China is moving to harden its drive for artificial intelligence self-reliance after recent talks between Donald Trump and Xi Jinping, according to a new commentary from the Taiwan-based industry publication Digitimes. The analysis argues that, with little sign of relief from US technology controls, Beijing will accelerate efforts to cut reliance on foreign chips, tools, and cloud services. That would lock in a strategy already underway: build domestic compute capacity, mandate homegrown software stacks, and steer public procurement toward local suppliers. For companies operating in China, the shift points to more localisation, tighter data controls, and a more fragmented global technology market. For households and public services, it likely means steady, but measured, rollouts of AI features built on Chinese infrastructure.
The commentary was published on 20 May 2026 by Digitimes, which tracks electronics and semiconductor supply chains across Asia. It follows recent talks between Donald Trump and Xi Jinping. The article frames those discussions as a turning point for Beijing’s next phase on AI, though it does not provide detailed readouts of the meeting or confirm new policy documents.
What the new analysis says about Beijing’s next steps
Digitimes’ note argues that Beijing views the current external environment as settled for now: US export rules remain in place, allied nations continue to align on equipment restrictions, and Chinese firms face ongoing uncertainty accessing advanced chips and software. In that context, the commentary says policymakers are doubling down on domestic alternatives across the AI stack (semiconductors, accelerators, networking, compilers, frameworks, and cloud platforms) so critical systems do not depend on supplies that can be cut off.
The thrust is not new. China has pushed “secure and controllable” technology for years. What appears to be new, according to the analysis, is pace and scope. The piece points to stricter procurement guidance for state bodies and state-linked sectors, broader substitution of foreign cloud services, and an expansion of domestic model training on China-based compute clusters. The aim is to stabilize development schedules and reduce exposure to last-minute rule changes outside China.
Export controls, chips, and compute: why self-reliance matters
US controls on advanced AI accelerators and chipmaking tools, first tightened in October 2022 and updated in October 2023, limit China’s access to the highest-performance GPUs and the most advanced lithography equipment. The Netherlands and Japan introduced related limitations on specific chipmaking tools, narrowing pathways for Chinese fabs to scale at the cutting edge. These rules make it harder to train the largest general-purpose models, and they push Chinese firms to optimise for efficiency over raw scale.
In response, Chinese companies have invested in domestic accelerators and systems integration. Firms have explored China-made AI chips, network fabrics tailored to local components, and software adaptations that make training and inference viable on restricted hardware. The Digitimes commentary suggests this work will broaden, with more emphasis on standardised domestic clusters and software stacks that ministries and state-owned enterprises can deploy widely.
China’s AI rulebook: controls at home shape the market
China has also set out rules to govern how AI is built and used domestically. The Cyberspace Administration of China published measures for generative AI services in 2023, building on earlier rules for recommendation algorithms and a “deep synthesis” regime for synthetic media. Providers must register certain systems, perform security assessments, and meet content management requirements. These rules set the operating conditions for local AI services even as access to foreign cloud and models narrows.
Data transfer controls pull in the same direction. Cross-border data flows from sectors deemed sensitive continue to face security reviews. Public procurement rules in critical infrastructure often favor domestic vendors. Together, these policies reinforce one message to Chinese firms: design systems that run on local compute, use local tools where feasible, and keep sensitive data within the country.
Implications for global supply chains and regional suppliers
A push for AI self-reliance affects suppliers well beyond China. Taiwanese, South Korean, and Japanese firms that sell into China’s data center and networking markets could see shifts in demand from foreign accelerators to domestic alternatives, and from cutting-edge nodes to mature-node components. Toolmakers in Europe and Japan face ongoing licensing regimes when selling to Chinese fabs, reinforcing a bifurcated trajectory: advanced capacity clustered outside China, expanded mature capacity inside it.
For multinational cloud and software providers, a hardened self-reliance agenda points to stricter licensing and carve-outs for China-specific operations, with separate infrastructure and compliance processes. That increases cost and complexity. It also raises the chance of duplicate investments (two parallel tech stacks that meet different national rules) reducing economies of scale and complicating global service levels.
What this means for businesses, universities, and public services in China
Enterprises that handle personal data, finance, energy, health, or transport already face strong localisation requirements. If Beijing accelerates AI self-reliance, banks and insurers may need to certify that risk models run on domestic hardware and approved platforms. Hospitals and city authorities may roll out AI assistance in diagnostics, call centers, or traffic control using local clouds and vetted model providers. Performance may vary compared with services built on the latest foreign accelerators, so teams will likely lean on model pruning, distillation, and careful task design to hit service-level targets.
Universities and research labs that rely on large model training will continue to adapt their research agendas. Expect more work on efficient training, edge inference, and domain-specific systems tuned to Chinese datasets and computing constraints. That could produce robust tools for practical tasks; translation, document processing, customer support, without chasing maximum model size.
Energy, data centers, and the cost of domestic capacity
Building AI capacity at home requires land, power, and cooling. Data center growth strains local grids and water supplies in some regions. Chinese planners have promoted “east-data-west-compute” projects to site energy-hungry facilities in areas with more space and renewable potential, then move results over high-speed networks to demand centers. A faster self-reliance push would likely extend those programs and raise the importance of grid upgrades and efficiency standards for servers and cooling.
Costs remain a key variable. Domestic accelerators and mature-node chips can lower exposure to export shocks, but total cost of ownership depends on performance, software maturity, and supply stability. Companies will watch whether standardised domestic clusters can deliver predictable throughput for training and inference. If they can, budgeting and rollout timelines for AI features in public services and consumer apps become easier to manage.
The policy signals to watch in the coming months
The clearest signs of a hardened agenda will come through official documents: procurement catalogues that list approved chips and platforms, guidance to financial and industrial regulators on acceptable AI stacks, and funding allocations for data center build-outs and semiconductor tooling. Further adjustments to US and allied export regimes would also shape the pace and direction of China’s plans.
Multinationals with China operations will monitor compliance updates on cross-border data, source-code disclosures, and security reviews for AI services. Local developers will track model registration processes and any changes in content requirements. Supply chain firms will look for signals on mature-node equipment orders, packaging capacity, and domestic accelerator roadmaps.
China’s AI strategy has long aimed at reducing dependence on foreign technology. The Digitimes commentary suggests that, after the recent Trump–Xi talks, Beijing sees little reason to wait for external relief. For businesses in and around China, the practical takeaway is straightforward: plan for more localization, more regulatory checkpoints, and more parallel infrastructure. For everyday users and public services, expect steady progress in AI-enabled features built on Chinese stacks, rather than sudden leaps. The balance between capability, cost, and control will define how fast those systems reach scale, and how far the global market continues to split into separate, self-reinforcing technology spheres.
