Pacific Money | Economy | East Asia

The Gas Inside Your AI Chip

The closure of the Strait of Hormuz has cut off a key source of helium, threatening the global semiconductor supply chain.

Since the closure of the Strait of Hormuz, most of the international headlines have focused on oil, liquefied natural gas (LNG), and shipping costs. But another, less visible crisis has been building in the gas supply chain that analysts have rarely modeled: helium. Qatar’s Ras Laffan Industrial City, the world’s largest LNG export facility, was forced offline after being struck by an Iranian drone in early March. Because helium is extracted as a byproduct of natural gas processing, the halt froze helium output simultaneously.

Qatar accounts for roughly one-third of global helium supply, and the Strait of Hormuz is the only maritime exit for what remains. Russia, once a meaningful alternative source through Gazprom’s Amur processing plant, has been unreliable since a catastrophic fire in 2021 damaged the facility, and subsequent export disruptions followed the Ukraine war. The global buffer that once existed is already gone. Even if the strait reopens, the damage to helium supply chains may not quickly reverse: Ras Laffan remains offline, stockpiles have been drawn down, and the disruption has already threatened production of the advanced semiconductor chips the global economy depends on.

A two-week ceasefire between the United States and Iran was announced on Tuesday, but passage through the strait remains blocked and fighting continues, and it remains unclear how long the ceasefire will hold. Regardless, the crisis has already exposed a category of vulnerability that neither governments nor industry had previously stress-tested: helium, a critical and non-substitutable input to advanced semiconductor production, is concentrated in a handful of countries and sits outside every supply chain risk model. The consequences extend from AI accelerators to the smartphones that billions depend on for basic economic life. Policymakers who have spent years hardening chip supply chains against geopolitical shock have been planning for the wrong crisis.

Helium is not a reagent – it does not etch silicon or develop photoresist – but it is indispensable in advanced node fabrication and has no near-term substitute. Fabs blow helium across the back of silicon wafers to maintain the thermal stability that precision at three nanometers and below requires. A professor of semiconductor devices at South Korea’s Sangmyung University said flatly that under current processes, there is no viable substitute for helium in wafer cooling. The Semiconductor Industry Association warned in 2023 that even a short disruption would produce shocks to global manufacturing.

The geographic concentration of this dependency is acute. Taiwan sourced 69 percent of its helium from Gulf Cooperation Council states in 2024. South Korea sourced approximately 65 percent from Qatar alone. Together, these two nations account for 36 percent of global semiconductor production capacity. South Korea’s exposure matters for a distinct reason: SK Hynix supplies roughly 60 percent of global HBM, the high-bandwidth memory inside every Nvidia AI accelerator, and Samsung operates its own advanced logic fabs at leading nodes. A helium shortage in South Korea constrains the memory that American AI hardware cannot currently be built without. Taiwan’s TSMC, meanwhile, produces roughly 90 percent of the world’s most advanced logic chips. The global helium supply chain operates on roughly 45 days of liquid inventory buffer before stockpiles boil off. As the Hormuz closure passed the four-week mark on March 28, the timeline is compressing.

The consequences for American AI are more immediate than they appear. Nvidia’s GPUs, Google’s TPUs, and the custom accelerators that Amazon, Microsoft, and Meta are deploying at scale all depend on TSMC’s Taiwan fabs at the 3nm and 2nm nodes. TSMC’s 3nm and 5nm production is fully booked through 2026, and Nvidia’s order backlog is measured in months. Any reduction in wafer starts tightens a constraint that was already structurally difficult before Ras Laffan went offline. Spot helium prices have already surged 70 to 100 percent; if the disruption extends beyond 60 days, analysts project contract price increases of 25 to 40 percent, costs TSMC passes to customers as higher wafer pricing. If the 45-day inventory clock runs out before Qatar restores output, TSMC faces the choice of rationing wafer starts, a decision with direct implications for Nvidia’s delivery schedule and the broader AI buildout.

The disruption will also affect Asia. Across the region, sovereign AI programs and regional hyperscalers are competing for the same constrained wafer supply. Japan’s RAPIDUS is attempting to set up domestic 2nm production with substantial government backing but remains years away from significant output and dependent on TSMC in the interim. Singapore, Malaysia, and India are all running national AI infrastructure that depends on imported accelerators built on advanced TSMC silicon. The Gulf states at the heart of this crisis are simultaneously among its most exposed customers: Saudi Arabia’s Humain initiative and the UAE’s G42 have both signed major Nvidia supply agreements in the past year. In other words, a helium shortage that forces wafer rationing hits every one of these programs, not just the American hyperscalers whose names appear most frequently in international press coverage.

China is in a different and more strategically significant position. SMIC and Huawei’s HiSilicon are pushing toward advanced nodes under U.S. export controls and depend on helium for any domestic fab capacity they have managed to build. But China has its own helium extraction capacity, concentrated in the Sichuan basin, which it has been quietly expanding. If the Hormuz disruption persists long enough, Beijing has the option of positioning itself as an alternative helium supplier to Asian manufacturers, chipmakers, and industrial users whose established supply chains have been disrupted.

