US-China AI Race Takes Centre Stage at Trump-Xi Summit: Chips, Safety and What Comes Next
By Shahnoor Saqib | SCN Analysis
WASHINGTON — When Donald Trump hosts Xi Jinping at the White House on Thursday, artificial intelligence will sit alongside trade, critical minerals, Taiwan and other strategic disputes on an agenda that increasingly looks less like a conventional bilateral summit and more like an attempt by the world's two largest powers to manage a technological rivalry neither believes it can afford to lose.
The immediate diplomatic objective is comparatively modest. U.S. Treasury Secretary Scott Bessent and Chinese Vice Premier He Lifeng have already discussed creating a bilateral AI dialogue and a notification mechanism through which Washington and Beijing could alert each other to serious AI incidents with national-security implications. Trump and Xi are expected to consider that proposal when they meet, but the broader competition over advanced chips, computing power, infrastructure, models and global AI standards will remain unresolved.
That distinction matters. An agreement on AI safety would not amount to an agreement to slow the technological race. Washington's stated policy is to maintain American leadership, while Beijing has made AI a central component of its technological and industrial ambitions. Both governments can therefore simultaneously cooperate on preventing catastrophic AI failures and continue competing aggressively over who builds the most capable systems.
The White House has confirmed Xi's state visit for September 24, his first to the White House in more than a decade. China's Foreign Ministry says Xi intends to hold in-depth discussions with Trump on bilateral relations and wider international issues, framing the visit as part of an effort to maintain what Beijing calls a relationship of "strategic stability."
AI has moved from commercial competition to national power
The central question surrounding the summit is no longer simply whether an American or Chinese company develops a better chatbot.
AI increasingly intersects with military systems, intelligence, cyber operations, biotechnology, industrial productivity, autonomous systems and critical infrastructure. The Trump administration's June national-security AI directive explicitly called for accelerated adoption of advanced AI across the U.S. national-security establishment and described maintaining technological superiority as a strategic objective.
America's broader AI Action Plan likewise identifies infrastructure as fundamental to maintaining leadership: semiconductor factories, data centres and the electricity required to operate them. The document specifically acknowledges China's rapid expansion of its power grid while arguing that the United States must accelerate construction of energy and computing infrastructure.
China approaches the issue differently in public language but with similarly high strategic stakes. Beijing describes AI as a major driver of technological and industrial transformation and has promoted international governance initiatives stressing development, security, controllability and wider access to AI technology. China's government released another international AI ethics governance plan in July, advocating risk controls, international cooperation and broader access for developing countries.
Behind those differing formulations lies the same geopolitical reality: neither Washington nor Beijing wants the other to establish an overwhelming technological advantage.
The chip battle is not going away
That makes semiconductors one of the hardest parts of the relationship.
Advanced AI requires enormous computing capacity, which in turn depends on sophisticated accelerators, semiconductor manufacturing equipment, networking technology, memory and data-centre infrastructure. Washington has used technology restrictions and other measures to protect what it considers strategically sensitive capabilities.
The Trump administration has also taken additional semiconductor measures this year. In January, Trump invoked Section 232 authorities concerning semiconductor imports and imposed a 25% tariff on certain advanced computing chips, including Nvidia H200 and AMD MI325X products under specified circumstances, while directing officials to negotiate broader semiconductor arrangements.
Yet China's response illustrates why export restrictions alone are unlikely to settle the AI contest.
Huawei said this month that domestic demand for its AI computing equipment exceeds its available production capacity. The company is preparing its Ascend 950DT AI chip and expects Chinese model developers to begin training models on systems using the processor next year. That does not establish technological parity with the most advanced U.S. ecosystem, but it shows how restrictions can also accelerate Beijing's effort to develop domestic alternatives.
Chinese developers have simultaneously demonstrated that constraints on access to the world's most advanced hardware do not prevent model development altogether. The result is a competition increasingly taking place across the entire technology stack: chips, software, algorithms, energy, data centres and applications.
For Trump and Xi, therefore, a comprehensive semiconductor settlement would require compromises touching the core of each country's national-security strategy. There is little evidence that either side is preparing to make such a concession at this summit. Indeed, U.S. Trade Representative Jamieson Greer said sophisticated AI-chip and semiconductor-equipment export controls were not part of the AI-mechanism discussions held in New York ahead of the leaders' meeting.
Where an agreement may actually be possible
The more realistic opening is AI risk management.
Bessent's proposal would establish some form of notification mechanism for AI incidents reaching a national-security threshold. Details remain unresolved — including what qualifies as such an incident, what information governments would disclose and how quickly they would communicate.
But the underlying logic resembles crisis-management mechanisms developed during earlier periods of strategic rivalry.
Washington and Beijing do not have to agree on who should lead the AI industry to recognise that an uncontrolled AI-related event could harm both countries. Potential areas for future discussion identified around the current talks include AI weaponisation, attacks against critical infrastructure, cyber risks and other high-consequence failures.
China has publicly acknowledged similar risks. Its Foreign Ministry said this month that Beijing takes both inherent and secondary AI risks seriously and supports safeguards intended to keep systems safe, reliable and controllable. China nevertheless prefers a multilateral framework with a significant United Nations role rather than governance dominated by Washington or a small group of technologically advanced states.
That creates a narrow but potentially consequential overlap.
