Strategic Coexistence: Navigating US-China AI & Semiconductor Race Requires Adaptive Controls & Friend-Shoring

The global contest for AI and semiconductor dominance between the United States and China is profoundly reshaping world economies, cybersecurity postures, and international relations over the next two decades. This strategic rivalry, characterized by export controls, massive R&D subsidies, and a concerted push for indigenous capabilities, is fostering technological fragmentation with significant economic and security implications. While national security concerns drive technology decoupling, the economic and innovation costs are substantial, with potential global GDP losses of up to 5% in severe fragmentation scenarios.

The dynamic nature of this competition is evident in policy recalibrations, such as the US decision to lift export restrictions on NVIDIA's H200 AI chips in late 2025 or early 2026, balancing national security with economic realities. The most likely scenario is one of limited competition or strategic coexistence, where targeted restrictions coexist with some market access and partial interoperability. This approach necessitates precise, adaptive controls, prioritizing friend-shoring, robust domestic talent development, and proactive, conditional engagement with 'tech middle powers' to build resilient innovation ecosystems and mitigate severe fragmentation risks.

The US-China competition for AI and semiconductor dominance is multifaceted, encompassing critical dimensions such as compute, models, adoption, integration, and deployment. This rivalry has been intensified by sweeping US restrictions on advanced semiconductor exports to China between 2022 and 2024, aiming to impede China's progress in cutting-edge AI capabilities essential for advanced weaponry, surveillance, and economic competitiveness.

China has responded with a massive state-backed push for indigenous production, exemplified by its 'Made in China 2025' initiative and over $150 billion in semiconductor investments since 2014. This drive has yielded significant results, with SMIC reportedly achieving volume production of its 5nm-class N+3 node without EUV tools by late 2025, a breakthrough that challenges previous assumptions about the efficacy of export controls. Huawei's HiSilicon division further demonstrated adaptive resilience by replacing over 13,000 sanctioned components in its products, including the Kirin 9000s processor.

The dynamic nature of this competition is also reflected in US policy shifts, such as the decision to lift export restrictions on NVIDIA's H200 AI chips in late 2025 or early 2026. This move suggests a continuous recalibration to balance national security objectives with economic considerations for US tech companies and the broader global economy. The strategic landscape is further complicated by China's R&D spending, which reached $1.03 trillion (PPP-adjusted) in 2024, surpassing the US's $1.01 trillion, indicating an accelerating pace of indigenous innovation.

This rivalry is fundamentally shifting the global technological landscape from a highly globalized model to one characterized by regionalized hubs and bifurcated supply chains. Both nations are scaling public investment in domestic fabs and AI research, with the US CHIPS Act allocating $52.7 billion for domestic semiconductor R&D and manufacturing. This trajectory points towards a future of technological divergence, where separate but advancing parallel tracks emerge.

The **United States** aims to maintain technological leadership, deny critical capabilities to adversaries, and protect national security. Its revealed preference includes imposing export controls while demonstrating tactical flexibility, as seen with the H200 chip reversal, to balance national security with the economic interests of its tech firms. The US leverages its technological prowess and economic influence, holding high veto power in global tech standards.

**China** is committed to achieving technological self-reliance, overcoming foreign dependencies, and becoming a global leader in AI and semiconductors. Its strategy involves massive state-backed investment in indigenous R&D and manufacturing, actively seeking to bypass foreign restrictions, and promoting its own digital infrastructure through initiatives like the Digital Silk Road. China also possesses high leverage and veto power in shaping its technological trajectory.

**Tech Middle Powers**, including the EU, Japan, South Korea, Taiwan, India, Israel, and the Netherlands, seek to protect their economic interests, maintain access to global markets, and carve out strategic niches. They are diversifying supply chains, attracting foreign direct investment from both blocs, investing in domestic tech capabilities, and developing independent regulatory frameworks, such as the EU AI Act adopted in May 2024. These nations hold medium leverage but lack veto power over superpower decisions.

