The Plant Is 1,200 Miles From the Fab

There is a number that describes the American market for ultra-high-purity hydrochloric acid better than its size does. It is 1,200 — roughly the road miles between the Arizona fab cluster, where TSMC runs a leading-edge fab, and the Texas, Louisiana and Mississippi chlor-alkali corridor where almost all of the country’s hydrochloric acid is made. Columbus sits about 1,100 miles from that corridor. Albany, 1,600. Boise, 1,900.

Pheonix Research sizes the market at USD 15.83 million on 4,516 tonnes in 2025, reaching USD 37.40 million on 7,760 tonnes by 2033, in the new report on US ultra-high-purity hydrochloric acid. Value grows 11.35 per cent a year against 7.00 per cent in volume. But the size is not the interesting part. The interesting part is why the geography looks like that, and what it says about where the value in this chemistry actually sits.

The acid exists because the chlorine needs somewhere to go

Hydrochloric acid is not made because someone wants hydrochloric acid. More than 90 per cent of US output comes out of chlorination chemistry — around 2.3 million Mg a year — where the acid is the acceptable form in which to dispose of chlorine left over from making something else. Chlorinated organics. Water treatment. The acid is a disposal route with a market attached.

So the supply curve is set somewhere a fab never looks. If you buy ultra-high-purity HCl, your supply is a byproduct stream whose volume is decided by demand for chlorinated solvents and municipal disinfection, on a coast, in plants built decades before the first Arizona fab broke ground. The same structural oddity runs through sulfuric acid, where output follows smelter and alkylation demand rather than the buyer’s need. It is the defining feature of the acid business, and the reason nobody can order a plant into existence beside a customer.

Sixty-two per cent of every tonne is water

At 36 to 38 per cent concentration, water is the majority of the product. Freight therefore has a real vote, and it is the argument people reach for first. It is also the weakest of the three reasons this market is local — because hydrochloric acid is dense, non-hazardous to move in the ordinary sense, and cheap per tonne. The reason the plant does not follow the fab is not the shipping bill.

The certificate is the product

The reason is contamination, and contamination is a function of distance. A sub-ppt acid is a specification, not an inventory item. Grades run from 1 ppm through 1 ppb to 1 ppt, and what a fab qualifies is the whole chain that produced them: which container the acid travelled in, how many times it was transferred, how long it sat where, which analytical method confirmed it on arrival. A three-day dwell in a tote is not a neutral event. It is a change to the product.

Which is why three grades of one molecule price as three different products. Electronic grade clears at roughly USD 1,600 a tonne, semiconductor grade at about USD 2,800, ultra or VLSI grade at about USD 5,500. Ultra grade is 35 per cent of 2025 volume and 54.92 per cent of value, and that mix — not tonnage — is what produces the 4.35-point spread between value growth and volume growth.

The same logic runs through the other wet chemistries on the same bench. Purified wet-process phosphoric acid and thermal phosphoric acid are bought against the hot phosphoric nitride strip at around 160 °C, where the specification is the entire purchase. Phosphorus oxychloride is the liquid dopant source loaded into a furnace tube for phosphorus diffusion — the other point in a fab where a certificate rather than a commodity changes hands.

What is being built near the fabs is the last mile

Two announced projects sit exactly where the demand is, and neither of them makes the molecule. KPPC Advanced Chemicals at Casa Grande: Phase 1 above USD 120 million, production from late 2027, roughly USD 500 million of revenue by 2035, with TSMC Arizona, Intel and Micron named as customers. Solstice, the former Honeywell electronic chemicals business, at Chandler: 200,000 square feet and a dedicated 50 MW substation. Purification, dilution, packaging, analysis, delivery. That is the last mile of a chain whose first mile is a chlor-alkali complex 1,200 miles away.

