Tech Stack Diplomacy: Policy Implications of the U.S. AI Export Strategy
The White House’s plan to extend its AI technology advantage through the export of U.S. technology relies on dated assumptions of its unipolar technological dominance. As other countries develop their own national AI and emerging technology strategies, the U.S. policy framework often puts it at odds with its partners’ national interests, creating a risk of ineffective technological alignment. Additionally, it creates an avenue for allies to hedge against U.S. influence by cultivating parallel relationships with China.
The U.S. strategy to lead globally in AI is outlined in America’s AI Action Plan, released in July by the White House. A core objective of the plan is to “export the entire U.S. tech stack” to secure allies and establish global dependency on U.S. technology ahead of competitors like China. This objective was reinforced by Executive Order 14320, which explicitly states the need for the U.S. to not only develop advanced AI capabilities but also to ensure that U.S. AI technologies, standards, and governance models are adopted worldwide to strengthen alliances and maintain technological dominance. The United States defines an AI tech stack as encompassing AI-optimized computer hardware; data pipelines and labeling systems; AI models and systems; measures to ensure the security and cybersecurity of AI models and systems; and AI applications for specific use cases. To implement its export objective, the action plan instructs the secretary of commerce to establish and implement an AI Exports Program to support the development and deployment of full-stack AI export packages.
The goal to export the U.S. tech stack is based on the premise that exporting a robust stack would provide tremendous advantages for the U.S. and its companies to expand its global AI footprint and guide standards, software, and advancements of the future. Additionally, it would solidify its reach and lock nations into dependency on U.S. models and technology before an ascending China has the chance to combat it. Michael Krastios, director of the White House Office of Science and Technology Policy, suggested that this objective was built upon his experiences in the first Trump administration witnessing Huawei’s global telecommunications expansion and the challenges U.S. companies had in combating it.
However, while it is important to expand technology to regions before China does, the overall export goal relies on a dated assumption of unipolar technological dominance. As nations around the world develop their own national AI and emerging technology strategies, highlighting the goals for digital sovereignty and global leadership, it remains untested if countries are willing to import a full tech stack. This assumption risks a resulting policy framework that is often at odds with the national interests of its partners, creating a risk of ineffective technological alignment. Additionally, it creates an avenue for allies to hedge against U.S. influence by cultivating parallel relationships with China.
Dual AI Action Plans
In August, China followed the release of the U.S. AI action plan with one of its own. The State Council of China’s Opinion on In-Depth Implementation of the “Artificial Intelligence +” Initiative (关于深入实施“人工智能+”行动的意见) details the requirements and objectives for China’s AI development over the next decade .It is apparent both nations view their AI priorities from different vantage points, which may inform the way they export their tech stacks globally.
The AI+ Plan sets specific targets, aiming for a 70% adoption rate of AI terminals and agents by 2027, a figure projected to increase to 90% by 2030. By 2035, these intermediate targets are designed to assist in “achieving basic realization of socialist modernity,” a national goal since 2017. The plan specifically targets the comprehensive integration of AI across six major sectors of society and the economy: science and technology, industry, consumption, livelihoods, governance capacity, and global cooperation.
China also aims to share its AI technology with other countries and leverage multilateral institutions, specifically the United Nations, as a vehicle for improving global AI governance standards. Furthermore, the Chinese plan calls for a renewed focus on promoting open-source AI development and providing support to Global South countries to enhance their own national AI capabilities. With these objectives, China hopes to communicate to the world that it is a reliable partner in AI and emerging tech and sees developing nations as crucial partners that can’t be left behind.
On the other hand, the U.S. plan prioritizes continued U.S. dominance of the sector, framing AI as an urgent international competition. Its strategy centers on exporting U.S. standards, values, and models globally to secure this leadership. It explicitly views its AI ecosystem, encompassing cutting-edge chips, models, and technical standards, as a comprehensive “full stack” offering to be promoted aggressively to allied nations to limit market access to adversaries. While the plan dedicates a pillar to “Leading in International AI Diplomacy and Security,” its framing is often adversarial, tasking the government to “Counter Chinese Influence in International Governance Bodies.” This approach, which focuses more on export competition than genuine, sustained multilateral collaboration or working through the United Nations and other global bodies to create consensus, leaves a strategic opening for China to gain a narrative win by promoting seemingly more inclusive concepts of shared governance in international forums.
In a period of intense rivalry, both the U.S. and China are actively leveraging control over critical areas of the global technology stack to gain strategic advantage. For the United States, this has centered on debates about export controls on essential inputs, most notably previous restrictions on the sale of advanced chips desired for Chinese AI development, resulting in the decision to allow the export of Nvidia’s H200 chips. Conversely, China has asserted control by introducing a rare-earth licensing regime, which was put on hold following talks between Trump and President Xi Jinping in November. This regime grants Beijing the power to dictate which companies and countries receive the resources necessary not only for semiconductor production but also for other critical, high-tech, and defense applications worldwide. The resulting volatile interdependency risks global technological bifurcation. The strategic ripple effects of this power struggle are profoundly influencing the policies and balancing acts of middle powers globally.
