Chip Fabs Now Consume More Helium Than MRI Machines. The AI Debate Never Noticed.
In short
Key takeaways
- The United States produces about 43% and Qatar about 33% of global helium, roughly 76% combined, and Ras Laffan ceased operations on 2 March 2026 after Iranian strikes, with further strikes on 19 March 2026.
- Helium cannot be effectively stored beyond 35 to 48 days in specialised cryogenic containers, and the current shortage is the fifth Brad Borggard, CFO of North American Helium, has counted in the past 20 years.
- Semiconductor manufacturing has surpassed medical MRI as the single largest consumer of helium, used in cooling, leak detection, plasma processing and advanced lithography.
- China imposed a temporary helium export ban on 10 July 2026 despite producing only about 15% or less of its own needs and importing roughly 85%.
- The US dropped helium from the final 2022 Critical Minerals List, 87 FR 10381, and sold the Federal Helium System assets in 2024, while Canada and Australia list it as critical and the EU is proposing to.
Helium is the only industrial gas that cannot be manufactured. It is a by-product of a handful of natural gas fields, it is chemically inert so nothing substitutes for it, and it boils off in transit so you cannot stockpile it. Qatar supplies about a third of it. Qatar's Ras Laffan complex was hit by Iranian strikes and ceased operations on 2 March 2026. On 10 July 2026 China imposed a temporary helium export ban. Airgas declared force majeure. Air Liquide warned of shortages.
And here is the connection that the AI infrastructure discourse has not absorbed: semiconductor manufacturing has overtaken medical MRI as the single largest consumer of helium on earth. That crossover has been reported, by Scientific American, Physics Today and Medscape. What has not followed it is the implication. The AI industry is downstream of an MRI coolant, and the people arguing about accelerator supply are not tracking that line.
Why there is no Plan B
Three properties make helium unlike every other strategic input. It cannot be manufactured: commercial helium is recovered almost entirely as a by-product of natural gas production from a relatively small number of gas fields, as Jack Lifton, Co-Chair of the Critical Minerals Institute, frames it: "Helium is not geologically scarce; competence is." It is chemically inert, the lightest inert element, and under any but the most extreme conditions it does not react with anything. And it cannot be stockpiled: it is difficult and expensive to store and boils off in transit.
There is no substitute for helium in deep cryogenic applications, meaning anything below roughly -429 Fahrenheit. That covers MRI superconducting magnets, NMR, and fusion research. That is a hard physical limit, not a commercial one.
The supply map and the buffer
USGS Mineral Commodity Summaries 2026 gives the concentration. US production was 81 million cubic metres of a roughly 190 million cubic metre world total in 2025, about 43% of the global figure per USGS data reported by Foreign Policy in April 2026. Qatar was 63 million cubic metres, about 33%. Russia is about 9.5%, with Algeria holding a meaningful additional share. China produces about 15% or less of its own needs and imports roughly 85% of its consumption. The US and Qatar together produce roughly 76% of world supply. Concentration that extreme in a non-commodity, non-manufacturable input is rare.
The buffer number is the one everybody should memorise: helium cannot be effectively stored beyond 35 to 48 days in specialised cryogenic containers, per AGBI on 25 March 2026. That is not a modelling assumption, it is a boil-off rate. Physics Today puts the recurring pattern the same way, listing global shortages in 2006-2007, 2011-2013 and 2018-2020, each of which disrupted MRI operations in multiple countries.
So the current disruption is not a novel event. Brad Borggard, CFO of North American Helium, counts it as the fifth in the past 20 years, which requires a 2022 disruption on the list as well. The response each time has been reactive.
The 2026 disruption, on a timeline
28 February 2026: Operation Epic Fury begins. 2 March 2026: Ras Laffan, the Qatari gas complex that supplies roughly one-third of the world's commercial helium, is attacked and ceases operations. 19 March 2026: further Iranian strikes. The supplier of a third of the world's helium is offline. NPR, 3 April: the blockade "is also disrupting the flow of about one-third of the world's supply of helium," and South Korea and Taiwan "are feeling the shortage the most. They're the biggest consumers of helium from Qatar."
Brad Borggard, CFO of North American Helium, on the arithmetic of it: "You can't make up for the fact that a third of world supply has been offline for a month." And on the pattern: "This helium shortage isn't the first. It's actually the fifth in the past 20 years. A lot of the supply is in geopolitically, I'll call them, challenged areas like the Middle East, like Russia."
Physics Today, Toni Feder, 30 April 2026: the shortfall "will be most severe during the next few months, before other sources can be integrated into the global supply." And the worse news: "Even when the conflict subsides and the Strait of Hormuz reopens, it will take months to get liquid natural gas and helium production back up to capacity."