**Multinational Tech Corporations** prioritize maximizing profits, maintaining global market access, and optimizing supply chains for efficiency and innovation. They are actively navigating export controls, diversifying manufacturing locations through 'friend-shoring,' and lobbying governments for policy adjustments. These corporations exert medium leverage but do not possess veto power.

**Developing Nations** aspire to access affordable technology for economic development and avoid becoming collateral damage in the tech rivalry. They often seek investment and technology transfer from both the US and China, adopting multi-alignment strategies, but remain vulnerable to a 'tech tax' and 'debt-trap diplomacy.' These nations typically hold low leverage and lack veto power.

The technological competition is driving new alliance formations, with the US and China anchoring rival technology blocs and pressuring other nations to align supply chains, standards, and data policies. This risks a more fragmented, less interoperable global order, undermining existing multilateral frameworks like the WTO and UN, which are ill-equipped to manage rapid technological decoupling.

The competition will intensify geopolitical rivalry for resource access, talent, and market share, with the strategic importance of Taiwan, a critical hub for advanced semiconductor manufacturing, remaining a flashpoint. Policy volatility, such as the H200 chip restriction reversal, suggests that identified red lines (e.g., advanced AI chip exports, critical infrastructure access) may shift, creating dangerous miscalculation risks.

In the security domain, both the US and China are rapidly developing advanced AI capabilities, which will be leveraged for offensive and defensive cyber operations. This includes AI-driven offensive cyber capabilities for automated vulnerability exploitation, deepfake-enabled social engineering, and sophisticated supply chain attacks, leading to an acceleration of cyber conflict and a reduction in reaction times.

The drive for indigenous semiconductor production, while reducing reliance on adversaries, also creates new, potentially less scrutinized supply chains, increasing vulnerabilities to hardware backdoors or software compromises. Fragmented security architectures, with separate chip architectures and AI development paths, will create new attack vectors and increase the attack surface for supply chain and firmware attacks. China's 'Digital Silk Road' is expanding surveillance infrastructure in over 60 countries, creating backdoors for espionage.

The weaponization of technology and export controls can provoke retaliatory cyber operations targeting critical infrastructure or supply chain chokepoints, leading to 'gray zone' sabotage, such as China subtly degrading US AI models via data poisoning. This dynamic risks an accelerated military-technological competition, leading to a new arms race in autonomous systems, cyber warfare, and intelligence capabilities, where technological parity between superpowers is inherently unstable and can increase incentives for risk-taking and escalation.

The economic costs of technological decoupling are substantial and measurable, with technological fragmentation potentially resulting in global economic output losses ranging from $0.6 trillion to $5.7 trillion, or about 5% of global GDP, particularly affecting many economies. This is driven by increased capital expenditure for semiconductor fabrication (fabs), duplicated R&D, inefficient supply chains, and a loss of economies of scale, leading to higher production costs for chips, potentially increasing by 20-40%.

Technological fragmentation also reduces trade volumes, diminishes economies of scale, and leads to inefficient resource allocation as countries duplicate efforts and forgo comparative advantages, directly impacting global GDP growth. While decoupling is intended to foster domestic champions, it paradoxically risks innovation stagnation by denying all players access to the world’s best talent, tools, and feedback loops, slowing the development of next-generation chips and AI models.

The US strategy of pressuring allies (Japan, South Korea, Netherlands) to align with export controls creates secondary sanctions risks for firms supplying China, while China's acceleration of indigenous semiconductor production and its 'Digital Silk Road' initiatives reshape global trade. Foreign direct investment in China’s tech sector has declined, while US and EU subsidies attract firms to relocate, leading to a 'tech tax' on developing nations forced to choose between tech stacks.

Increased production costs due to localized, less efficient supply chains, coupled with reduced global competition, will translate into higher prices for AI-enabled products and semiconductor components. This will feed into global inflationary pressures, particularly in technology-dependent sectors, impacting consumer purchasing power and corporate profitability globally.

