Before the 2026 Strait of Hormuz crisis became another oil-price story, about one-third of the world’s helium moved through that same narrow waterway. That is not an energy footnote. Helium is one of the gases semiconductor fabs use to keep advanced chip manufacturing stable, clean, and cold enough to run. Qatar accounted for roughly 33% of global helium supply before the disruption, which means the AI chip supply chain entered this crisis with a dependency that was physical, concentrated, and easy to ignore until ships stopped moving normally.[1]
The Strait of Hormuz disruption should not begin and end with crude. Oil matters, of course. In early March 2026, coverage of the crisis centered on oil above $110 per barrel, roughly 20% of global oil supply at risk, 170 stranded ships, an 18% decline in air cargo, and freight-rate increases of 25–35%.[2] Those are serious numbers. They are also the familiar part of the story.
The less convenient part is that the AI buildout is not made from compute alone. It requires specialty gases for fabs, sulfur and sulfuric acid for copper refining, copper for power infrastructure, maritime access for chemicals, and capital flows from Gulf exporters that have become important buyers and funders of AI infrastructure. Hormuz is not merely a route for hydrocarbons. In 2026, it is a stress test for the materials ledger behind AI.
The Crisis Is Dated, But The Exposure Is Not Abstract
The relevant window runs from March through July 2026. By March, the disruption had already produced the oil, shipping, and air-cargo effects that made the Strait of Hormuz visible to general markets.[2] A ceasefire announced on April 8, 2026 did not return the system to normal; by July, shipping had not fully resumed and the situation remained fragile.[1]
Rerouting only solves a narrow slice of the problem. Pipeline alternatives can move only 15–20% of the oil usually shipped by sea, and there are no meaningful strategic reserves for many chemicals and critical materials that move through the same disruption zone.[3] That distinction matters. Oil markets have strategic reserves, substitution pathways, and a deeply watched futures curve. Semiconductor gases and chemical inputs tend to have thinner buffers, more specialized logistics, and procurement teams that discover the shortage before the market does.
There was also a direct digital-infrastructure signal. Iran struck Oracle and AWS data centers in the region in March 2026, causing service disruptions.[4] Those strikes do not prove that global cloud resilience is failing, and they should not turn this analysis into a cybersecurity piece. They do, however, puncture the comfortable assumption that data-center risk in this crisis is only upstream, indirect, or theoretical.
Helium Is The Fab Constraint Hiding In Plain Sight
Helium’s role in semiconductor manufacturing is easy to understate because it rarely appears in the polished diagrams of AI infrastructure. It is not the chip, the server, the rack, or the power purchase agreement. It is one of the enabling industrial gases without which leading-edge fabrication becomes harder to schedule and harder to stabilize.
That is why Qatar’s share is so important. Before the Iran conflict, about one-third of global helium supply transited the Strait of Hormuz, and Qatar itself accounted for roughly 33% of global supply.[1] A disruption there does not need to eliminate all helium shipments to create trouble. It only needs to force fabs and gas suppliers into allocation decisions: which customers receive contracted volumes, which sites pay for replacement supply, which tools run at full schedule, and which expansion plans quietly slip.
South Korea shows how quickly the issue moves from commodity map to production planning. As of the available 2026 reporting, South Korea said it had sufficient helium reserves only through June 2026, while SK Hynix and Samsung were working to secure supply from the United States.[5] That reserve window should be treated carefully. It is not proof, on July 21, 2026, that Korean fabs are out of helium. It is proof that the buffer identified earlier in the crisis was time-bound, and that buyers had already begun looking outside the Gulf route before the summer.
For AI chips, the operational risk is not usually a cinematic halt where every fab goes dark at once. It is more likely a chain of rationing and reprioritization. Advanced memory and logic customers with stronger contracts or strategic value get protected first. Smaller customers see delivery dates move. Engineering lots wait. Maintenance and qualification schedules become harder to coordinate. The shortage shows up as degraded flexibility before it shows up as an empty building.
This is the part of the AI infrastructure conversation that strategy decks tend to skip. The model roadmap depends on accelerator availability. Accelerator availability depends on advanced fabs and packaging capacity. Those fabs depend on gases and chemicals that procurement teams cannot simply summon from a cloud console when a maritime chokepoint tightens.
Sulfur Turns The Data-Center Story Into A Copper Story
The sulfur channel is less familiar than helium, which makes it more revealing. Nearly half of global seaborne sulfur passes through the Strait of Hormuz.[1] Sulfur is used to make sulfuric acid, and sulfuric acid is essential for copper refining.[1][6] Copper then becomes part of the electrical equipment, cabling, transformers, substations, and grid connections that data centers need before a single GPU can do useful work.
That sequence is not elegant, but it is the sequence that matters: sulfur shipment, sulfuric acid availability, copper refining, electrical infrastructure, data-center energization. A disruption at the first step does not instantly cancel every data-center project. It creates a squeeze in the industrial inputs that feed an already crowded buildout schedule.
| Hormuz-linked input | Immediate industrial use | AI infrastructure consequence |
|---|---|---|
| Helium | Semiconductor fabrication | Less flexibility for advanced chip and memory production schedules |
| Sulfur and sulfuric acid | Copper refining and chemical processing | Pressure on data-center power equipment, cabling, grid connections, and electrification timelines |
| Oil export revenue | Gulf fiscal and sovereign investment capacity | Risk to reportedly committed AI infrastructure funding if disruption is prolonged |
Copper is already one of the least forgiving materials in the AI data-center buildout. The hard part is not just buying servers. It is getting power to the site, moving through interconnection queues, securing transformers and switchgear, and building enough redundancy that the facility can run at the load profile promised to customers. When sulfuric acid supply tightens, the bottleneck does not need to announce itself as an “AI” shortage. It may arrive as delayed copper refining, higher input costs, or longer lead times for the electrical package.