The China export ban, read correctly
China's Commerce Ministry and customs agency announced a temporary export ban on helium, effective immediately, on 10 July 2026, under the Foreign Trade Law of the PRC. No elaboration given. Context from Trivium China's research note: "China only produces about 15% or less of its own helium, importing much of it from Qatar, which generates roughly one-third of the world's supply."
Gary Ng, senior economist at Natixis: the measure is "intended to protect the local industry, especially as it is critical to chipmaking," and is "likely more to secure supply for local usage than for political reasons." Cameron Johnson of Tidalwave Solutions read it the way it deserves: "The fact that they are now banning exports basically tells me that they know there's simply not enough helium to do what they need to do."
The same reporting supplies the qualification: as China is a relatively small exporter of helium, the move may not have a significant global impact. The export ban's significance is the signal, not the volume. A country is prioritising chips over export revenue, which only makes sense if it believes the domestic shortage is severe.
The demand shift the AI debate has not absorbed
Kornbluth Helium Consulting's usage data, as reported by Physics Today, Scientific American and Medscape in 2026: semiconductor manufacturing has surpassed medical MRI as the largest consumer of helium. The rest of the distribution puts low-temperature physics, chemistry, biomedical and other research together at about 10% of total use.
MRI's roughly 30% share of global commercial helium demand is the figure everyone knows, and it is the figure the entire framing rests on. Nobody writing about AI has followed that framing to its conclusion. One honesty note before building on it: the underlying usage breakdown is still one consultancy's estimate, so the load-bearing verification is corroboration from gas suppliers or fab procurement patterns. What has changed is that the crossover is no longer a single-source claim, it is reported independently by Physics Today, Scientific American and Medscape. Treat it as settled enough to plan around and not settled enough to stop checking.
What it does to fabs
Helium is used in four distinct semiconductor processes: cooling, leak detection, plasma processing, and advanced lithography. Without it, "many fabrication steps simply cannot be performed economically," per InvestorNews.
Where the operator voice is thin, say so. Phil Kornbluth, whose firm supplies semiconductor fabs, is the clearest public voice on this shortage and his framing is the one I would act on: "the situation is still quite fluid." Scotten W. Jones of Semiconductor Manufacturing Economics is concerned about the shortage but adds the nuance that matters: even if this drives prices, "the fundamental issue is whether the shortage becomes a crisis." Helium is a percentage of fab cost, so the pressure is on continuity of supply at specific process steps, not on unit economics.
Who is exposed: South Korea and Taiwan, the largest consumers of Qatari helium and the two economies most dependent on uninterrupted advanced-node production. The two most advanced semiconductor clusters on earth, the same two that dominate HBM and advanced packaging, are the two most exposed to a Qatari gas outage.
The market response and the queue
Airgas, the US packaged gases distributor, declared force majeure on helium shipments at 12:01am ET on 17 March 2026, after Qatar suspended production. Air Liquide warned helium shortages would persist, in its H1 2026 results. Nippon Sanso announced on 19 June 2026 an average helium price increase of 30% or more, effective July 2026.
The institutional response is recycling: after previous crises, institutions invested in recovery equipment, and many researchers report they are not yet having difficulty, pointing to protection through recycling, stockpiling and long-term contracts. But the allocation order is the actual social contract, and it reads like a joke that is real. Semiconductors and medical imaging are at the top of the pecking order for allocations. Researchers are lower down, along with party balloon sellers. Party balloon sellers. William Halperin, Northwestern's cryogenics director: "The time scale for response in the market as we see it as users is quite long." Facilities in Japan lacking reliquefying capability are the most exposed.
The policy gap, which is the strangest part
Helium is absent from the final 2022 Critical Minerals List, 87 FR 10381, published 24 February 2022, and it had already been dropped from the 2021 draft. Canada and Australia list it as critical. The EU is proposing to add it. The US sold the Federal Helium System to Messer LLC on 27 June 2024, against a $423.35 million bid, with $460 million transferred by December 2024, so the strategic reserve cushion is smaller and the market depends more on private operators, storage and logistics discipline. Meanwhile USGS estimates hundreds of billions of cubic feet of recoverable helium remain in US reservoirs, so the resource is not the constraint. The resource is abundant and the policy treats it as scarce in exactly the wrong direction.
Global Policy ran the strongest outside framing, Zamani and Farzanegan on 19 May 2026, arguing helium should be treated the way strategic minerals are. Their core point: helium has fewer buffers than oil. Oil has strategic reserves, multiple suppliers, liquid spot markets and route flexibility. Helium has boil-off losses, no substitution, concentrated liquefaction capacity, and a 35-to-48-day storage horizon. Their conclusion: "A local disruption to extraction, liquefaction, port operations, insurance or shipping can therefore become a global industrial constraint."
