Helium Crisis 2026: How Long Each Industry Probably Has Before Shortages Turn Serious

Most writeups on the Israel-Iran conflict and helium stop at one line: Qatar supplies about a third of the world’s helium. True, but that fact by itself does not tell you when things actually start breaking.

The number I kept coming back to was much smaller and more concrete. A fully loaded liquid helium ISO container holds roughly 40,000 liters. A mid-sized regional hospital with three MRI scanners can burn through around 1,500 liters a month, even if it’s reasonably careful about boil-off. Once you start thinking in containers instead of abstract global market share, the problem looks different. This is not a party-balloon shortage story. It is closer to a medical and semiconductor logistics story with a delay built into it.

So what follows is the practical version: how the bottleneck works, which sectors get hit first, what alternate supply can and cannot do, and what people should be doing on a 7-day, 30-day, and 90-day horizon. The numbers here are pulled from USGS production data, published QatarEnergy export volumes, distributor allocation history from the 2012 and 2019 shortages, and publicly available semiconductor process documentation. Some of the industry inventory estimates are still estimates, obviously. Nobody is posting their exact tank levels online.

The chokepoint is physical, not theoretical

The part I think gets missed in a lot of coverage is that this is not mainly a production problem. It is a transport and packaging problem. That changes the timeline a lot.

Qatar’s Ras Laffan Industrial City is the biggest single concentration point for helium on earth. Helium there is captured as a byproduct from North Field gas during LNG processing. QatarEnergy’s output is around 45 million cubic meters per year in crude-equivalent terms, which works out to roughly 30 to 32 percent of global refined supply. Two liquefaction trains at Ras Laffan turn that into liquid helium, load it into cryogenic ISO containers at about -269°C, and send it out through the Strait of Hormuz.

That strait is only 21 miles wide at its narrowest point. Every bulk helium shipment leaving Qatar has to pass through it. There is no pipeline workaround. Air freight is not a serious option for these volumes. And there is no clean reroute that just magically skips the Persian Gulf.

The chain looks simple on paper, but every step adds constraint:

• Helium is separated during gas processing. You do not drill a helium-only well in this system. If you want more helium, you generally need more natural gas production too.
• Then it gets purified and liquefied. Crude helium starts around 50 to 70 percent purity and has to be refined up to 99.9999 percent before liquefaction.
• Then shipping. Containers move from Ras Laffan to terminals like Zeebrugge in Belgium, Iwaki in Japan, and facilities in Texas and Louisiana. Gulf-to-Europe runs about 18 to 22 days. Asia-Pacific is more like 10 to 15.
• After that, inland distribution. Cryogenic truck, then cylinders or dewars to the end user.

The ugly detail is boil-off. Liquid helium evaporates continuously, even in very good containers. ISO containers lose something like 0.5 to 1 percent per day. So if a shipment sits delayed for 30 days, the problem is not just lateness. It may arrive with 15 to 30 percent less helium than when it was loaded.

That is the multiplier effect here. A Strait disruption does not merely pause supply. It erodes inventory already in motion while also cutting off replenishment behind it.

Who gets hit first

Shortages do not turn critical everywhere at once. This is one of the few useful things to be blunt about. Different sectors have very different inventory buffers, contract priority, and substitution options.

Global liquid helium consumption is about 160 million liters a year. Qatar accounts for around 32 percent of that, or roughly 51 million liters annually. If the Strait closes completely, that volume drops out of the replenishment pipeline within about 3 to 4 weeks after the last vessels clear.

After that, it becomes a sector-by-sector countdown.

Medical imaging

MRI systems are the obvious one, but they are not all equally exposed.

A standard 1.5T or 3T clinical MRI magnet usually holds 1,500 to 2,000 liters of liquid helium. Newer sealed-magnet designs like Siemens Altea and GE’s Freelium systems use dramatically less, in some cases about 0.7 liters, and are effectively zero-boil-off. The catch is that these are still a minority of the installed base.

Large academic centers often have helium recovery systems. Community hospitals usually do not. That difference matters more than people realize. A hospital without recovery can keep losing helium even when the scanner is just sitting there charged and idle.

My best estimate: well-equipped academic centers with recovery can probably stretch 3 to 5 months without replenishment. Community hospitals without recovery are in a much worse spot, more like 6 to 10 weeks before they start worrying about quench risk or cutting scan schedules. In practice, I would expect MRI delays at smaller facilities to show up around week 8 to week 12 of a sustained disruption.

Semiconductor fabs

This is the part most general-interest coverage tends to wave away, which is annoying because it matters.

Helium shows up in several wafer-processing steps: carrier gas in CVD, purge gas in EUV lithography chambers, cooling in ion implantation, leak detection all over the place. The most helium-sensitive fabs are the advanced-node ones, especially 3nm, 5nm, and 7nm, because EUV tools need ultra-pure helium environments.

