish leadership in AI.

The narrowing performance gap between leading US and Chinese large language models suggests that access to talent, compute and even state-backed capital alone may not determine who leads the next phase of AI competition.

In consumer AI, where success depends on acquisitions, cross-border investment, developer ecosystems and rapid international deployment, firms with both access to capital and the freedom to deploy it may hold the longer-term advantage. China’s financial model may continue to support AI innovation, but by constraining how capital can exit, scale and integrate globally, it may also restrict the country to a narrower path to AI leadership.

Hadley Spadaccini is Fulbright Scholar at National Tsing Hua University.

This work was supported by the Fulbright Program.