The Paradox of Conditional Access in U.S.–China AI Chip Controls
Author
Demetria Hu
Editor
James Lautens
Publications Lead
Gianluca Mandarino
Introduction
In mid-January 2026, tensions over advanced AI chips have increasingly taken shape through customs enforcement and administrative discretion. On January 13, the Trump administration formally approved exports of Nvidia’s H200 AI chips to China under new security conditions, subject to third-party review, quantity caps tied to U.S. supply, and firm-level security requirements. Within 24 hours, Chinese authorities instructed customs agents that Nvidia’s H200 chips were “not permitted to enter the country,” and domestic tech firms were cautioned against purchasing them unless necessary. Two weeks later, on January 28, China approved its first batch of several hundred thousand H200 chips during Nvidia CEO Jensen Huang’s visit to China, with approvals allocated primarily to ByteDance, Alibaba, and Tencent.
While AI chips are only one component of a broader U.S.–China economic security policy, they offer a clear example of how leverage is exercised via managed exposure. In this paper, “managed exposure” refers to the policy strategy employed by Washington, while “conditional access” describes how recipient states and firms experience that strategy in practice. Access to advanced AI chips has become a negotiable instrument, granted conditionally, priced politically, and governed by administrative discretion rather than stable rules. This paper argues that conditional access contains a structural paradox: by signalling reversibility, it preserves formal control while encouraging hedging behaviour that erodes leverage over time.
The U.S. Approach: Export Controls and Managed Exposure
Since late 2022, the U.S. has deployed export controls to constrain China’s access to advanced computing capacity, placing H200-class chips under presumptive controls. The U.S. Commerce Department’s Bureau of Industry and Security (BIS) has tightened rules on advanced integrated circuits, end-use licensing, and related technologies. These measures aim not only to limit China’s ability to acquire cutting-edge chips, but also to slow its capacity to produce or optimize them domestically.
Rather than fully cutting off access, U.S. policy has evolved toward managed exposure: allowing some shipments under restrictive conditions that aim to protect national security interests while preserving economic engagement. The approval of H200 exports fits this pattern. According to reporting, the rules require that China cannot receive more than 50% of the chips sold to U.S. customers, and that shipments would be subject to independent third-party review. Recipients would also be required to demonstrate security procedures to prevent restricted end uses, including military, intelligence, or weapons-related applications.
For Washington, managed exposure preserves leverage while avoiding the economic costs of outright decoupling. This approach embeds strategic discretion directly into administrative processes. Access decisions are contingent and reversible, with enforcement mechanisms that can be adjusted as circumstances change. Because permissions are neither time-bound nor rule-based, access under this export control and licensing framework cannot reliably anchor long-term investment, infrastructure planning or supply-chain commitments.
China’s Response: Discretionary Friction and Conditional Acceptance
China’s initial response to the H200 approval took the form of blocking shipments at customs. Rather than signalling outright rejection, the move reflected caution toward reliance on access governed by external discretion. It signalled that Chinese authorities appeared to interpret conditional access itself as a source of vulnerability.
The subsequent approval of a limited batch of H200 imports clarified this posture. Rather than treating the approval as the beginning of stable access, Chinese authorities confined imports to a small number of large platform firms and did not issue any public indication of broader reopening, limiting the clearance to firm-specific allocations. By doing so, Beijing treated access as a tactical exception rather than a basis for long-term dependence.
This response directly illustrates the paradox at the centre of conditional access as leverage. Because access was discretionary and reversible, it failed to generate the dependence necessary for leverage to operate over time. Instead of incorporating H200 access into long-term planning, Chinese authorities treated it as provisional and continued to prioritize parallel options. In doing so, they reduced the strategic value of Washington’s approval: access that is not institutionalized cannot easily be withdrawn or threatened.
This stance reflects a strategic assessment of how leverage operates under conditional access. Where access is perceived as contingent, actors face strong incentives to hedge rather than commit. China retains other levers it can adjust, including controls on upstream inputs and informal administrative guidance that shapes firm behaviour. While such measures are often opaque in real time, they are consistent with longer-term planning frameworks such as the Fifteenth Five-Year Plan, which emphasizes technological self-reliance as a core component of the national development strategy.
Spillovers and Third-Country Pressure
The H200 episode illustrates broder dynamics: conditional access generates spillovers that extend beyond the immediate parties, reshaping incentives for firms and governments that must plan around regulatory uncertainty. When access to advanced compute is conditional, actors not directly targeted by controls adjust their behaviour preemptively.
At the firm level, uncertainty surrounding export policy influences investment, sourcing, and production decisions across global supply chains. For example, suppliers to Nvidia reportedly paused H200 component production after the Chinese customs response, illustrating how anticipated enforcement shifts can propagate upstream even without formal bans. Firms hedge by delaying commitments, diversifying sourcing, or reallocating capacity toward less politically exposed markets.
These dynamics also place pressure on third-country firms and governments operating across both the U.S. and Chinese regulatory spheres. In these countries, decisions regarding compliance, supply chain restrictions, or market orientation are increasingly influenced by anticipation of regulatory reversals. Manufacturing and logistics hubs in Southeast Asia, including Vietnam and Malaysia, have absorbed additional electronics and semiconductor-adjacent investment as firms accelerate existing diversification strategies in response to regulatory uncertainty, a trend documented throughout 2025 and expected to intensify in 2026. Rather than producing clean decoupling, conditional access encourages diversification and regulatory alignment, fragmenting markets along political lines.
These spillovers feed back into leverage itself. When firms and third countries plan around uncertainty rather than permissions, the influence of conditional access diffuses. Control is preserved formally, but its practical impact diminishes as actors hedge rather than commit.
Looking Ahead
As both Washington and Beijing increasingly rely on discretionary controls, the risk of miscalculation shifts from overt escalation to cumulative friction. The primary challenge is no longer abrupt policy shocks, but the gradual erosion of predictability as actors anticipate restrictions that may never materialize.
For policymakers, this suggests that the effectiveness of conditional access depends not only on enforcement capacity, but on the credibility and coherence of signalling. Frequent recalibration or episodic approvals risk encouraging hedging behaviour that undermines leverage by default. Where access is previewed as unreliable, recipients and third parties alike have incentives to treat permissions as temporary convenience rather than foundations for long-term planning.
This dynamic complicates the strategic use of export controls. While discretion preserves flexibility and avoids the political costs of outright bans, it also narrows the margin for stable expectations. Over time, conditionality that is too fluid may weaken influence by accelerating precautionary adjustment across firms and supply chains, even in the absence of formal escalation.
Conclusion
This paper explores how uncertainty, discretion, and reversibility change behaviour even when access technically exists. The H200 episode illustrates how competition over advanced semiconductors is increasingly governed through conditional access rather than categorical prohibition. By embedding discretion into export control regimes, policymakers preserve flexibility and retain formal leverage. Yet this same design feature reshapes incentives in ways that encourage hedging rather than commitment.
As firms, third countries, and states plan around uncertainty rather than permissions, leverage becomes harder to sustain. Access that is discretionary and reversible may influence behaviour in the short term, but its strategic value erodes as actors reduce dependence and develop parallel options. In this environment, the central challenge is not whether exposure can be managed, but whether leverage derived from conditional access can remain effective once uncertainty itself becomes the dominant organizing principle.