The consequences also extend beyond AI infrastructure. The same advanced nodes that produce Nvidia GPUs also supply the mobile chipsets inside virtually every smartphone sold in emerging markets. Qualcomm’s Snapdragon series and MediaTek’s Dimensity line, which together power the majority of affordable Android handsets across Southeast Asia, South Asia, and Sub-Saharan Africa, are fabricated at TSMC and Samsung on nodes where helium is non-negotiable. A prolonged wafer shortage that forces fabs to ration starts will not prioritize consumer handsets over AI accelerators and data center chips, where margins are higher and contracts are longer. For the roughly 3.5 billion people who rely on smartphones as their primary gateway to mobile banking, agricultural price information, remittances, and health services, a tightening of mobile chipset supply represents a significant development constraint.

Taiwan and South Korea are not without options. The most actionable near-term lever is U.S. helium: the United States holds the world’s largest reserves and has domestic production capacity that could, under emergency coordination, be redirected toward allied fab ecosystems. The Chip 4 alliance linking the United States, Taiwan, South Korea, and Japan already provides channels through which a supply commitment could be negotiated and operationalized faster than building new sourcing relationships from scratch.

The second lever is recycling: leading-edge fabs already operate helium recovery systems capable of recycling 80 to 90 percent of process helium, but recovery rates vary significantly across facilities, and governments could co-finance rapid upgrades as part of an emergency package. The third lever is stockpiling: neither Taiwan nor South Korea maintains a dedicated strategic helium reserve, but could establish one along the lines of Taiwan’s crude oil reserve regime. Domestic chip makers, including SK Hynix and Samsung, reportedly hold approximately six months of helium supply, suggesting the model is already understood by industry, although such reserves have yet to be formalized by governments.  None of these options resolves the immediate shortage, but the combination represents a credible medium-term option if the Strait of Hormuz remains closed over the longer term.

In sum, what Hormuz has revealed is not a single vulnerability but a category of vulnerability previously outside the frame of semiconductor supply chain planning: critical, non-substitutable process inputs that are concentrated in a small number of countries, and never incorporated into the stress tests that governments and industry have previously run on the sector.

Get to the bottom of the story

Subscribe today and join thousands of diplomats, analysts, policy professionals and business readers who rely on The Diplomat for expert Asia-Pacific coverage.

Get unlimited access to in-depth analysis you won't find anywhere else, from South China Sea tensions to ASEAN diplomacy to India-Pakistan relations. More than 5,000 articles a year.

Unlimited articles and expert analysis

Weekly newsletter with exclusive insights

16-year archive of diplomatic coverage

Ad-free reading on all devices

Support independent journalism

Already have an account? Log in.

Since the closure of the Strait of Hormuz, most of the international headlines have focused on oil, liquefied natural gas (LNG), and shipping costs. But another, less visible crisis has been building in the gas supply chain that analysts have rarely modeled: helium. Qatar’s Ras Laffan Industrial City, the world’s largest LNG export facility, was forced offline after being struck by an Iranian drone in early March. Because helium is extracted as a byproduct of natural gas processing, the halt froze helium output simultaneously.

Qatar accounts for roughly one-third of global helium supply, and the Strait of Hormuz is the only maritime exit for what remains. Russia, once a meaningful alternative source through Gazprom’s Amur processing plant, has been unreliable since a catastrophic fire in 2021 damaged the facility, and subsequent export disruptions followed the Ukraine war. The global buffer that once existed is already gone. Even if the strait reopens, the damage to helium supply chains may not quickly reverse: Ras Laffan remains offline, stockpiles have been drawn down, and the disruption has already threatened production of the advanced semiconductor chips the global economy depends on.

A two-week ceasefire between the United States and Iran was announced on Tuesday, but passage through the strait remains blocked and fighting continues, and it remains unclear how long the ceasefire will hold. Regardless, the crisis has already exposed a category of vulnerability that neither governments nor industry had previously stress-tested: helium, a critical and non-substitutable input to advanced semiconductor production, is concentrated in a handful of countries and sits outside every supply chain risk model. The consequences extend from AI accelerators to the smartphones that billions depend on for basic economic life. Policymakers who have spent years hardening chip supply chains against geopolitical shock have been planning for the wrong crisis.

Helium is not a reagent – it does not etch silicon or develop photoresist – but it is indispensable in advanced node fabrication and has no near-term substitute. Fabs blow helium across the back of silicon wafers to maintain the thermal stability that precision at three nanometers and below requires. A professor of semiconductor devices at South Korea’s Sangmyung University said flatly that under current processes, there is no viable substitute for helium in wafer cooling. The Semiconductor Industry Association warned in 2023 that even a short disruption would produce shocks to global manufacturing.

