Executive Takeaway

In January 2026, the Bureau of Industry and Security (BIS) shifted license reviews for certain advanced artificial intelligence (AI) chips (including the Nvidia H200-class and equivalent chips) from a policy of denial to a case‑by‑case review conditioned on certain technical limits, supply integrity, and verifiable controls. License approvals, if any, will hinge on regulator‑credible evidence and enforceable license conditions.^[1] Although some market participants have described the policy change as a loosening of US export controls, the practical effect is more limited. As White House official Michael Kratsios emphasized this week, the United States “isn’t opening the floodgates” for H200 exports to China—policy fundamentals remain unchanged despite the shift to case‑by‑case review, and approvals will hinge on verifiable controls and stringent license conditions.^[2]

The BIS move is better understood as a procedural recalibration rather than a policy reversal. It likely reflects an effort to enhance US bargaining leverage as President Trump prepares for renewed diplomatic engagement with China this spring. Specific license outcomes, however, will be determined through technical BIS review rather than by headline policy announcements, and any approved license almost certainly will be subject to stringent compliance and license conditions. The revised posture aligns with the American AI Exports Program’s emphasis on programmatic, enforceable governance for full‑stack American AI exports.

This development reinforces themes that Alvarez & Marsal’s National Security, Trade, and Technology (NSTT) practice has emphasized in recent publications, including the growing importance of enforceable controls, operational credibility, and execution discipline in US technology policy.^[3] From an implementation perspective, companies navigating this shift will need to align US export-control expectations with on-the-ground information security, systems governance, and operational realities abroad.

Implications and Consistency With AI Export Program Trends

Under the revised policy, BIS will review license applications for H200-class and equivalent chips destined for China or Macau on a case-by-case basis, provided applicants can demonstrate compliance with multiple threshold conditions, including:

• “Sufficient US supply” attestations supported by auditable shipment and demand data
• Performance limitations confirming that chips operate below defined advanced-computing thresholds
• Foundry capacity protections showing that exports will not divert production from US or allied end users
• Aggregate shipment caps limiting China and Macau volumes to no more than 50% of comparable US shipments
• Third-party technical testing conducted by qualified, US-headquartered laboratories
• Robust know-your-customer, physical security, and remote-access controls, including restrictions on infrastructure-as-a-service, training, and inference use

BIS will continue to apply a presumption of denial for exports to end users owned by, headquartered in, or controlled from Country Group D:5 jurisdictions and will maintain heightened scrutiny of downstream access and use.^[4] These requirements are consistent with the tenets of the American AI Exports Program’s model: export full‑stack capability with embedded governance, continuous monitoring, and independent validation.

What Did Not Change

Several fundamentals remain unchanged:

• There is no categorical reopening of AI chip exports to China and Macau.
• Enforcement risk and compliance expectations remain elevated.
• Licenses are not assured even where threshold criteria are met.

The policy change effectively moves from automatic rejection to discretionary review, while preserving denial authority where national security concerns persist. Moreover, given the policy attention to AI-enabling technology exports to China, any approved license almost certainly will include stringent compliance license conditions.^[5]

A&M NSTT has previously discussed many of the underlying dynamics driving the license review policy shift in “What the US AI Action Plan Means for Export Controls and US National Security,” which highlights the move toward programmatic, enforceable governance models.^[6]

Why Update the License Review Policy Now

The timing of the revision is notable—it follows months of internal debate over AI diffusion, allied coordination, and China-related risk. Most notably, it coincides with renewed diplomatic engagement planning by the United States.

From a policy standpoint, the revised review framework:

• Enhances US leverage in bilateral negotiations
• Avoids locking the administration into a categorical denial posture
• Signals conditional flexibility without committing to approvals
• Reflects growing interagency focus on restoring order to America’s AI policy, moving from ad hoc signals to repeatable and predictable licensing outcomes for companies

For companies and investors, the implication is clear. The specific details of each transaction, rather than the policy announcement, will determine outcomes. Companies operating in or with China-facing environments should expect continued scrutiny of how approved technology is secured, accessed, and governed in practice, not just how it is described in licensing submissions.

For additional insights into the juncture of geopolitics and compliance outcomes, see A&M NSTT’s “What the 2025 US National Security Strategy Means for Transnational Companies, Investors, and Advisors.”^[7]

What Companies and Investors Should Do Now

1. Treat licensing as a high-friction, evidence-driven process.
   Applicants should expect intensive scrutiny of supply data, performance characteristics, end-use/end-user controls, and security architecture. Unsupported assertions are unlikely to succeed.
2. Prepare for operational, not paper, compliance.
   BIS expectations increasingly focus on provable, enforceable controls, including identity assurance, physical access restrictions, remote-access governance, logging, and continuous monitoring, particularly in operating environments located in or connected to China and Macau.
3. Model commercial scenarios conservatively.
   Transaction timing, valuation, and revenue assumptions should reflect uncertain approval timelines, conditional or partial authorizations, and ongoing oversight obligations.
4. Integrate export strategy with broader China risk management.
   Companies should recognize that this BIS policy shift intersects with outbound investment scrutiny, data-security regimes, and supply-chain controls. Fragmented compliance materially increases risk.

Readers can further explore how strategic technology controls are increasingly embedded within broader economic and investment frameworks in “America First Investment Policy: Disruption and Opportunity.”^[8]

How A&M Helps Clients Navigate This Shift

A&M’s National Security, Trade, and Technology practice supports companies and investors at the execution layer of US technology controls, where policy objectives meet operational reality.

In the AI and advanced-compute context, NSTT supports clients and counsel with:

• License strategy and application readiness, including performance-threshold analysis, shipment-ratio tracking, and sufficient-supply attestations
• Security-by-design implementation aligning physical, cyber, compute, and access controls with BIS expectations
• Independent verification, audits, and investigations that provide regulator-credible assurance
• End-use and end-user governance, including infrastructure, training, and inference controls

A&M NSTT coordinates these efforts with A&M colleagues based in China and Hong Kong who bring deep expertise in information security and systems security. Our partnership allows clients to align US export-control expectations with practical, on-the-ground implementation, while maintaining US regulatory credibility and trusted performance across jurisdictions.

Bottom Line

The BIS policy change does not reopen the door to unrestricted AI chip exports to China and Macau. Instead, it replaces a bright-line denial rule with discretionary review, while significantly increasing execution demands for industry.

For companies and investors, success will depend less on headlines and more on whether their technology, security, and governance models can withstand government scrutiny and demonstrate trusted performance.