TSMC, Samsung Foundry, and Intel Foundry are not buying this stuff casually on the spot market. They tend to have long-term offtake agreements and stronger inventory buffers than almost anyone else. Gartner estimated in 2023 that leading-edge fabs keep roughly 60 to 90 days of critical gas inventory on-site.

So the near-term picture is not “chip production stops next week.” It is more like 60 to 90 days before advanced fabs start feeling genuine production pressure. Older-node fabs have a bit more room. Some steps can use nitrogen or argon in partial substitution, depending on process. Those facilities may be able to muddle through for 4 to 6 months.

Research labs and NMR

This group gets squeezed fast.

NMR spectrometers are helium-hungry in a very boring, relentless way. A 500 MHz magnet holds around 60 to 80 liters of liquid helium and boils off continuously. Labs without recovery systems often refill every 2 to 4 weeks. Universities usually buy under annual distributor contracts, and those contracts nearly always include allocation language.

That mattered in 2019. Several major US universities reportedly got allocation notices cutting contracted volumes by 20 to 40 percent within six weeks.

Runway here is short: about 4 to 8 weeks before labs start cutting NMR hours or shutting instruments down selectively. Quantum computing labs are in a similar position. If you’re running dilution refrigerators at millikelvin temperatures, your substitution options are basically nonexistent.

Industrial users

These customers usually get pushed to the back of the line first.

Helium-shielded arc welding matters in aerospace, especially with aluminum, stainless, and titanium. Leak detection matters too. Fiber optic manufacturing uses helium in cooling glass preforms during drawing, though some plants have already pushed partial nitrogen substitution where they can.

So yes, they get hit early. But the consequences are uneven. Expect allocations to start within 2 to 6 weeks. For many industrial applications, operators can adapt with process changes or gas substitution. For titanium aerospace welding, not so much.

One practical way to think about it: there is probably a 10-to-12-week gap between the first industrial allocation notices and the first noticeable MRI capacity crunch at smaller hospitals. That gap is the warning period. A lot of procurement teams are going to waste it.

Three conflict scenarios

Duration is everything here. A 30-day disruption and a 90-day disruption are not the same problem scaled up. At 90 days, the system starts behaving differently.

Scenario A: short disruption, 1 to 4 weeks

The closest comparison is probably the 2019 tanker attacks. Those pushed spot pricing up 15 to 25 percent without causing major clinical shortages because distributors still had some buffer.

If 2026 looks like that, I would expect industrial allocation notices, spot prices up maybe 20 to 40 percent, and plenty of anxiety in research labs. MRI operations at well-supplied sites probably keep running normally. Once shipping resumes, the system starts recovering within 2 to 3 weeks.

Scenario B: medium disruption, 1 to 3 months

This is the one that looks most plausible to me, though I might be wrong about that if de-escalation happens faster than markets expect.

Here distributor buffers get used up. Leading-edge fabs pull harder on strategic reserves. NMR labs cut operating schedules or shut down some instruments. Community hospital MRI capacity starts slipping, maybe 15 to 30 percent in some regions as facilities try to avoid quench risk. Spot prices could hit 2 to 3 times contract levels, roughly in line with what happened during the 2012 dislocation tied to BLM stockpile policy.

And no, alternate suppliers are not riding in like some tidy Econ 101 chart. The US, Algeria, and Russia all matter, but in this time window they do not come close to replacing Qatar’s volume.

Scenario C: prolonged conflict, 3 months or more

This is where it stops being a shortage story and starts becoming a systems story.

Leading-edge semiconductor output begins to tighten in a measurable way. AI chip availability gets pinched 4 to 6 months later, once fab inventories burn down and the lag from wafer production to finished devices works through the pipeline. MRI backlogs stretch into weeks in many markets. Research universities pause NMR work. A lot of smaller users simply get frozen out.

Could new North American projects help? Eventually, yes. Canada’s Weil Group, North American Helium in Saskatchewan, Desert Mountain Energy. Together they might add 5 to 7 million cubic meters a year. But meaningful ramp-up is more like 12 to 18 months. That does not solve a 90-day emergency.

The replacement-capacity story needs a little less optimism and a little more math. Qatar’s 45 million cubic meters per year is not replaceable on short notice. The US produces around 40 million cubic meters, but much of that is already committed. Algeria’s Skikda plant does about 12 million. Russia’s Amur plant has serious theoretical capacity, around 60 million cubic meters a year at full output, but logistics and geopolitics make it a messy option for Western buyers.

Realistically, short-horizon replacement capacity is maybe 10 to 15 percent of what Qatar normally supplies. Not 100 percent. Not even close.

Also, slightly spicy opinion: people who talk about “the market finding alternatives” in this context usually mean they have not looked at how helium is actually moved and stored.

