The US Didn’t Lift the Chip Ban — It Changed Who Decides
US export controls advanced computing semiconductors: what changed
The US Quietly Rewired the Chip Gate — What the New BIS License Review Policy Changes for Advanced Computing Semiconductors
The Commerce Department’s Bureau of Industry and Security (BIS) filed a rule—scheduled for Federal Register publication on January 15, 2026—that changes how the US reviews export license applications for certain advanced computing semiconductors bound for China and Macau. It swaps a default “no” posture for “case-by-case” review, but only inside a tightly defined box.
The immediate headline will read like a softening. The operational reality is stranger: the rule creates a narrow, conditional lane, and then surrounds it with new measurement, certification, and documentation choke points that can slow shipments just as effectively as an outright ban.
The story turns on whether “case-by-case” becomes a workable approvals pipeline—or a compliance bottleneck that only a handful of transactions can survive.
Key Points
BIS is revising its license review policy for exports of certain advanced computing commodities to end-users in China or Macau, shifting from a presumption of denial to case-by-case review within specific performance thresholds.
The case-by-case lane applies to chips with total processing performance (TPP) below 21,000 and total DRAM bandwidth below 6,500 GB/s, with examples in the rule text including NVIDIA’s H200 and AMD’s MI325X.
The policy change is conditional: exporters must certify US supply sufficiency, that exports won’t divert foundry capacity away from US end users, and that shipments to China/Macau stay within a 50% cap relative to shipments to US end users.
BIS is adding a third-party testing lab requirement in the United States to verify performance specs, plus expanded Know Your Customer (KYC) and controls around unauthorized remote access.
What looks like permissive language is, in practice, a new regime of definitions, proofs, and audit trails that determines who can ship, how fast, and at what compliance cost.
Background
US export controls on advanced chips sit inside the Export Administration Regulations (EAR). BIS can require exporters to apply for a license, then decides whether to approve.
A crucial concept is license review policy. This is not the same thing as “is a license required?” It is BIS telling industry what its default posture will be when a license application arrives. A presumption of denial means the government starts from “no,” and approvals are rare. Case-by-case means BIS will evaluate applications individually—still discretionary, but not pre-judged as a denial.
The January 2026 move targets a particular slice of “advanced computing” exports to China and Macau: chips under specific performance thresholds, and only if the exporter can meet detailed conditions. The rule also makes clear that certain related risk areas remain treated harshly, including scenarios involving covered entities and end-use risks.
Analysis
Political and Geopolitical Dimensions
The near-term geopolitical effect is messaging: Washington can say it has not “opened the gates,” while also signalling that it can calibrate pressure with more granularity than a blanket posture allows. Beijing, meanwhile, gets a clearer read on what the US considers “too capable” versus “commercially tolerable.”
Two plausible scenarios emerge quickly. First, a selective approvals scenario: BIS approves a limited number of shipments that cleanly fit the thresholds and paperwork requirements, using the lane to shape behaviour without conceding strategic advantage. Signposts would include early approvals clustered around a small set of product SKUs and a consistent set of compliance undertakings.
Second, a deterrence-by-process scenario: case-by-case exists on paper, but the combined friction of testing, certification, and auditability causes most transactions to stall or be abandoned. Signposts would include long application cycles, repeated “requests for information,” and a growing gap between announced commercial intent and actual export volumes.
Economic and Market Impact
Markets tend to hear “case-by-case” as upside for revenue and supply. Operations hear it as cost, delay, and uncertainty. The rule effectively creates a compliance premium: the ability to ship may depend on who can afford the testing, the counsel, the documentation build, and the ongoing monitoring overhead.
For chipmakers and their ecosystem, the biggest economic question is not whether “some exports” can happen, but whether the lane is stable enough to plan around. If approvals are unpredictable, companies may treat the lane as opportunistic rather than strategic—useful for one-off transactions, not for a dependable regional revenue base.
Two scenarios follow. In a managed predictability scenario, BIS approvals and denials settle into a pattern that compliance teams can model; firms then reintroduce China/Macau into quarterly planning with cautious volume assumptions. Signposts would be repeatable license conditions and fewer bespoke negotiations per application.
