Semiconductor Fab Resource Intensity - Not Just Energy

A leading-edge semiconductor fab is not primarily a building full of machines. It is a continuous-operation resource conversion system that transforms electrical power, ultrapure water, process gases, and specialty chemicals into patterned silicon wafers — consuming each of these inputs at rates that have no industrial precedent outside of large petrochemical complexes and aluminum smelters. The fab's resource requirements are not incidental to its operation; they are the primary determinants of where it can be built, how quickly it can reach full production, and what the real cost of semiconductor manufacturing is beyond the capital equipment line.

Understanding fab resource intensity as an integrated system — rather than as separate engineering problems — is essential for semiconductor supply chain analysis. The resource constraints interact: a site with abundant power but insufficient water cannot support a leading-edge fab. A site with power and water but inadequate specialty gas supply infrastructure will face production ramp delays. A fab that solves power, water, and gas supply but lacks adequate heat rejection capacity (cooling towers, dry coolers) will face thermal management constraints on tool density and operating hours. No single resource constraint can be evaluated in isolation from the others — and the compound constraint profile of a proposed fab site determines its viability more accurately than any single infrastructure metric. See: Fab OPS Overview

Resource Consumption Profile — Leading-Edge Fab at Full Production

Resource	Consumption at leading-edge fab (N3–N5, 300mm)	Consumption at mature node fab (28nm–180nm)	Comparator	Site selection constraint character
Electrical power	200–600 MW continuous peak demand; 1–3 TWh/year	40–150 MW continuous; 0.2–0.8 TWh/year	Equivalent to 90,000–270,000 US households continuous; a single EUV scanner draws ~1 MW; cleanroom HVAC is 30–50% of total load	Hard constraint — no substitute; requires dedicated transmission infrastructure; US grid interconnection queues are 3–5 years; competes directly with AI datacenter load growth for the same grid capacity
Ultrapure water (UPW)	10–20 million gallons/day gross production; 2–10 MGD net withdrawal after recycling	2–8 MGD gross; 1–4 MGD net withdrawal	Leading-edge fab gross UPW consumption approaches that of a small city; requires 18.2 MΩ·cm resistivity — the theoretical maximum purity for liquid water	Hard constraint — UPW cannot be purchased or stockpiled; requires reliable local water source year-round; water stress at TSMC Arizona (Sonoran Desert, Colorado River basin) is the most acute resource constraint at any CHIPS Act site
Bulk process gases (N2, Ar, O2, H2, He)	N2: 1–3 million SCFD; Ar: 200,000–500,000 SCFD; He: 20,000–80,000 SCFD; H2: 100,000–300,000 SCFD; O2: 50,000–150,000 SCFD	N2: 200,000–800,000 SCFD; other gases proportionally lower	N2 and Ar supplied by on-site ASU (cannot be trucked at this volume); He is a non-renewable resource — lost to atmosphere permanently on release; He supply is geographically concentrated in Qatar, US, and Russia	Moderate infrastructure constraint — on-site ASU requires 10–20 year supply contract with Air Liquide, Linde, or Air Products; He supply vulnerability is the most acute single-gas risk; ASU construction is a long-lead item in fab commissioning sequence