The Role of Middle Powers
U.S. and Chinese dominance over the most advanced layers of the tech stack appear to forge a bipolar contest between the world’s biggest powers. The U.S. leads in high-end compute (semiconductors, data centers) and frontier models, while China excels in AI adaptation, hardware manufacturing, and in open-weight AI. This dominance necessitates that any other country must coordinate with one or both powers to build its own AI systems. However, this structure has not yet resulted in a purely bifurcated competition between allied blocs, despite framing of the AI race by OpenAI and others as “democratic vs. authoritarian.” Instead, middle powers have carved out significant room to maneuver by strategically balancing relations with both, while simultaneously developing sovereign components of the tech stack to advance their own national goals.
With their pursuit of “sovereign AI” many nations aim to free themselves from dependence on the U.S. and China by developing advanced models and uplifting their domestic workforce. However, the current global structure severely constrains this ambition. U.S. and Chinese companies control the most critical sectors of the technological stack (compute, chips, and foundational models), leaving most countries without truly independent options for AI development. Consequently, countries with limited AI and compute capacity are becoming geopolitical battlegrounds for technological expansion, a dynamic evident in programs like China’s Digital Silk Road and the U.S.-aligned OpenAI for Countries initiative, as well as its claims against Chinese companies of interference in international AI outreach.
Middle powers in the AI space have strengths of their own. The European Union, for instance, acts as a powerful regulatory pole, while key member states leverage technological strengths: France contributes through advanced AI models like Mistral, and the Netherlands houses ASML, a crucial manufacturer of high-end semiconductor production equipment. Concurrently, Gulf states such as the United Arab Emirates and Saudi Arabia are leveraging massive sovereign wealth to invest heavily in AI compute infrastructure and talent, actively pursuing their goal of becoming regional and global AI leaders. While middle powers don’t possess the total capability to compete with the U.S. and China, they utilize economic and technological relationships to seize opportunities to expand their role in the tech space and secure necessary components of the tech stack for themselves. This can be seen most prominently through the UAE’s AI policy and partnerships, which represent the first major test of the White House’s export strategy.
The UAE’s AI Hedge: A National Imperative
The UAE has prioritized becoming a global AI leader as part of its economic transformation into non-oil sectors. To guide this effort, it adopted the UAE National AI Strategy 2031 in 2019, aiming to build a reputation as an AI destination, attract global talent, and establish itself in the rapidly growing field. It also created the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), a research institute aiming to train talent and build AI systems. In November, Microsoft’s AI Diffusion Report identified the UAE as the global leader in AI adoption with 59.4% usage among working-age adults. To fulfill its ambitious goals, the UAE has resisted fully embracing the entire tech stack of an individual country, as its ultimate objective remains building its own sovereign AI capabilities. While the U.S. has succeeded in building closer relations with Abu Dhabi, its effort to export its tech stack has been insufficient and instead could work towards building a future competitor.
Faced with U.S. and Chinese dominance in the tech sector, and with its own abundance of cash, the UAE’s strategy allows it to integrate itself into the tech stacks of the most powerful nations, then, utilizing its talent, companies, and leverage, forge a network of AI models and systems built on cutting-edge technology that could compete with the best U.S. and Chinese models. Additionally, the UAE looks to seize the gap amongst frontier models adapting to Arabic language and cultural contexts in the Middle East region.
The UAE’s technology partnerships demonstrate a clear hedging strategy, though pursuing it has required significant concessions in specific sectors. The primary example is the recent U.S.-UAE AI Acceleration Partnership, which signals a major deepening of the UAE’s tech relationship with the U.S. Central to this deal is the planned 10-square-mile U.S.-UAE AI campus that will host Stargate UAE, a 1-gigawatt compute cluster built by the Emirati firm G42 and operated by OpenAI and Oracle. As part of the deal, Nvidia is set to export up to 500,000 of its advanced Grace Blackwell GB200 systems annually, and the UAE will enable access to ChatGPT nationwide, a success for OpenAI’s global expansion. In total, this partnership amounts to an export of the U.S. tech stack, as the UAE will receive hardware, models, software, applications, and compliance standards. However, Abu Dhabi’s larger AI strategy blunts its effectiveness, based off the White House’s AI plan is blunted by Abu Dhabi’s larger AI strategy.
A Dual-Track AI Strategy
The central sticking point for the U.S.-UAE partnership involved G42. As a private company, G42 is tasked with attracting foreign AI investment and is consequently more vulnerable to partner scrutiny. The company’s prior investments in and partnerships with Chinese firms raised alarms, specifically the risk that China could gain access to restricted chips. Following years of pressure, G42 was ultimately compelled to divest from its Chinese holdings to secure the U.S. partnership. This decision reflects the UAE’s calculation that sacrificing relationships in China was a necessary cost to access the best U.S. tech. Furthermore, the success of this alignment was cemented by Microsoft’s $1.5 billion minority investment in G42 in 2024, which was followed by a further commitment of $13.7 billion in overall investments and spending through 2029.