This is also where the usual cloud-native language becomes actively misleading. A hyperscale region is not a placeless abstraction. It is a real estate decision attached to a grid connection, cooling system, substation plan, generator strategy, fiber route, and equipment procurement schedule. If copper-intensive power infrastructure slips, so does the compute capacity that was supposed to arrive after it.
The Cost Shock Does Not Land Evenly
Energy prices still matter, but they do not hit every part of the AI stack with the same force. Chip fabrication is expensive and power-intensive, yet leading fabs have room to absorb some of the shock. TSMC’s energy bill was about $3 billion in 2024, under 3% of revenue, which means even a doubling of power costs would be painful but not structurally fatal for that specific business model.[5]
That fact should not be generalized to every AI infrastructure project. A marginal data-center project in Asia or Europe has a different cost stack, a different revenue certainty, and often a tighter financing case. Available analysis indicates that planned data centers in Asia and Europe could face a 10–20% total-cost increase under the post-closure energy environment, enough to threaten projects already sitting near the edge of viability.[5]
The United States is in a different position. US data centers are partially cushioned by domestic gas and helium sources, and by the contrast between roughly $4 per MMBtu domestic gas and about $13 per MMBtu European and Asian spot prices after the closure.[5] That does not make US projects immune to transformer shortages, copper constraints, or global chip allocation. It does mean that the same Hormuz shock can be a survivable cost increase in one region and a project-killing change in another.
This asymmetry is where procurement and site-planning teams need to be more precise than the market narrative. “AI infrastructure exposure” is not one bucket. A Taiwanese fab buying industrial gases, a Korean memory producer rebuilding helium supply, a European data-center developer exposed to gas prices, and a US hyperscaler with domestic energy access are not carrying the same version of Hormuz risk.
Gulf AI Capital Is A Cash-Flow Question, Not A Write-Off
Only after the physical channels are visible does the capital channel make sense. Gulf AI investment commitments reportedly totaled about $300 billion before the disruption, according to reporting cited in 2026 analysis.[5] That figure should be handled as reported committed capital, not as a guaranteed loss and not as proof that every announced project is equally exposed.
The mechanism is straightforward. Prolonged oil export disruption crimps the cash flows that support public budgets, sovereign investment vehicles, and large-scale industrial policy commitments. Some AI projects may already have secured funding. Some may be politically protected. Others may depend on future disbursements, vendor financing, or procurement phases that become easier to slow than to cancel publicly.
This is not the same kind of risk as helium allocation or sulfuric acid availability. A fab gas shortage creates a production constraint. A copper-refining bottleneck creates an equipment and construction constraint. A sovereign cash-flow squeeze creates a sequencing constraint: which projects move first, which are deferred, which are resized, and which vendors wait longer for purchase orders that were previously treated as bankable demand.
What To Track While The Situation Keeps Moving
The practical response is not to declare that AI infrastructure is doomed. The buildout has too much strategic backing, too much customer demand, and too many substitution efforts already underway for that kind of conclusion. The response is to stop tracking Hormuz as only an energy headline.
- For semiconductor procurement teams, the live questions are helium source diversification, allocation priority, reserve duration, and whether replacement US supply is contracted or merely being pursued.
- For data-center developers, the questions are copper-linked equipment lead times, sulfuric acid and refining exposure, transformer and switchgear availability, and whether higher regional energy prices break the project model.
- For AI infrastructure investors, the questions are which Gulf-backed commitments are funded, which depend on future hydrocarbon cash flows, and which vendors are counting reported commitments as firm demand.
- For risk teams, the questions need dates attached: March shipping conditions are not July shipping conditions, and a reserve window that lasted through June cannot be treated as permanently open or permanently exhausted without new evidence.
The Strait of Hormuz crisis is not just an energy shock to watch from the side of the AI market. It is a materials-and-capital stress test for the assumptions behind the AI buildout. The useful view is commodity by commodity, region by region, and date by date: helium for fabs, sulfuric acid for copper, copper for power infrastructure, gas prices for project economics, and Gulf cash flows for reported AI commitments.
References
- AGBI / The Asia Group analysis on Hormuz disruption and AI supply chains, AGBI / The Asia Group, July 2026.
- Slimstock analysis on Strait of Hormuz disruption and shipping impacts, Slimstock, March 2026.
- Roland Berger analysis on Strait of Hormuz rerouting capacity, Roland Berger, March 2026.
- Data center service disruption reporting on Oracle and AWS regional strikes, Data Center Dynamics / Amazon, March 2026.
- Epoch AI analysis of AI infrastructure exposure to the Strait of Hormuz crisis, Epoch AI, April 2026.
- World Economic Forum analysis on commodity impacts beyond oil, World Economic Forum.
Comments
Join the discussion with an anonymous comment.