The concentration arithmetic
The checkable bound on this crisis: US 43% plus Qatar 33% is 76%, plus Russia's roughly 9.5%. With Qatar offline and the US at full rate, the world runs on roughly 67% of a normal year's supply, minus whatever logistics friction takes. That is the size of the hole. The volumes make it concrete: Qatar at 63 million cubic metres out of a roughly 190 million cubic metre global total in 2025, against the US at 81 million. And China has announced plans to add 250 million standard cubic feet a year of domestic capacity from May 2026, which helps, eventually, at the pace China builds things, which is not fast enough to be a 2026 answer.
Where this could be wrong
Concede the strongest counter-argument first: semiconductor helium use is a rounding error in fab cost, so this is not a pricing constraint. Partly true. Helium is a small share of unit cost. The argument is not about price, it is about continuity of supply at specific process steps, where "cannot be performed economically" is the operative phrase. Second: the US is 43% with hundreds of billions of cubic feet in reserve, so this is a solved problem. US production is the mitigation and it is real, but helium is absent from the final 2022 Critical Minerals List and the federal reserve was sold in 2024, so policy attention and buffer have both declined while production share stayed high, and the constraint is partly logistics: liquid helium needs specialised transport equipment that is itself in short supply, and boil-off makes long hauls lossy. Third: Qatar will restart. Probably, and Physics Today notes it will take months even after the strait reopens, so nobody should be predicting permanent shortage.
And the split between war-specific and structural. War-specific: the March 2026 outage, the China ban, force majeure. Structural: the 43/33 split, the 35-to-48-day storage horizon, the absence of substitution, the allocation economics. Only the second set justifies a standing conclusion, and one verification carries this whole thesis: if the Kornbluth semiconductor-over-MRI claim cannot be corroborated, it stays one consultancy's estimate reported by three outlets and the other findings carry the piece.
Five weeks of inventory
Anyone with a helium dependency, a lab, a fab, a cryogenics facility, has a 35-to-48-day inventory and no strategic reserve behind it. The largest consumer of the gas changed industries from medicine to semiconductors, and the industry that inherited the problem has not mentioned it. A third of supply went offline and the only people who noticed immediately were the people who cannot make more of it. That buffer number is the thing to remember, and Physics Today's own framing is the right last word on timing: even when the war ends, it will take months to get production back to capacity.
On this page
- Why there is no Plan B
- The supply map and the buffer
- The 2026 disruption, on a timeline
- The China export ban, read correctly
- The demand shift the AI debate has not absorbed
- What it does to fabs
- The market response and the queue
- The policy gap, which is the strangest part
- The concentration arithmetic
- Where this could be wrong
- Five weeks of inventory
Sources
- USGS Mineral Commodity Summaries 2026 (helium production by country)USGS, 2026
- Helium supply crunch puts MRI services at risk amid Qatar disruptionsMohamed Elashi, Euronews, 2026
- Strait of Hormuz closure deflates global helium supply (NPR, 3 Apr 2026)NPR, 2026
- Helium users brace as shortages begin (Physics Today, Toni Feder, 30 Apr 2026, DOI 10.1063/pt.ea34547464)Toni Feder, Physics Today, 2026
- The AI boom is dangerously dependent on helium (semiconductor overtakes MRI as largest consumer, 18 Mar 2026)Scientific American, 2026
- Could the Iran conflict disrupt MRI services (9 Jun 2026)Medscape, 2026
- Helium, Semiconductors, and the Strategic Risk of Hormuz (Global Policy, 19 May 2026)Zamani and Farzanegan, Global Policy Journal, 2026
Frequently asked questions
Why is helium different from other industrial gases?
Because it cannot be manufactured. Commercial helium is recovered almost entirely as a by-product of natural gas production from a small number of gas fields with economically recoverable concentrations. If those fields or their processing facilities go offline, there is no practical substitute, because helium is inert and nothing else will do the job below roughly minus 429 Fahrenheit. It also cannot be meaningfully stockpiled: it is difficult and expensive to store and boils off in transit, giving an effective buffer of only 35 to 48 days in specialised cryogenic containers. Unlike oil there is no large strategic global inventory to stabilise a market.
Is helium actually used in chipmaking?
Yes, in four distinct places, and this is the part most people miss. Helium is used for cooling, leak detection, plasma processing, and advanced lithography. Without it many fabrication steps cannot be performed economically. Semiconductor manufacturing has now surpassed medical MRI as the single largest consumer of helium worldwide, ahead of low-temperature physics and research at roughly 10 percent of total use, per usage data from Phil Kornbluth's consultancy as reported by Physics Today, Scientific American and Medscape in 2026.
Is the United States safe?
Safer than most. USGS data puts US production at about 43 percent of the global total and estimates hundreds of billions of cubic feet of recoverable helium remain in US reservoirs, with the country still the largest producer. But two things weaken that. Helium is absent from the final 2022 Critical Minerals List, 87 FR 10381, published 24 February 2022, while Canada and Australia list it as critical and the EU is proposing to. And the US sold the Federal Helium System assets in 2024, so the market has less official cushioning and depends more on private operators, storage and logistics discipline.