The AI buildout has a helium problem too

This part is easy to miss because the timing is lagged.

Every H100, B200, MI300X, and whatever else is getting poured into AI clusters depends on chips made through helium-intensive processes. TSMC’s advanced-node production uses helium in EUV lithography, ion implantation, leak detection, and related vacuum handling. ASML’s NXE-series EUV scanners require helium at six-nines purity, 99.9999 percent, and consume about 20,000 liters per tool per year in high-volume operation.

There are around 160 EUV tools deployed globally as of 2025, concentrated mostly at TSMC, Samsung, and Intel. EUV alone uses around 3.2 million liters a year. That is not a giant share of total helium demand, but at the process level it is extremely hard to substitute away.

So a long disruption does not instantly choke AI chip supply. The fab inventory buffer buys time. But if the Strait stays effectively closed for 3 months, you start setting up a chip squeeze that likely shows up in Q4 2026 or Q1 2027. The lag from helium shortage to visible AI infrastructure pain is probably 6 to 9 months.

If I were modeling AI datacenter build costs, I would be watching helium spot pricing as an early warning signal. Not the only one, but a real one.

Why 2026 is different from prior shortages

The earlier helium shortages each had their own flavor.

The 2006 shortage was largely demand-side. MRI adoption grew faster than liquefaction capacity. The 2012 shortage was tied to US federal helium policy and the drawdown of the BLM stockpile, which once held 32 billion cubic feet. The 2019 shortage came from plant maintenance outages plus lower BLM sales. All three episodes pushed spot prices up hard, roughly 50 to 150 percent.

What makes 2026 uglier is the combination of factors.

First, the BLM reserve is much more depleted now, so one of the old shock absorbers is weaker. Second, the Strait of Hormuz is a single physical chokepoint, not a bunch of scattered technical outages. Recovery depends on geopolitics, not just restarting equipment. Third, demand has become more helium-dependent in exactly the wrong places: more MRI installations, more EUV semiconductor production, more quantum computing labs.

The likely result is a shortage that moves faster and less evenly than the older ones did. Asia-Pacific probably feels it first and hardest because it is more exposed to Qatari supply than North America is. Japan and South Korea stand out here, especially where hospitals and semiconductor fabs are competing within the same regional supply picture.

What to do in the next 7, 30, and 90 days

This part does not need much dressing up.

In the next 7 days

If you manage helium procurement, ask your distributor for allocation status in writing. Specifically ask whether force majeure clauses have been triggered on your contract. That tells you very quickly whether your “normal supply relationship” is still normal.

Over the next 30 days

Hospitals should figure out which MRI systems actually have helium recovery and which do not. Do not assume your network knows this cleanly. A surprising number of organizations have the data scattered across service vendors, facility engineering, and radiology admin.

If a system does not have recovery, start active boil-off monitoring and define an inventory threshold that triggers reduced scan scheduling before you hit a crisis point.

Fabs and labs should do leak audits. This is not glamorous work, but it is probably the fastest way to extend inventory. In older systems, 10 to 20 percent losses from undetected leaks are not rare.

Looking out 90 days

Evaluate reliquefaction seriously. Portable helium reliquefiers from suppliers like Quantum Design or Cryomagnetics can capture boil-off and turn gaseous helium back into liquid on-site. Capex is roughly $150,000 to $400,000 depending on size. At spot prices around $20 to $30 per liquid liter, up from a current-ish $10 to $15 in normal conditions, the payback gets a lot better very quickly.

For NMR-heavy universities, this is probably already rational. If disruption lasts beyond 60 days, it starts looking urgent.

The cliff arrives late

This is the part helium shortages always seem to hide until the last minute. Things can look mostly normal for 6 to 10 weeks after the disruption starts because inventory buffers are still doing their job. Then those buffers empty, replenishment still is not back, and the shortage suddenly feels real.

The rough sequence to watch is:

• spot prices moving more than 50 percent above contracted rates, usually around weeks 2 to 4
• distributor allocation notices to industrial customers, around weeks 3 to 6
• NMR operating cuts at universities, around weeks 6 to 10
• community hospital MRI capacity problems, around weeks 8 to 14

If the first two are happening, the back half is probably coming unless shipping resumes.

For public data, the USGS Mineral Commodity Summaries helium chapter is still one of the cleaner sources. BLM helium operations reports are useful too. The Compressed Gas Association’s market material is worth checking if you can get access, especially for prior allocation history.

The bigger point, though, is hard to ignore once you see it. A lot of very advanced infrastructure, MRI diagnostics, AI chip manufacturing, quantum research, depends on a gas that is extracted as a side product from one giant gas field and then pushed through one narrow strait. That is not resilience. That is a single point of failure dressed up as a global market.