Specialty process gases (etch, deposition, cleaning)	NF3: hundreds to thousands of metric tons/year per fab; PFCs (CF4, C2F6): hundreds of metric tons/year; SiH4: tens of thousands of kg/year; WF6, TiCl4, TMA: tool-specific quantities	Proportionally lower; fewer etch and CVD steps per wafer; NF3 consumption tracks CVD tool count	NF3 (GWP 17,200) and SF6 (GWP 23,500) make specialty gas supply and abatement the primary Scope 1 GHG story — not electricity; SiH4 is pyrophoric (ignites spontaneously in air); AsH3 and PH3 are acutely toxic at sub-ppm concentrations	Supply chain constraint rather than geographic constraint — specialty gas supply networks exist globally but concentration in specific suppliers (SK Materials for NF3; limited SiH4 and WF6 producers) creates no-stockpile vulnerability; specialty gas abatement is a continuous operating cost and environmental obligation
Process chemicals (acids, bases, solvents, photoresist, CMP slurries)	H2SO4: hundreds of thousands of liters/year; HF: tens of thousands of liters/year; photoresist: thousands of liters/year per node; CMP slurry: millions of liters/year across all CMP steps	Proportionally lower; fewer cleaning and planarization steps; more established supply chain	Photoresist qualification takes 12–18 months — supplier switching is not possible mid-production; CMP slurry continuous-flow requirement means delivery cannot be interrupted without risking defect introduction; HF is acutely toxic at any skin contact level (systemic fluoride poisoning)	Supply chain constraint with qualification lock-in — not a geographic constraint but a procurement and logistics constraint; concentrated supplier base for photoresist (JSR, Fujifilm, TOK — all Japan-headquartered) creates geopolitical supply chain exposure; chemical delivery qualification timelines make diversification slow
Heat rejection (cooling water, dry coolers)	Chiller plant heat rejection: 300–800 MW thermal; cooling tower water consumption: 1–5 MGD evaporative loss; process tool cooling water: dedicated closed-loop systems per tool	Proportionally lower; chiller plant heat rejection 60–300 MW thermal	A leading-edge fab's heat rejection load is equivalent to a mid-sized power plant's cooling requirement; cooling tower evaporative loss adds to the fab's net water withdrawal — relevant for water-stressed sites (Arizona); dry coolers (air-cooled) eliminate water consumption but require more land and electrical energy	Climate-dependent constraint — hot climates (Arizona, Texas) require more cooling energy and more cooling water per unit of heat rejected; temperate climates (Oregon, Germany, Ireland) have lower cooling loads and potential for free cooling; arid sites with water stress face a tension between cooling water conservation and cooling system efficiency
Compressed air (pneumatic systems)	Instrument air: 100,000–500,000 SCFD; tool pneumatic supply: continuous; FOUP transport and robot actuation: continuous	Proportionally lower tool count	Compressed air is a derived energy product — each SCFD of compressed air at 100 psig represents approximately 0.1–0.15 kWh of electrical energy consumed by compressors; compressor waste heat recovery (80–120°C) can offset some facility heating load	Not a site selection constraint — compressed air is generated on-site by electrically driven compressors; no external supply dependency; efficiency improvement through VSD compressors and heat recovery is an operational cost lever

