The Helium Crisis That Could Choke AI's Chip

The technology industry has a helium problem, and most people working in AI don't know it yet.

A missile strike at Qatar's Ras Laffan facility has taken offline roughly a third of the world's helium supply—the same helium that's irreplaceable in advanced semiconductor manufacturing. The plant has been damaged. Some of the infrastructure may take five years to rebuild. And the specialized containers carrying helium to chip fabs in South Korea and Taiwan have a shelf life: 35 to 48 days before the liquid helium vaporizes and becomes useless.

This isn't a story about balloons at birthday parties. This is about whether the trillion-dollar AI infrastructure build can actually proceed as planned.

Why Helium Matters

Analyst Nate B. Jones, in a detailed breakdown of the crisis, explains that helium plays multiple critical roles in chip fabrication that have no substitutes. During plasma etching—when material is scraped from silicon wafers to form transistor structures—fabs blow helium over the back of the wafer to maintain precise temperature uniformity. According to Georgetown University researcher Jacob Fieldboy, whom Jones quotes: "You have to maintain a constant temperature over the wafer being processed to maintain that uniformity. And helium is the only thermal conductor that makes it possible at that scale."

Helium is also the only gas small enough to reliably test vacuum chamber seals. Its atomic diameter is the smallest of any element, which means if helium leaks, you know about contamination problems before they ruin an entire production run.

And crucially, per-chip helium consumption increases with each new generation of semiconductors. The most advanced fabs—the ones producing high-bandwidth memory for AI accelerators—consume between 5,000 and 20,000 cubic meters of helium per month. The purity requirement is extreme: 99.9999% pure, known as 6N grade. Only a handful of facilities worldwide can produce it.

Qatar's Ras Laffan complex was one of them.

The Scope of Disruption

Qatar Energy has acknowledged that 14% of Qatar's helium capacity is permanently damaged, with reconstruction timelines stretching to five years. The country was also planning to bring online Helium 4, which would have been the world's largest helium plant. That timeline, originally set for 2027, is now unknown.

The damage matters because helium production is inextricably linked to liquefied natural gas processing. Helium emerges as a byproduct when LNG is cryogenically distilled. You can't separate the two operations. So the LNG disruption and the helium disruption share the same recovery timeline.

South Korea imported two-thirds of its helium from Qatar, according to the Korean International Trade Association. SK Hynix and Samsung Electronics—the world's two largest memory chip manufacturers—depend on that supply to produce the high-bandwidth memory that goes into every Nvidia GPU, every AMD AI accelerator, every Google TPU.

TSMC in Taiwan faces a different but equally serious problem. The company imports the bulk of its energy as LNG and maintains only 11 days of gas reserves. Taiwan imports 97% of its energy. When asked about the Qatar situation, TSMC told reporters it "doesn't currently anticipate notable impact" and is "monitoring the situation."

That's the kind of statement companies issue when they have to say something but don't want to trigger panic.

The Questions No One Can Answer

Helium consultant Phil Kornbluth offered what Jones characterized as "the most optimistic scenario" at a Gas World webinar: a shutdown lasting two to three months, potentially stretching longer if companies start relocating cryogenic equipment out of Qatar entirely. That kind of supply chain rearrangement would take years.

Helium spot prices have already doubled. Contract surcharges are up 30%.

The central problem isn't whether chip fabs will shut down completely—they won't. These facilities are engineered for resilience, and their supply chain managers are resourceful. The question is whether they'll operate at reduced capacity, whether chip deliveries will arrive late, whether scaling production to meet surging AI inference demand becomes impossible. And how much of the cost increase gets passed to customers.

Jones frames it clearly: "The question for anyone planning here, for whether you're planning for your personal compute budget, your IT supply budget, whether you're planning for a hyperscaler and you're looking at your chips, it's not do the fabs shut down completely. That is not what I'm saying. The fabs will find a way to continue to operate. The problem is do they operate at lower capacity?"

Those second-order effects compound the longer Qatar remains offline. And no one can predict when the conflict ends or when repairs can safely begin under the threat of continued attacks.

The Geopolitical Angle

China appears positioned to benefit from this crisis in ways that will reshape the semiconductor landscape.

The country has been negotiating for years with Russia over the Power of Siberia 2 pipeline, which would deliver LNG from Siberia directly to China. Price disputes have stalled the project. But with Qatar's supply disrupted and China facing the same energy constraints as everyone else, Beijing now has strong incentive to finalize that deal.

If China secures cheap domestic energy through the Russian pipeline while everyone else pays premium prices for constrained global LNG supply, Chinese chip fabrication suddenly becomes more cost-competitive. China is also accelerating domestic helium production. A plant in Guangdong has achieved 6N certification and can now supply ASML-approved helium to desperate fabs.

China's current helium production is tiny—1.2 million cubic meters in a market measured in billions. But it's growing fast, accelerated by this crisis.

The structural outcome: China could control both its energy costs and its helium supply while South Korea and Taiwan—which have no domestic helium sources—remain dependent on a global market that just lost a third of its capacity.

Jones points out that the U.S. ended sales from its Federal Helium Reserve in 2023, removing supply from the market just when it's needed most. Russia can't sell its helium because of ongoing attacks related to the Ukraine conflict. Alternative suppliers exist, but qualifying new helium sources at 6N purity standards takes months to years, not days.

What This Means

I've covered enough technology cycles to recognize the pattern: a single point of failure in a complex global supply chain, invisible until it breaks, suddenly threatening an entire industry's forward momentum.

The AI boom has proceeded on the assumption that compute capacity is primarily a money problem. Hyperscalers issue bonds, sign power purchase agreements, order chips by the millions. Sergey Brin said on record he'd rather see Google go bankrupt than lose the AI race.

But some things can't be solved by throwing money at them. You cannot manufacture helium on demand. You cannot speed up the reconstruction of damaged industrial infrastructure in an active conflict zone. You cannot bypass the physics of chip fabrication that require this specific noble gas.

The companies issuing reassuring statements today will face different questions in four months, in eight months. The planners designing data centers for 2027 and 2028 now need to factor in energy costs that were unthinkable three weeks ago.

And somewhere in Guangdong, a Chinese helium plant that was a footnote in the global supply chain is becoming strategically important.

Bob Reynolds is Senior Technology Correspondent for Buzzrag