**Full Decoupling (Probability: 25-30%)**: This scenario envisions a complete separation of technological ecosystems, characterized by severe trade restrictions, minimal collaboration, and distinct, incompatible standards. It would lead to the most severe economic losses, potentially exceeding 5% of global GDP, maximum innovation loss, and heightened geopolitical instability, creating a 'digital iron curtain' globally. Cybersecurity would face persistent, high-intensity attacks, and international relations would be fragmented and unstable.

**Limited Competition (Strategic Coexistence/Partial Decoupling) (Probability: 55-60%)**: This is the most probable scenario, involving targeted decoupling in critical national security sectors (e.g., advanced military AI, leading-edge chip manufacturing) while maintaining some economic and technological exchange in less sensitive areas. It projects moderate economic fragmentation (1-3% global GDP reduction), selective innovation loss, and the formation of distinct but not entirely separate tech blocs. Geopolitical competition would remain high but managed, with episodic cyber confrontation, allowing for strategic denial while mitigating the most severe economic blowback.

**Technological Divergence (Integrated within Limited Competition)**: This sub-scenario, with a high likelihood, suggests parallel advancement where both sides achieve significant capabilities through separate development paths. The US would likely maintain advantages in AI models and software ecosystems, while China achieves hardware self-sufficiency faster than anticipated, leading to significant but not absolute innovation loss. This implies a shift from outright denial strategies to focusing on innovation acceleration.

**Selective Cooperation (Technological Détente) (Probability: 10-20%)**: This scenario involves limited cooperation on global challenges (e.g., AI safety, climate modeling) that require shared data or computational resources, while competition continues in other areas. It would result in lower economic costs, shared threat mitigation, and accelerated innovation, fostering a more stable, rules-based international order. However, the current geopolitical climate makes widespread cooperation challenging, limiting it to specific, mutually beneficial, non-military applications.

The primary risk is that the pursuit of technological security may ultimately undermine the innovation that creates genuine security advantages, with potential global GDP losses of up to 5% and measurable innovation reduction. This creates a critical trade-off between short-term security gains and long-term economic and technological competitiveness.

Export controls, while intended to deny access, are demonstrably leaky, as evidenced by the smuggling of NVIDIA H100 GPUs to China throughout 2024. This circumvention not only undermines the effectiveness of controls but also accelerates indigenous innovation in target nations, potentially making denial strategies counterproductive over time.

The instability of technological parity between great powers, where neither side can achieve decisive advantage, increases incentives for risk-taking and escalation beyond current levels. This 'asymmetric technological stagnation' could lead to more dangerous outcomes, including intensified cyber warfare and geopolitical instability.

Developing nations face disproportionate harm, risking a widening digital divide and being forced to 'choose sides' in technology standards, affecting their sovereignty and economic policy. They are vulnerable to becoming battlegrounds for technological influence, facing a 'tech tax' and potential debt traps from infrastructure loans.

Policy incoherence, such as the tactical recalibration seen with the H200 chip reversal, can create investment uncertainty for multinational corporations. Over-reliance on domestic champions may also breed inefficiency and rent-seeking, distorting market signals and leading to suboptimal capital allocation.

For the **United States and Allies**, a policy of strategic coexistence with precise, adaptive controls is recommended. This involves strengthening minilateral tech alliances like the Chip 4 and Quad AI Working Group to counter China’s tech influence. Export controls should be narrowly tailored to cutting-edge AI chips (e.g., NVIDIA B100, AMD MI400), while allowing limited commercial sales of older models to avoid full decoupling and maintain industry revenue for R&D.

Investment in resilience is paramount, including diversifying supply chains through 'friend-shoring' to trusted partners such as India, Vietnam, and Mexico. Aggressive domestic R&D, supported by the $52.7 billion CHIPS Act, must be paired with open science, talent visas, and co-innovation platforms to accelerate next-generation capabilities. Proactive engagement with 'tech middle powers' is crucial, offering incentives like technology transfer in non-sensitive areas and supporting their indigenous capabilities, such as India's planned $15 billion in semiconductor investments by 2030.