The Compound Constraint — Why Site Selection Is Hard

Each resource constraint in the table above has a different geographic character. Power is a transmission infrastructure problem — solvable in principle anywhere with a 3–5 year grid interconnection timeline. Water is a watershed problem — not solvable by infrastructure investment if the local watershed cannot sustain the required net withdrawal rate. Specialty gas supply is a logistics and supplier network problem — addressable through investment in supply chain infrastructure. Heat rejection is a climate problem — hot arid sites face a structural tradeoff between cooling efficiency and water consumption.

The compound constraint profile of a fab site — the combination of all these requirements simultaneously — is what makes leading-edge fab site selection genuinely difficult and why the apparent abundance of potential CHIPS Act sites does not translate into an abundance of viable sites for leading-edge manufacturing. The table below evaluates the five primary CHIPS Act fab sites against the compound constraint profile.

Fab site	Power	Water	Heat rejection climate	Specialty gas logistics	Seismic	Overall resource profile
TSMC Fab 21 — Phoenix, AZ	Moderate — APS grid expansion required; solar PPA potential strong	Critical constraint — Sonoran Desert, Colorado River basin allocation; 100% recycle commitment required	Challenging — extreme summer heat increases cooling load; dry coolers viable but land-intensive; cooling tower water use conflicts with water stress	Adequate — major industrial gas infrastructure in Southwest; specialty gas supply chain buildout required for leading-edge node volume	Low — favorable; Basin and Range Province geology; primary vibration source is HVAC, not seismic	Strong power and seismic; water is the binding constraint; heat rejection compounded by water stress; the most resource-constrained CHIPS Act site by water metric
Intel Ohio One — New Albany, OH	Challenging — AEP Ohio / PJM interconnection queue congestion; 400–800 MW at full buildout requires new transmission	Adequate — Ohio River watershed; Columbus municipal supply; not water-stressed; Great Lakes basin access	Favorable — temperate climate; free cooling viable for significant fraction of year; cooling tower water not a stress factor	Adequate — Midwest industrial gas infrastructure; ASU siting on campus feasible; specialty gas supply chain buildout at greenfield site	Low — stable craton geology; not a material constraint	Water and heat rejection favorable; power interconnection is the binding constraint given PJM queue congestion; Intel Ohio's primary ramp risk is grid interconnection timeline, not resource availability
Samsung Taylor — Taylor, TX	Moderate — ERCOT islanded grid reduces federal interconnection complexity; wind and solar abundant in Texas; 2021 winter storm vulnerability requires microgrid resilience investment	Moderate — Williamson County is growth-stressed; Little River / Granger Lake supply adequate for current buildout; reclaimed water use targeted for cooling systems	Challenging — hot Texas summers increase cooling load; not as extreme as Arizona; moderate humidity increases cooling tower efficiency vs. dry-bulb-only systems	Strong — existing Samsung Austin S2 fab provides gas supply chain infrastructure baseline; Williamson County industrial corridor has established chemical and gas logistics	Low — stable craton; induced seismicity not a concern at Taylor site	Balanced resource profile with no single binding constraint; grid resilience (ERCOT winter storm exposure) is the primary reliability risk rather than resource availability; existing Austin S2 supply chain infrastructure is a genuine site advantage
Micron Clay — Clay, NY	Strong — National Grid / NYPA; NYPA hydropower access provides low-carbon physical renewable electricity; grid infrastructure adequate for Micron's planned buildout	Strong — Onondaga County; Oneida Lake / Oswego River watershed; Great Lakes basin proximity; not water-stressed; best water supply position among CHIPS Act sites	Favorable — temperate upstate New York climate; low summer design temperature; free cooling viable for majority of year; cooling tower water not a stress factor	Moderate — industrial gas infrastructure in upstate NY is less developed than Southwest or Midwest; requires supply chain buildout; GlobalFoundries Malta (nearby) provides some regional infrastructure precedent	Very low — Laurentian Craton; best seismic profile among CHIPS Act sites	The strongest overall resource profile among CHIPS Act sites — favorable on power (NYPA hydro), water, heat rejection, and seismic simultaneously; specialty gas logistics buildout is the primary infrastructure gap; the best physical decarbonization pathway of any CHIPS Act site
GlobalFoundries Malta — Malta, NY	Adequate — National Grid service territory; NYPA access; mature infrastructure from existing fab operation since 2012	Strong — Hudson River watershed; Saratoga County water supply; not water-stressed; established water treatment infrastructure from existing fab	Favorable — temperate upstate New York; free cooling applicable for significant portion of year	Strong — established specialty gas supply chain from 14+ years of fab operation; Air Products and Linde on-site or adjacent infrastructure; the most mature gas supply chain among CHIPS Act sites	Very low — same Laurentian Craton as Micron Clay	Strong resource profile with established infrastructure advantage from existing fab operations; specialty gas supply chain is the most mature among CHIPS Act sites; expansion constrained by fab technology node (mature node, not leading-edge) rather than resource availability

Heat Rejection — The Underappreciated Resource Constraint

Heat rejection is the resource constraint that receives the least attention in semiconductor policy and supply chain analysis — and is increasingly material as fab power density and power consumption per unit of floor area increase at leading-edge nodes. Every watt of electrical power consumed in a fab is ultimately converted to heat that must be removed from the facility. A leading-edge fab consuming 400 MW of electricity rejects approximately 350–380 MW of heat to the environment (the remainder leaves as product and waste streams). This heat rejection requirement shapes the cooling infrastructure of the fab as fundamentally as power consumption shapes its electrical infrastructure.