In a volatility premium scenario, approvals become sporadic and politically sensitive, pushing firms to keep China/Macau revenue off the “core forecast,” which then influences investor guidance and capex planning. Signposts would be uneven quarter-to-quarter disclosures and continued emphasis on “policy risk” as a material uncertainty.
Technological and Security Implications
The rule’s technical heart is definitional. It uses performance criteria—most prominently TPP and total DRAM bandwidth—to draw a boundary between what is treated as strategically unacceptable and what can be considered for approval. That boundary is not just engineering trivia; it becomes the legal line that determines licensing posture.
Security also appears in the compliance obligations: KYC procedures and controls designed to prevent unauthorized remote access, particularly where advanced chips could be placed into shared infrastructure environments. The practical effect is to shift part of the national security burden onto exporters and consignees, who must translate policy goals into enforceable operational controls.
A tight control scenario would see companies building hardened delivery models—segmented customers, locked-down deployment environments, and contractual constraints designed to satisfy the government’s comfort tests. Signposts would include standardized security attestations and repeated references to remote access constraints in license paperwork.
A leakage anxiety scenario would see BIS treat the remote access risk as too hard to mitigate in practice, leading to denials even where the raw performance numbers qualify. Signposts would be denials tied to deployment models (cloud, colocation, multi-tenant environments) rather than chip specs alone.
What Most Coverage Misses
The overlooked hinge is not “easier exports.” It is who gets to define compliance reality. The rule’s case-by-case lane is governed by measurable thresholds and certifications, and those measurements become the gate. If a company cannot prove, document, and independently validate that the chip sits below the defined criteria—and that the transaction satisfies supply, foundry-capacity, security, and KYC undertakings—then “case-by-case” is irrelevant.
In other words, the new posture moves the battleground from headline geopolitics to compliance mechanics: how performance is characterized, how supply sufficiency is justified, how foundry diversion is argued, and how remote access is controlled and evidenced. A firm can be “allowed in principle” and still fail in practice because the compliance story is not auditable.
That has a second-order effect: it quietly advantages players with mature export-control tooling and disciplined documentation pipelines. Smaller firms and smaller customers may be priced out of the lane. The policy shift, therefore, can reshape competitive dynamics even if the total number of approved exports remains limited.
Why This Matters
In the short term (the next 24–72 hours and the coming weeks), the impact is triage. Companies will re-check product classifications, map eligible SKUs to the thresholds, and decide whether the case-by-case lane is worth the cost of building out the required testing and certification workflow. Compliance teams will update internal guidance and accelerate customer segmentation—who is even eligible to be approached, supported, or supplied.
In the longer term (months to years), the rule is a template: export controls that look flexible on paper but are enforced through measurement, third-party validation, and operational undertakings. That approach can travel. It can be extended to other chip categories, adjacent components, or even other “dual-use” technologies where the US wants control without a universal prohibition.
Upcoming decision points are less about speeches and more about practice: whether BIS approvals begin to appear consistently, whether the conditions remain stable, and whether enforcement actions signal a tightening interpretation of what counts as acceptable risk.
Real-World Impact
A multinational hardware supplier reroutes internal resources to build a “China/Macau eligible” product bin, only to discover the real constraint is lab capacity and certification timing. Shipments become a scheduling problem, not a factory problem.
A cloud-adjacent customer is told it can only be served if it can demonstrate strong controls against unauthorized remote access. It reorganizes its infrastructure offerings—creating a segregated service tier—because the compliance story has to survive scrutiny.
An export compliance manager rewrites contracts so that security undertakings, audit rights, and end-use representations are enforceable. Deals take longer, sales cycles slow, and the compliance function becomes a revenue gatekeeper.
The Next Week’s Stress Test
This rule will be judged fast. Not by what it says, but by whether it produces a credible approvals pathway that companies can execute without betting the business on discretionary outcomes.
If case-by-case approvals start to flow in a consistent pattern, firms will treat the lane as narrow but real—and build around it. If approvals are rare, slow, or unpredictable, the lane will function as a signalling device and a paperwork trap, not a commercial channel.
Watch for three signposts: early approvals (or silence), the emergence of common “standard conditions” attached to licenses, and whether companies publicly describe the process as workable rather than merely “available.” The historical significance is that chip power is no longer controlled only by engineering and manufacturing—it is increasingly controlled by definitional law and compliance throughput