Despite the significant attention focused on G42, the company is only one of several AI players in the UAE competing to develop and launch models for the broader Middle East – models that all stand to benefit from the U.S.-UAE partnership. To navigate the geopolitics of U.S.-China technology competition, the UAE appears to be employing a dual-track approach. Alongside the private track led by G42, there is a separate public track in which UAE-funded organizations often take cues from China’s strategy by offering low-cost, open-weight models as the foundation of their AI ecosystems. For instance, the Falcon AI system, developed by the government-funded Technology Innovation Institute (TII) in Abu Dhabi, has achieved success with its releases, including a dedicated Arabic version.
The public track of the UAE’s strategy also includes models built on top of other base models. MBZUAI’s open-weight model K2 Think achieves performance competitive with systems like ChatGPT and DeepSeek while maintaining a low-cost and lightweight design. This achievement reflects a deliberate hybrid approach by the UAE, strategically integrating elements from both major global technology ecosystems. Specifically, the model utilizes the Chinese stack by building on Alibaba’s Qwen 2.5 AI model and crediting DeepSeek for its fine-tuning methodology in establishing its reasoning system. It also incorporates the American stack and the Emirati stack by being trained and tested on hardware from U.S. chipmaker Cerebras and deployed via G42-operated data centers. The ultimate result is not a purely U.S.-aligned offering, but a unique product that merges contributions from both leading AI powers, while firmly asserting the UAE’s own sovereign ambitions in the emerging landscape. While G42 has been under more scrutiny for its dealings with Chinese companies, MBZUAI has managed to continue a strong relationship with individuals and organizations closely tied to the CCP and its political objectives.
The UAE maintains its intent to pursue strong technological collaboration with China despite being compelled on a case-by-case basis to divest from or downplay certain relationships with Chinese firms. While the U.S. is successfully exporting components of its tech stack to the UAE and securing important concessions in the geopolitical competition, the UAE is unlikely to abandon its largest trading partner and a leading AI power that supplies the foundation for many of its national systems, such as its extensive Huawei 5G network. Therefore, it accepts U.S. restrictions to ensure access to essential cutting-edge AI hardware, but still actively works to grow its technological and economic collaboration with China through public entities tied to the government.
Strategic Implications for U.S. Tech Leadership
False Assumptions
The objective to export the U.S. tech stack relies on two key assumptions: 1) countries will gladly accept U.S. export packages to build their AI foundation and 2) countries must choose between U.S. or Chinese tech dominance. On the ground, however, it is apparent neither assumption holds true for many middle powers and emerging regions. While the UAE is unique in its ability to raise capital and attract large-scale AI investments, the U.S-UAE partnership is less impactful because of Abu Dhabi’s dual-track strategic approach between the U.S. and China. Its private track has been successful in forging lucrative deals with top U.S. companies to build domestic digital infrastructure that will raise its own profile. By importing advanced chips and opening to nationwide ChatGPT access, it is adhering to key elements of the American AI export objectives. However, the public track still appears to maintain a strong connection to China through mirroring its open-weight model strategy and continuation of data center projects with leading Chinese companies.
Building Up Future Competitors
Without a detailed tech stack export strategy, the U.S. risks building up future competitors that will hold a level of sovereignty over data centers, and introduce models aimed to defeat leading American companies. Moreover, models created in states like the UAE are already using parts of the Chinese tech stack at the base of its models, detracting from the America AI Action Plans’ objectives to prioritize American values, standards, and model dominance. While significant vendor lock-in is achieved through deals like the U.S.-UAE partnership, exporting the tech stack beyond desired infrastructure and hardware remains a challenge.
The Importance of Open-Weight Models
The ability for companies and countries to easily build on Chinese models emphasizes the significance of open-weight models and their scaling and deployment advantages over proprietary models. China and its companies do not have to be present to spread its models and resources. This creates an opening for China to maintain a foothold in countries like the UAE despite the pressure for companies like G42 to cease its partnerships with Chinese firms. While it forms a difficult obstacle for China to navigate, it won’t be possible to keep China from continuing to make reasonable gains in the Gulf region and beyond that are in line with its current technological capacity. It also does not prevent the creation of new Chinese-built data centers in the UAE, as proven by Alibaba’s new venture.
Chinese open-weight models have also made inroads in the United States. For instance, Airbnb CEO Brian Chesky made note that the company’s AI customer service agent relies heavily on Alibaba’s Qwen model, as it’s fast and cheap. Other reporting suggests that many Silicon Valley start-ups are inclined to use Chinese models for similar reasons.
Overall, the UAE case is unique due to Abu Dhabi’s technological capacity and capital, which extends beyond other middle powers and emerging states.
Policy Recommendations
The views expressed in this article are those of the author and not an official policy or position of New Lines Institute.