The first purpose-built American UHP HCl plant is on the coast too. Regional Acid Production at Port Lavaca, Texas — announced in May 2025, operations targeted for early 2026, 1 ppm, 1 ppb and 1 ppt grades in IBCs and HDPE drums — was sited next to the chlorine, not next to the customer. That is the tell. Even a plant built explicitly for semiconductor-grade acid chose the byproduct stream.

The size of the prize, stated plainly

A 100,000 wafer-per-month fab consumes 312 tonnes a year, about USD 1.09 million. The building it sits in costs billions. That ratio is the market’s entire character. The 2033 market is USD 21.57 million of new annual revenue over eight years, against tens of billions of announced fab capex in the same window.

It is price-inelastic for a mundane reason — it is a rounding error on the fab’s cost sheet — and unforgiving for a serious one: failure is measured in scrapped wafers, not in a negotiation. Nobody argues hard over a line item that small until it fails a certificate, and by then the argument is not about price.

Where the forecast can break

Two levers move this number, and only one of them is demand.

The first is intensity: how much acid a wafer actually takes. At 0.15 kg per wafer the 2025 market is USD 9.9 million. At 0.36 kg it is USD 23.7 million. Nothing else in the model shifts the answer by a factor of two. The report uses 0.24 kg, derived from the SC-2 ratio and single-wafer dispense volumes and blended across leading-edge and mature flows, and states the range rather than presenting the midpoint as measured.

The second is capacity timing. The build starts from SEMI’s 33.6 million wafers per month of global installed capacity in 2025 (200mm-equivalent), takes a 10.5 per cent US share, converts at the 2.25 area ratio to 18.82 million 300mm-equivalent wafer starts, and adds about 660,000 300mm-equivalent wafers per month of announced US capacity by 2033, net of disclosed slippage. The slippage is not a footnote. Intel’s first Ohio fab was announced for 2025 and now points at 2030 or 2031. Samsung’s Taylor fabs point at 2027 for meaningful volume. Two 65,000 wafer-per-month fabs arriving five years late removes roughly 1,900 tonnes — 40 per cent of the 2025 base — from the early window.

That is why the bottom-up lands at about 11.5 per cent of world capacity by 2033, against the SIA and BCG trajectory of 14 per cent by 2032 that the industry quotes. Tool spending is the upstream driver of all of it, and the estate sizes that separately in semiconductor manufacturing equipment.

The node decides the price, the wafer decides the volume

Volume here follows wafer starts. Price follows node mix, because it is the leading edge that consumes ultra grade. That is what EUV lithography represents in this market: not a competing forecast, but the reason the grade mix tilts. Every wafer that moves to a node defined by extreme-ultraviolet lithography is a wafer whose cleaning sequence will not be satisfied with electronic grade.

Which is also why a 4,500-tonne national market supports the pricing it does. The volume is trivial. The specification is not.

Three things that would change the answer

1. Inland chlor-alkali capacity. If a chlorine source were ever built near the fabs, the dwell-time logic would loosen. A market of this size cannot justify one, so it would have to arrive for another reason entirely — which is exactly why it has not.
2. A second qualified source per fab. Fabs qualify supply chains and prefer two of them. Every new qualification is a share shift between suppliers, not market growth; it moves who earns the USD 1.09 million, not how large it is.
3. Grade migration faster than capacity migration. If ultra grade takes share faster than US capacity grows, the value line detaches from the volume line. That is the 4.35-point gap, and it is the only mechanism in the model that works in this market’s favour without a single new wafer.

The molecule is a commodity. The certificate is not.

Everything about this market follows from one asymmetry: the acid is abundant, cheap and made for reasons that have nothing to do with semiconductors, while the certificate that makes it usable is scarce, slow to earn and easy to lose. The chlor-alkali corridor will keep making hydrochloric acid whatever happens to American fab construction. What will not happen on its own is somebody standing between that corridor and an Arizona fab with a qualification file, an analytical lab and a clean container.

That is the business. Not the molecule — the distance between the molecule and the specification, and the discipline of keeping it short.