The geographic concentration of this dependency is acute. Taiwan sourced 69 percent of its helium from Gulf Cooperation Council states in 2024. South Korea sourced approximately 65 percent from Qatar alone. Together, these two nations account for 36 percent of global semiconductor production capacity. South Korea’s exposure matters for a distinct reason: SK Hynix supplies roughly 60 percent of global HBM, the high-bandwidth memory inside every Nvidia AI accelerator, and Samsung operates its own advanced logic fabs at leading nodes. A helium shortage in South Korea constrains the memory that American AI hardware cannot currently be built without. Taiwan’s TSMC, meanwhile, produces roughly 90 percent of the world’s most advanced logic chips. The global helium supply chain operates on roughly 45 days of liquid inventory buffer before stockpiles boil off. As the Hormuz closure passed the four-week mark on March 28, the timeline is compressing.

The consequences for American AI are more immediate than they appear. Nvidia’s GPUs, Google’s TPUs, and the custom accelerators that Amazon, Microsoft, and Meta are deploying at scale all depend on TSMC’s Taiwan fabs at the 3nm and 2nm nodes. TSMC’s 3nm and 5nm production is fully booked through 2026, and Nvidia’s order backlog is measured in months. Any reduction in wafer starts tightens a constraint that was already structurally difficult before Ras Laffan went offline. Spot helium prices have already surged 70 to 100 percent; if the disruption extends beyond 60 days, analysts project contract price increases of 25 to 40 percent, costs TSMC passes to customers as higher wafer pricing. If the 45-day inventory clock runs out before Qatar restores output, TSMC faces the choice of rationing wafer starts, a decision with direct implications for Nvidia’s delivery schedule and the broader AI buildout.

The disruption will also affect Asia. Across the region, sovereign AI programs and regional hyperscalers are competing for the same constrained wafer supply. Japan’s RAPIDUS is attempting to set up domestic 2nm production with substantial government backing but remains years away from significant output and dependent on TSMC in the interim. Singapore, Malaysia, and India are all running national AI infrastructure that depends on imported accelerators built on advanced TSMC silicon. The Gulf states at the heart of this crisis are simultaneously among its most exposed customers: Saudi Arabia’s Humain initiative and the UAE’s G42 have both signed major Nvidia supply agreements in the past year. In other words, a helium shortage that forces wafer rationing hits every one of these programs, not just the American hyperscalers whose names appear most frequently in international press coverage.

China is in a different and more strategically significant position. SMIC and Huawei’s HiSilicon are pushing toward advanced nodes under U.S. export controls and depend on helium for any domestic fab capacity they have managed to build. But China has its own helium extraction capacity, concentrated in the Sichuan basin, which it has been quietly expanding. If the Hormuz disruption persists long enough, Beijing has the option of positioning itself as an alternative helium supplier to Asian manufacturers, chipmakers, and industrial users whose established supply chains have been disrupted.

The consequences also extend beyond AI infrastructure. The same advanced nodes that produce Nvidia GPUs also supply the mobile chipsets inside virtually every smartphone sold in emerging markets. Qualcomm’s Snapdragon series and MediaTek’s Dimensity line, which together power the majority of affordable Android handsets across Southeast Asia, South Asia, and Sub-Saharan Africa, are fabricated at TSMC and Samsung on nodes where helium is non-negotiable. A prolonged wafer shortage that forces fabs to ration starts will not prioritize consumer handsets over AI accelerators and data center chips, where margins are higher and contracts are longer. For the roughly 3.5 billion people who rely on smartphones as their primary gateway to mobile banking, agricultural price information, remittances, and health services, a tightening of mobile chipset supply represents a significant development constraint.

Taiwan and South Korea are not without options. The most actionable near-term lever is U.S. helium: the United States holds the world’s largest reserves and has domestic production capacity that could, under emergency coordination, be redirected toward allied fab ecosystems. The Chip 4 alliance linking the United States, Taiwan, South Korea, and Japan already provides channels through which a supply commitment could be negotiated and operationalized faster than building new sourcing relationships from scratch.

The second lever is recycling: leading-edge fabs already operate helium recovery systems capable of recycling 80 to 90 percent of process helium, but recovery rates vary significantly across facilities, and governments could co-finance rapid upgrades as part of an emergency package. The third lever is stockpiling: neither Taiwan nor South Korea maintains a dedicated strategic helium reserve, but could establish one along the lines of Taiwan’s crude oil reserve regime. Domestic chip makers, including SK Hynix and Samsung, reportedly hold approximately six months of helium supply, suggesting the model is already understood by industry, although such reserves have yet to be formalized by governments.  None of these options resolves the immediate shortage, but the combination represents a credible medium-term option if the Strait of Hormuz remains closed over the longer term.

In sum, what Hormuz has revealed is not a single vulnerability but a category of vulnerability previously outside the frame of semiconductor supply chain planning: critical, non-substitutable process inputs that are concentrated in a small number of countries, and never incorporated into the stress tests that governments and industry have previously run on the sector.

Alvin Camba, Ph.D. is the Lead Scientist and Director of Research at Lyvi, and a nonresident fellow in the Indo-Pacific Security Initiative at the Atlantic Council’s Scowcroft Center for Strategy and Security.

semiconductor industry

Semiconductor manufacturing

U.S.-Israel war on Iran

© The Diplomat