Reinvigorating multilateral tech governance through forums like the OECD and G7 is necessary to preserve interoperability and security, including developing AI safety standards to prevent misuse, such as bans on autonomous weapons and deepfake regulations. This approach aims to find the most effective balance where national security objectives are met with the least economically damaging tools.

For **China**, the strategy should focus on accelerating self-sufficiency in semiconductors through entities like SMIC and Huawei HiSilicon to reduce dependence on US tech. Expanding the Digital Silk Road can lock in developing nations through AI infrastructure loans, and leveraging 'tech diplomacy' can offer preferential market access to nations adopting Chinese standards.

**Developing Nations** should adopt a 'multi-alignment' strategy, engaging with both US and Chinese tech ecosystems to avoid over-dependence. Investing in local AI talent, such as India’s AI skilling programs, and negotiating data sovereignty agreements are crucial to prevent foreign surveillance overreach and reduce brain drain.

An **interagency task force** (US) should be established within two weeks to develop granular, adaptive policy frameworks for export controls, investment incentives, and international tech partnerships. Policy effectiveness and market shifts should be reviewed at six-month intervals, with a comprehensive strategic review every two years, adapting controls and engagement strategies as needed.

Precise quantification of long-term innovation loss under partial decoupling scenarios remains a significant limitation, as economic models struggle to fully capture the dynamic and adaptive responses of global actors. The exact pace, yield, and cost-effectiveness of China's advanced semiconductor manufacturing, particularly for SMIC's 5nm-class and future 3nm capabilities, are subject to ongoing debate and lack complete independent verification.

The degree of autonomy and influence that 'tech middle powers' can truly exert under sustained pressure from both superpowers is uncertain. It is unclear whether they can consistently act as independent arbitrators or will eventually be forced into more definitive alignments, impacting the resilience of diversified supply chains.

The unpredictability of future technological breakthroughs, such as in quantum computing or advanced general artificial intelligence (AGI), could fundamentally alter the competitive landscape and render current policy frameworks obsolete. Furthermore, the extent to which domestic political pressures in the US and China will consistently override economic pragmatism remains a critical unknown, influencing the likelihood of extreme decoupling scenarios.

The specific impact of AI-driven cyber warfare on global stability, including the frequency and severity of attacks on critical infrastructure and the potential for rapid, autonomous escalation, is an evolving and largely unquantified risk. The role of non-state actors in disrupting the global tech order also presents a significant blind spot.

**Verified Claims:** The H200 chip restriction reversal in late 2025 or early 2026; the US CHIPS Act allocating $52.7 billion; China's semiconductor investments exceeding $150 billion; the 1973 OPEC Oil Shock; China's R&D spending reaching $1.03 trillion (PPP-adjusted) in 2024, surpassing the US's $1.01 trillion; Huawei's HiSilicon developing the Kirin 9000s processor and replacing 13,000 components post-US restrictions; SMIC achieving 5nm-class production (N+3 node) by late 2025 without EUV; NVIDIA H100 smuggling occurring throughout 2024; and the EU AI Act adopted in May 2024, entering into force in August 2024, with full application by August 2027.

**Partially Verified Claims:** The potential global GDP loss of up to 5% due to fragmentation is cited from World Economic Forum reports, which refer to financial system fragmentation, not solely technological fragmentation, but indicate the scale of potential economic impact.

**Contradicted/Unverified Claims:** The assertion that the US-Soviet Cold War military-industrial complex and space race started in 1950 is incorrect, as these began earlier with the Soviet Union's first nuclear test in 1949. The claim that telecommunications infrastructure competition (e.g., Huawei 5G restrictions) started in 2015 is also incorrect, with US government restrictions on Huawei beginning in 2017. The specific figure of a 60% increase in cyberattacks since 2020 is unverified, though a general increase in cyberattacks is widely reported. Reports of SMIC achieving 3nm mass production by 2026 remain speculative and unconfirmed by independent sources for high-volume, high-yield production.