US closes Southeast Asia AI-chip loophole, Korea caught in crossfire
The US just slammed shut a semicondutor export-workaround routed through five Southeast Asian and Latin American countries. Samsung and SK Hynix face a costly choice between Washington's demands and their fastest-growing revenue channels.
The loophole that built a shadow supply chain
Washington has quietly closed one of the most important gray-market workarounds in global semiconductor trade. On November 17, the Bureau of Industry and Security (BIS) published an interim final rule that strips the export-control exemptions previously available to design houses and packaging facilities operating in five countries: Singapore, Malaysia, Indonesia, Thailand, and Brazil.
The effect is immediate for Korea. Six South Korean companies — Samsung Foundry, Doosan Testna, LB Semicon, Nepa, SFA Semiconductor, and Steco — sit on the BIS-approved OSAT list. They had relied on a regulatory shortcut known as the “c-hyung” exemption, which let them package chips without prior government permission as long as the final transistor count stayed below a threshold (roughly 30 billion). That shortcut is now gone for any fabless designer headquartered in the covered countries.
Only one path remains: the narrowest one. A Chinese or Russian-linked company must now obtain explicit BIS authorization before accessing advanced processing or packaging in Korea — or abandon those processes altogether.
What changed, technically
Before this rule, two regulatory provisions carved out exceptions from the Advanced Computing Chip designation under EAR code ECCN 3A090.a. Sub-paragraph b allowed a foundry to self-verify that its customer was not a prohibited entity. Sub-paragraph c let an approved OSAT certify that the packaged chip fell below the transistor-density trigger. Together, they formed what industry insiders called the “fast track” — a process by which smaller fabless firms in allied or neutral countries could route production through Korea without waiting months for US export licenses.
The new rule eliminates both shortcuts for entities connected to the five named countries. Fabs and designers headquartered there can still qualify if they win formal “approved designer” status under sub-paragraph a, but that designation requires a separate, facility-by-facility BIS pre-export license application. In practice, that means weeks or months of delay per order, higher compliance overhead, and a de facto barrier to entry for anyone who cannot absorb that friction.
Violations carry severe penalties. Companies found shipping without proper authorization face the Entity List or a Denial Order — the equivalent of financial exile in the semiconductor world, cutting off access to American equipment, software updates, and spare parts.
Who wins, who loses
China is the primary target, and the rule achieves exactly what it sets out to do: it severs the pathway by which Beijing-affiliated firms used shell companies in Southeast Asia and Latin America to circumvent export controls. Before this change, a Shenzhen-based AI startup could incorporate in Singapore, contract Samsung Foundry for a 14-nanometer logic chip, and ship the packaged result to Korea without any US license. After January 1, 2027 — the rule’s effective date — that route no longer exists unless the Singapore entity obtains individual BIS approval.
South Korea is the unintended collateral damage. The current customer base for Korean foundries and OSAT houses in the five affected countries may be small today, but it is growing fast. As the regulatory burden shifts toward prescreening every prospective client — a KYC-style check on ownership structure, ultimate beneficial owners, and end-use — Korean providers will face rising costs even for transactions with no national-security risk whatsoever.
One semiconductor industry source told etnews that existing Samsung Foundry and domestic OSAT clients remain concentrated in North America, with negligible exposure to the Southeast Asian corridor so far. But the same source acknowledged that the rule creates a new headwind for customer acquisition. Any Korean house pursuing design-house clients in Singapore or Bangkok now carries the compliance cost of pre-screening every transaction against a regulatory regime that treats proximity to China as a proxy for risk.
For Samsung specifically, the economics matter. The company’s foundry business depends on filling capacity across a range of nodes, including 16 nm and 14 nm logic chips. Some of those orders come from smaller designers who historically rode the fast track. Strip away that channel, and Samsung faces a steeper sales cycle, more administrative load per wafer, and potentially lower utilization rates at certain fabs.
The cloud-remote-access dimension
The BIS rule contains a second provision that deserves equal attention. It requires US-based cloud providers to obtain advance permission before offering large-scale computing performance (measured against total computational power thresholds) to enterprises headquartered in designated security-risk countries, including China and Macau. This effectively blocks Chinese companies from renting frontier AI compute on US-hosted clouds — a route that had become increasingly popular as on-premises chip access tightened.
Taken together, the two provisions form a pincer: you cannot physically ship restricted chips through Southeast Asia, and you cannot remotely access restricted compute through US cloud infrastructure. Both close simultaneously, both target the same strategy — using geographic and digital arbitrage to circumvent US export controls — and both hit Korea indirectly by raising the cost of doing business for any Korean supplier serving clients in the five named countries.
Beijing’s response, and what comes next
Chinese Foreign Ministry spokesperson Guo Jiakun called the measures “economic coercion” and warned that China would take “necessary countermeasures.” The language is standard for this type of exchange, but the substance matters. Beijing sees the rules as an overreach of extraterritorial jurisdiction — applying US law to transactions involving non-American companies, equipment, and end-users. That is precisely the accusation Europe has leveled at similar US export controls, and it suggests the measure may deepen friction with American allies beyond the chip sector.
For Korea, the practical question is simpler than the geopolitical one. The rule takes effect January 1, 2027. Korean foundries and OSAT houses now have roughly three months to restructure their sales pipelines, build out compliance teams, and decide whether to retain borderline customers or shed them. The companies that already have robust KYC infrastructure — Samsung in particular, given its scale — will adapt faster. Smaller Korean OSAT operators may struggle with the administrative burden relative to their margins.
The broader implication extends beyond Korea. Any advanced-manufacturing hub that relies on US technology — Taiwan, Japan, the Netherlands — faces the same compliance pressure going forward. The US has signaled that it will continue tightening the net around Chinese access to frontier compute, using export controls as a tool of industrial policy rather than simply national security. For Korea, the message is blunt: alignment with Washington is mandatory, and the cost of that alignment falls entirely on the Korean supply chain.