PERSPECTIVES | What you shouldn’t miss in the Trump-Xi summit: The bargain over AI dominance

This article appeared in UTS:ACRI's Perspectives on May 12 2026. Perspectives is the commentary series of the Australia-China Relations Institute at the University of Technology Sydney (UTS:ACRI), offering research-informed viewpoints on developments and debates in the Australia-China relationship.

Weeks before US President Donald Trump’s May 13-15 summit with People’s Republic of China (PRC) President Xi Jinping in Beijing, the technological battlefield was already shifting.

On April 24, the PRC’s leading AI firm, DeepSeek, released V4, underscoring Beijing’s push to build a self-sufficient AI stack, from models and software to domestic chips. Days later, the PRC’s National Development and Reform Commission blocked Meta’s proposed acquisition of Manus, a PRC-originated AI agent firm that had moved its headquarters and core team to Singapore, though its early codebase was developed in Beijing and Wuhan. Then, on April 30, the US Federal Communications Commission tightened compliance restrictions on electronics-testing laboratories in mainland China and Hong Kong, drawing sharp protests from Beijing.

These moves show Washington and Beijing laying down markers in a contest to shape and ultimately control the next generation of AI ecosystems.

The coming leaders’ summit will inevitably feature familiar disputes over Taiwan, tariffs and trade deficits. Yet the most consequential bargaining may revolve around the less immediately visible – chips, AI models, supply chains, standards and control over the technological infrastructure that will underpin economic and military power for decades to come.

For years, the prevailing assumption in Washington was that the US would maintain a substantial and durable lead in AI. Silicon Valley had access to the world’s most advanced chips at scale, the deepest pools of venture capital and the densest concentration of frontier AI laboratories. The PRC, despite its large domestic market and significant state support, was generally expected to lag behind in large language models.

That assumption is now under strain.

The PRC’s progress in AI over the past five years has extended beyond consumer applications to research capability, computing infrastructure and ecosystem building, especially in industrial AI, including humanoid robotics. Last year, DeepSeek challenged assumptions in Western policy and technology circles by demonstrating that Chinese firms could still produce highly capable open-source models despite severe restrictions on access to advanced chips. Its latest release, DeepSeek‑V4, may prove even more significant. The model reportedly integrates closely with Chinese chipmakers such as Huawei and Cambricon, tying domestic semiconductors more directly to the PRC’s AI-development stack.

This matters because the future AI race will not simply be about who builds the best chatbot, but on the development of rival technological ecosystems.

The US model is vertically integrated, proprietary and capital-intensive, relying heavily on NVIDIA chips, hyperscale cloud providers and closed systems controlled by firms such as OpenAI, Anthropic and Google DeepMind. The PRC model is taking shape along different lines, emphasising open-source development, cost-efficient models, industrial deployment and tight alignment with state industrial policy.

The Manus episode illustrates how strategic this competition has become. Manus is an AI agent platform, a category of systems designed to translate human instructions into machine-executable tasks. Unlike conventional chatbots, AI agents are intended to perform actions rather than simply generate responses.

That is precisely why Beijing intervened.

PRC authorities reportedly reviewed the deal through several regulatory frameworks, including foreign investment security screening, technology export controls and cross-border data regulations. Officials ultimately concluded that Manus’ agentic capabilities, particularly its semantic-recognition functions, fell within sensitive ‘key technology’ categories requiring prior approval for any outbound transfer.

The broader signal was clear. Beijing will not easily allow its most promising AI firms, engineers and core technologies to migrate offshore or be absorbed into the US-led technology stack. Even attempts to internationalise firms through Singapore-based restructuring may no longer offer protection from the PRC’s increasingly assertive interpretation of technological sovereignty.

The decision also carried a domestic message. Chinese entrepreneurs are being told that the era of freely arbitraging between Chinese talent and US capital is ending.

Washington, meanwhile, is sending its own signals. US export controls on advanced semiconductors were designed to slow the PRC’s AI development by limiting access to cutting-edge computing power. The results, however, have been mixed. The controls have increased costs for Chinese firms and constrained access to frontier hardware but they have also accelerated Beijing’s efforts to develop domestic alternatives.

The irony is becoming increasingly obvious. The harder the US tries to preserve technological dominance through restriction, the stronger the PRC’s incentives become to build an independent ecosystem.

This dynamic now extends beyond chips. The US dominates frontier graphics processing units (GPUs) and cloud infrastructure. The PRC dominates significant portions of critical mineral supply chains and controls much of the world’s rare earth processing capacity. Both sides increasingly view these dependencies as sources of bargaining leverage. Trump may view advanced NVIDIA chips as a negotiating card, especially as the PRC’s domestic alternatives continue to mature. Beijing, meanwhile, understands that rare earths remain indispensable for advanced manufacturing, defence systems and clean-energy technologies.

This is the paradox of weaponised interdependence: each coercive action gives the other side stronger incentives to reduce its vulnerability, thereby weakening the coercive power of the original move. The result is an increasingly fragmented global technology order.

That fragmentation carries substantial costs for countries and firms outside the two major powers. If rival AI ecosystems diverge further, companies may eventually need to comply with two sets of technical standards, governance rules and software architectures. Countries such as Australia could find themselves forced to operate simultaneously within US and PRC technological stacks.

At the same time, full technological decoupling remains unlikely. Despite rapid progress, the PRC still depends on foreign lithography equipment and remains behind Taiwan and South Korea at the frontier of semiconductor manufacturing. The US, meanwhile, cannot entirely insulate itself from PRC supply chains, markets and industrial capacity.

This is what makes the Trump-Xi summit so important. AI is no longer simply one issue within the broader US-PRC relationship. It is becoming the infrastructure that shapes productivity, military capability, industrial systems and geopolitical influence itself.

Neither side can afford to concede leadership. But neither side can fully isolate itself from the other without imposing substantial costs on both the global economic system and itself.

The real question facing Trump and Xi is therefore not whether competition can be avoided. It cannot. The question is whether the world’s two AI superpowers can prevent strategic rivalry from hardening into long-term technological rupture.

Note: Some of the ideas developed in this article were first discussed in the author’s May 9 2026 interview with Radio France Internationale (RFI) for its Mandarin-language programme.