The sand that runs the world
There is one more layer below the machines and the chemicals. The materials. Not the processed chemicals, but the raw inputs — the silicon feedstock, the rare earths, the minerals that everything upstream depends on. Some of those materials really do come from only a handful of places on Earth. And the story of where they are, who controls them, and what "controlling a material" actually means turns out to be far more interesting than the simple narrative about China and rare earths.
There is a popular version of the silicon story that goes roughly like this: silicon comes from sand, sand is everywhere, therefore silicon is not a bottleneck.
That version is almost entirely wrong — it mistakes the raw material for the finished product. The sand in a child's sandpit and the wafer inside a GPU are both technically silicon. The distance between them is several years of processing, multiple countries of specialised manufacturing, and purity requirements that make pharmaceutical-grade anything look casual.
Let's walk through what actually has to happen before silicon becomes a chip — and where the real chokepoints are.
The complete journey #
The silicon supply chain is multi-step. It is five very different ones, each requiring a completely different capability:
Ordinary quartzite (silica rock)
↓
Metallurgical-grade silicon metal
↓
Electronic-grade polysilicon
↓
Single-crystal silicon ingot
↓
300 mm semiconductor wafer
↓
TSMC / Samsung fabrication
↓
AI processor
The first step is genuinely abundant. The last four are not.
Stage 1 — Silicon metal #
Quartzite is reduced with carbon in an electric arc furnace at very high temperatures. What comes out is metallurgical-grade silicon — roughly 98 to 99% pure. This material goes into aluminium alloys, solar cells and various industrial processes.
China accounts for a very large share of global production. But this is a production cost and capacity concentration, not a geological one. Silicon-bearing rock is common. The same reduction process could be run — expensively — in Norway, Brazil, Australia, Canada, the United States and many other countries.
If Chinese silicon metal disappeared from global supply, it would be expensive and disruptive to replace. But the process is not geologically locked to China, and alternatives exist.
The real chokepoint comes in the next step.
Stage 2 — Electronic-grade polysilicon #
Metallurgical silicon is nowhere close to pure enough for a chip. It must be chemically converted, distilled, and deposited again to remove virtually every impurity.
The target purity for semiconductor-grade polysilicon can approach:
99.999 999 999 %
Eleven nines. A level of purity that means roughly one foreign atom per billion silicon atoms.
At this level, the contaminants that matter are not what you would normally measure — they are single atoms of boron, phosphorus, metals, carbon and oxygen that alter the electrical properties of the final crystal. A batch that appears chemically pure by ordinary standards may still be rejected because it behaves differently during crystal growth.
The qualified producers of electronic-grade polysilicon are a short list:
| Company | Country | Key facility |
|---|---|---|
| Wacker Chemie | Germany | Burghausen, Bavaria |
| Hemlock Semiconductor | United States | Hemlock, Michigan |
| Tokuyama | Japan | Tokuyama, Yamaguchi |
| OCI | South Korea | Korean and Malaysian facilities |
A new polysilicon plant must reach the required purity level and also it must demonstrate batch-to-batch consistency, controlled contamination during handling and packaging, and compatibility with the specific crystal-growth processes of its customers. Qualification against a leading wafer producer takes years even after the chemistry works.
Electronic-grade polysilicon belongs on the same list as EUV lithography and advanced packaging: technically concentrated, slow to expand, expensive to replicate.
Stage 3 — The quartz story #
This is where the story gets genuinely strange — and where a widely circulated claim needs to be corrected.
You have probably heard that a small mining district in North Carolina contains the only deposits of sufficiently pure quartz to sustain semiconductor manufacturing, and that a disruption there would halt chip production globally. The story is compelling. It is also significantly overstated.
Here is what is actually true.
The Spruce Pine district in the mountains of western North Carolina does contain unusually pure natural quartz. Two major companies mine and process it there:
- Sibelco* (a Belgian company with US operations at Spruce Pine)
- The Quartz Corp (Norwegian-registered, also operating at Spruce Pine)
And here is the part the popular narrative usually omits:
Spruce Pine quartz is generally not the source of the silicon atoms inside the chip.
What Spruce Pine quartz is used for is crucibles.
Grow a silicon crystal — which we will get to in a moment — and you need to melt several hundred kilograms of polysilicon in a vessel that can withstand temperatures above 1,400°C without contaminating the melt. That vessel is a fused-quartz crucible. To make a crucible that will not leach impurities into liquid silicon at extreme temperatures, you need quartz of extraordinary purity. Spruce Pine supplies a significant share of the quartz used to make those crucibles.