The United States and China could remain technological competitors while agreeing that certain failures are dangerous to both.
Such an arrangement would resemble strategic guardrails more than technological cooperation.
The summit cannot separate AI from trade
AI is also entangled with the wider economic negotiations.
Trump and Xi reached agreements during their May meeting in Beijing covering trade, investment, Chinese purchases of U.S. agricultural goods and aircraft, and critical-mineral supply. The White House said at the time that China had committed to addressing American concerns about rare-earth supplies and related restrictions.
Four months later, several elements remain unresolved. Reuters reported after this week's preparatory talks that U.S. officials gave no indication of progress on some agricultural, aircraft and critical-mineral commitments.
Rare earths matter directly to the technology rivalry because modern computing infrastructure depends on complex global supply chains extending far beyond semiconductor fabrication.
This creates a form of mutual vulnerability.
The United States retains major advantages in advanced AI technology and key parts of the semiconductor ecosystem. China possesses enormous manufacturing capacity and important positions in critical-material supply chains while rapidly developing its own computing ecosystem.
Neither country's leverage is absolute.
That is one reason Thursday's meeting may produce selective agreements rather than a comprehensive technological settlement.
Taiwan remains underneath the technology conversation
There is another strategic issue that cannot be removed from the AI equation: Taiwan.
Taiwan's role in advanced semiconductor manufacturing means that security tensions across the Taiwan Strait carry implications extending far beyond conventional military calculations. A serious disruption would affect global electronics and AI infrastructure.
Taiwan is expected to be among the major issues at this week's summit, alongside trade, critical minerals and AI.
The two governments approach Taiwan from fundamentally different positions, making it unlikely that an AI discussion will substantially alter that dispute.
But the increasing dependence of modern economies and militaries on advanced computing makes stability around semiconductor supply chains even more strategically important.
What happens next?
The first thing to watch is whether Trump and Xi formally approve a continuing U.S.-China AI dialogue.
If they do, the summit's significance will not necessarily lie in a detailed agreement announced Thursday. The more important development could be the creation of a durable channel through which technical and national-security officials meet regularly.
Second is the proposed AI incident-notification mechanism. The critical details would include what triggers notification, whether military AI incidents are included, how confidential information is protected and whether either side has an obligation to respond.
Third is whether AI becomes separated into two diplomatic tracks.
One could address risk and safety, where limited cooperation may be possible.
The other would cover strategic competition — chips, computing infrastructure, investment restrictions, advanced models and technology exports — where the disagreements are substantially deeper.
That separation may prove necessary because attempting to negotiate the entire AI relationship as one package could make agreement considerably harder.
Fourth is the trajectory of semiconductor restrictions after the summit. Any tightening or relaxation of U.S. controls would affect China's access to computing power, while Chinese progress in domestic accelerators will determine how durable American hardware advantages remain.
Fifth is infrastructure.
The AI contest is increasingly becoming an electricity and construction contest as much as a software race. Training and operating increasingly capable systems requires enormous data centres, energy generation, cooling infrastructure and high-speed networks. America's own AI strategy explicitly identifies this physical buildout as necessary to sustain leadership.
China's ability to construct infrastructure at scale means that the long-term competition cannot be measured solely by which country produces the most capable model in any particular month.
From an AI race to an AI order
The deeper consequence of the Trump-Xi meeting could emerge well beyond Washington.
Countries across Asia, Europe, the Middle East, Africa and Latin America increasingly face decisions about whose chips, cloud infrastructure, models, standards and governance frameworks they will use.
Washington wants an international environment that supports American technological leadership. Beijing is simultaneously presenting AI as a global development opportunity and promoting governance frameworks emphasising multilateralism, sovereign choice and broader technological access.
That means the competition is not only about developing artificial intelligence.
It is increasingly about shaping the ecosystem surrounding it.
The country that builds highly capable AI systems matters. So does the country whose hardware powers overseas data centres, whose companies provide cloud services, whose standards governments adopt and whose regulatory concepts become embedded in international institutions.
This is where the U.S.-China AI contest could become more consequential than a conventional technology race.
A summit of competition and restraint
The paradox facing Trump and Xi is that success in AI could increase the need for cooperation between them.
More powerful models create greater strategic value but also potentially greater systemic risks. Washington may seek to maintain technological superiority while still wanting Beijing to communicate when a serious AI incident threatens U.S. interests. Beijing can pursue technological self-reliance while reaching the same conclusion about American systems.
The Trump-Xi summit therefore does not need to end the AI competition to change its character.
A plausible immediate outcome is much narrower: institutionalising communication, defining categories of dangerous incidents and establishing enough transparency to reduce the risk that an AI failure is mistaken for deliberate hostile action. Whether the two governments can move even that far will depend on a level of trust that remains limited.
What comes after Washington is consequently more important than any ceremonial declaration made during the state visit.
If a bilateral AI mechanism survives subsequent disputes over Taiwan, chips, trade and security, the United States and China could begin constructing guardrails around a technological competition that both expect to continue.
If it does not, the two largest AI powers may continue accelerating capabilities while possessing few agreed mechanisms for dealing with the consequences when those systems create a cross-border crisis.
That is why AI's place at the Trump-Xi summit matters. The question before the two presidents is not simply which country leads the AI era. It is whether two strategic competitors can build even limited rules for managing a technology that each increasingly regards as central to economic and national power.