Your Next GPU Needs This Noble Gas (And Iran Just Blew Up 30% of It)
Nobody thinks about helium until it’s gone.
You think about silicon. You think about TSMC and EUV lithography and NVIDIA’s latest ridiculous GPU. You think about DRAM prices and the Strait of Taiwan. You do not, at 3 AM, lie awake worrying about a colorless noble gas that makes balloons float.
You should start.
What Just Happened
On March 12th, Iranian drone strikes hit Ras Laffan Industrial City in Qatar. If you don’t know what Ras Laffan is, it’s roughly equivalent to “what if you put Saudi Aramco, BASF, and half of OPEC in one zip code.” It’s Qatar’s main industrial hub — home to QatarEnergy’s LNG processing facilities, petrochemical plants, and critically, the world’s largest helium liquefaction operations.
QatarEnergy immediately declared force majeure. Translation: “we cannot fulfill our contracts, and legally, it’s not our fault.” Which is technically true when someone shoots drones at your facility.
The result: approximately 30% of global helium supply has been suddenly removed from the market. SK Hynix — one of the world’s largest DRAM manufacturers — has maybe two weeks of inventory before the shortages start hitting their fabrication lines. They’re now “forced to diversify,” which is a very polished way of saying they’re scrambling.
South Korea gets 64.7% of its helium from Qatar. Samsung and SK Hynix together produce most of the world’s memory chips. You do that math.
Wait, Helium? For Chips?
Yes. Helium. The stuff from party balloons.
Here’s what most people don’t know: helium is not optional for semiconductor manufacturing. It’s not some niche material you can substitute with whatever’s on the shelf. It’s deeply embedded into the fabrication process:
Wafer cooling. During lithography, silicon wafers need to be held at precise temperatures. Helium’s extraordinary thermal conductivity makes it ideal for transferring heat away from wafer chucks. Try doing 3nm node lithography when your temperature control is off by half a degree.
Inert purging. Many process chambers need an ultra-pure inert atmosphere. Argon works for some of this, but helium’s lower atomic mass makes it better for certain applications — and it’s what most existing equipment is tuned for.
Leak detection. Vacuum systems in fabs are tested with helium mass spectrometers. You don’t find a leak any other way. A leak in a fab is a bad time.
Crystal growth. The Czochralski process — how you grow silicon ingots — uses helium to control the atmosphere and cooling during crystal pulling.
Ion implantation. Some implanter architectures use helium as a cooling medium. No helium, no implanting dopants into wafers, no transistors.
Every major fab in the world — TSMC, Samsung, SK Hynix, Intel, Micron — runs on helium. And about a third of that helium just stopped flowing.
The Supply Problem Nobody Prepared For
Here’s what makes this particularly maddening: this wasn’t unforeseeable. Qatar controls roughly a third of global helium because its North Field — the world’s largest natural gas reservoir — contains unusually high concentrations of helium (around 0.3% by volume, which is genuinely rare). When Qatar massively expanded its LNG operations over the last 20 years, it became the dominant helium supplier almost by accident.
The semiconductor industry knew this. The 2012 shutdown of Qatar’s original Helium-2 plant caused serious shortages. Every supply chain analyst worth their spreadsheet had “Qatar single-point-of-failure” circled in red.
Then everybody… kept relying on Qatar anyway. Because the gas was there, the price was right, and the alternative — building supply chains around the US Gulf Coast, Russia’s Amur plant, or Algeria — requires the kind of long-term thinking that quarterly earnings reports don’t reward.
Russia’s Amur helium plant was supposed to diversify global supply when it came online. Then the Ukraine war happened. So much for that plan.
The Two-Week Clock
When Tom’s Hardware says “two-week clock,” they mean it literally. Fabs maintain helium inventories measured in days to weeks, not months. The gas is expensive to store (it requires cryogenic temperatures — ironic, since it’s needed to achieve cryogenic temperatures) and has historically been available on reliable long-term contracts. Building massive buffer storage wasn’t economical.
SK Hynix running out of helium doesn’t mean a press release. It means production lines stop. It means wafers don’t get processed. It means DDR5 and HBM3 don’t get made. It means the GPU in your next workstation doesn’t exist.
The Strait of Hormuz is also disrupted right now, which means shipping alternative helium sources from the US to Korean fabs takes longer than usual. There’s no quick fix here.
The Part Nobody Wants to Say Out Loud
The semiconductor industry spent years worrying about China/Taiwan geopolitics and pandemic disruptions. The risk models all had TSMC seizure scenarios, export controls, rare earth restrictions. The industry hardened supply chains for those risks.
Nobody hardened the helium supply chain because helium feels like a footnote. It’s not an exotic rare earth metal. It’s element number two on the periodic table. It literally floats away into space if you let it escape into the atmosphere.
And yet here we are. A drone strike on a gas processing plant in the Persian Gulf is now a potential disruption to the global production of memory chips, GPUs, and basically everything with a transistor in it.
The fragility isn’t in the obvious places. It’s in the noble gases, the process chemicals, the neon for laser excimer sources (also mostly from Ukraine, by the way — that was a 2022 reminder nobody quite absorbed), the specialty resists, the helium that keeps the whole thing running.
Your chips run on entropy. Specifically, the entropy of assuming everything will be fine.
It won’t always be fine.
Sources: Tom’s Hardware, HN discussion — 547 comments and counting.