The dependency is real. If the crucible contaminates the melt, the crystal is ruined. But the nature of the dependency is crucibles and furnace components — not silicon feedstock.
Spruce Pine is not the only possible source of high-purity quartz. Synthetic fused silica can be produced, and other natural deposits exist. But Spruce Pine is established, qualified, at industrial scale, and deeply integrated into the supply chains that depend on it. That makes it very difficult to replace quickly — just not for the reason usually given.
Stage 4 — Growing the crystal #
Electronic-grade polysilicon is not a form that a chip fab can use directly. It must be transformed into a nearly perfect single crystal — one continuous lattice of silicon atoms oriented in exactly one direction across an object the size of a large loaf of bread.
The standard method is the Czochralski process,[1] named after Jan Czochralski,[2] the Polish chemist who discovered it in 1916:
Polysilicon loaded into fused-quartz crucible
↓
Heated until fully molten (~1,420°C)
↓
Seed crystal lowered to touch the melt surface
↓
Crystal slowly rotated and pulled upward
↓
Silicon solidifies onto the seed, growing downward
↓
Cylindrical single-crystal ingot forms
↓
Ingot removed, cooled and inspected
The result is a crystal ingot roughly 300 mm in diameter and up to two metres long, weighing hundreds of kilograms. The entire process runs in a carefully controlled atmosphere, free of vibration, with the pull rate and temperature managed to nanometre-level precision.
Then it gets sliced into discs — wafers.
Stage 5 — The 300 mm wafer #
Slicing is only the beginning. Each disc must then be lapped (ground flat), etched (to remove surface damage), polished (to optical-level smoothness), and cleaned to a standard where particle counts are measured in single digits across a 300 mm surface.
The wafer must meet extremely tight specifications for:
- crystal orientation
- diameter and thickness uniformity
- flatness — measured at nanometre scales across the entire surface
- edge profile (to prevent chipping during handling)
- surface roughness at atomic scales
- oxygen concentration — controlled deliberately during crystal growth
- metallic contamination limits far below parts per billion
- resistivity uniformity across the disc
A wafer that looks identical to a qualified wafer may still fail qualification because its oxygen content is slightly off, or because its surface roughness is measurably different at atomic scales, or because it generates a handful of particles during handling.
The leading wafer producers are:
| Company | Country | Position |
|---|---|---|
| Shin-Etsu Chemical | Japan | Global leader in semiconductor-grade wafers |
| SUMCO | Japan | Major 300 mm producer |
| GlobalWafers | Taiwan | Significant integrated producer |
| Siltronic | Germany | Premium ultra-pure wafers |
| SK Siltron | South Korea | Korea's principal wafer supplier |
Japan dominates this list. Shin-Etsu and SUMCO together supply a large majority of global 300 mm semiconductor wafers. Both have built their positions over decades. A new entrant would need to qualify against specific foundry processes — which means years of testing before a single commercial wafer ships.
The actual chokepoint map for silicon #
Reframing the entire silicon supply chain by what is actually constrained:
Geologically concentrated #
| Chokepoint | Companies | Note |
|---|---|---|
| Ultra-high-purity quartz for crucibles | Sibelco, The Quartz Corp (both at Spruce Pine, NC) | Used for crucibles and furnace components, not as the silicon feedstock itself |
Technically concentrated #
| Chokepoint | Key companies | Countries |
|---|---|---|
| Electronic-grade polysilicon | Wacker, Hemlock, Tokuyama, OCI | Germany, US, Japan, South Korea |
| 300 mm semiconductor wafers | Shin-Etsu, SUMCO, GlobalWafers, Siltronic, SK Siltron | Japan, Taiwan, Germany, South Korea |
| Crystal-growth process knowledge | Same wafer companies | Primarily Japan and Taiwan |
Capacity concentrated #
| Chokepoint | Notes |
|---|---|
| Metallurgical-grade silicon production | China holds a large production share, but the process is replicable elsewhere |
The corrected conclusion #
The popular narrative — "silicon is scarce because it only exists in a few places" — is wrong. Silicon-bearing rock is genuinely common.
The correct narrative is this: the constraint is processing, not geology. Achieving semiconductor-grade purity, growing single crystals reliably, and producing wafers at atomic-scale tolerances requires years of accumulated process knowledge, qualified supply chains, and specific manufacturing infrastructure that takes a very long time to rebuild elsewhere.
This is a subtle but important distinction. A country could theoretically import quartzite from many sources. It cannot quickly replicate the process yield, qualification status and manufacturing maturity of Wacker's Burghausen plant or Shin-Etsu's Japanese facilities.
The bottleneck is in the factory.