Heat rejection technology	Operating principle	Water consumption	Climate sensitivity	Fab application
Evaporative cooling towers	Hot condenser water from chiller plant is cooled by evaporating a fraction of the water in the tower; latent heat of evaporation carries heat to atmosphere; most thermodynamically efficient cooling method in hot dry climates	1–5 MGD evaporative loss at leading-edge fab scale; plus blowdown (1–2 MGD) to control dissolved solids concentration; total cooling tower water consumption is the second-largest water use at a fab after UPW	Performance degrades at high wet-bulb temperature (humid conditions); performance improves in hot dry conditions (Arizona cooling towers are effective despite hot ambient); efficiency highest when wet-bulb temperature is low	Primary heat rejection at most leading-edge fabs; chiller plant condenser water cooling; multiple tower cells with N+1 redundancy; tower blowdown treatment and ZLD increasingly required at water-stressed sites
Dry coolers (air-cooled heat exchangers)	Hot fluid circulated through finned heat exchanger coils; ambient air blown across fins by fans; heat rejected to atmosphere without water evaporation; thermodynamically less efficient than evaporative cooling at equivalent ambient temperature	Zero water consumption — the primary advantage at water-stressed sites; eliminates cooling tower blowdown and evaporative loss	Performance limited by ambient dry-bulb temperature — cannot cool condenser water below ambient + approach temperature (~5°C); in hot climates (Arizona summer: 45°C ambient) dry coolers cannot achieve the chilled water temperatures required for precision cleanroom conditioning without mechanical refrigeration assist	Process tool cooling circuits (closed-loop, separate from cleanroom HVAC cooling); data hall and UPS cooling where water elimination is prioritized; hybrid systems pairing dry coolers with adiabatic pre-cooling (water mist on air inlet) reduce water consumption by 80–90% vs. full evaporative cooling
Free cooling (economizer)	When outdoor wet-bulb or dry-bulb temperature is sufficiently low, cooling tower or dry cooler can provide condenser water cool enough to satisfy chiller plant demand without operating compressors; chiller compressors are bypassed, reducing electrical consumption by 70–80% during free cooling hours	Same as underlying cooling tower or dry cooler; free cooling is an operating mode, not a separate hardware type	Free cooling hours per year depend strongly on climate: Oregon (Hillsboro): 3,000–4,000 hours/year; Germany (Dresden): 2,500–3,500 hours/year; Arizona (Phoenix): 500–1,000 hours/year; Texas (Taylor): 1,500–2,500 hours/year; temperate climates capture significantly more free cooling hours	Incorporated into chiller plant controls at all fabs with cooling towers; free cooling reduces chiller plant electrical consumption and extends chiller equipment life; temperate climate sites (Oregon, upstate New York, Germany) capture the most free cooling benefit — a genuine OPEX advantage over desert or subtropical sites
Waste heat recovery	Rather than rejecting all process heat to atmosphere, capture high-grade waste heat (80–160°C from compressors, thermal abatement, and some process tools) for facility heating, UPW system heating, or absorption cooling; reduces both heat rejection load and heating energy consumption	Reduces cooling tower load proportionally to heat recovered — indirect water consumption reduction	Most beneficial in cold climates where facility heating demand is high; less beneficial in hot climates where cooling demand dominates and heating demand is low; semiconductor fabs in Oregon, New York, Germany, and Ireland have the strongest case for waste heat recovery integration	Compressor waste heat recovery for facility heating (80–120°C); thermal abatement exhaust heat recovery (300–600°C exhaust — high-grade heat but chemically contaminated, limiting recovery options); district heating from fab waste heat (TSMC and Intel European sites exploring community heat supply from fab waste heat)

Resource Intensity Trends — Node Scaling Effects

Resource consumption per fab does not scale linearly with process node advancement. Each new process generation increases the number of process steps per wafer, the number and type of gases and chemicals consumed, and the power required per tool — but the wafer output per unit of floor area may also increase through tool throughput improvements. The net effect on resource consumption per wafer start is upward at each node transition, meaning that the resource intensity of the global semiconductor industry increases with both volume growth and node advancement simultaneously.

Resource	Node scaling driver	Trend per wafer start	Trend per fab
Electrical power	EUV scanner introduction (~1 MW/scanner); High-NA EUV (~1.5–2 MW/scanner); increasing plasma tool count per wafer at advanced nodes (more etch and deposition steps); HVAC energy scales with cleanroom classification stringency	Increasing — each node transition adds process steps and higher-power tools; EUV introduction was the largest single-node power intensity step change in fab history	Increasing — fab peak demand has grown from ~100 MW (180nm era) to 400–600 MW (N3–N5 EUV era); projected to exceed 600 MW at 2nm fabs with High-NA EUV
Ultrapure water	More cleaning steps per wafer at advanced nodes; more CMP steps (each requiring UPW rinse); EUV developer and rinse chemistry adds UPW demand; advanced node wafer cleaning specs are more stringent, requiring more rinse cycles	Increasing — 5nm node wafers require approximately 2–3× more UPW per wafer than 28nm era wafers due to additional cleaning and CMP steps	Increasing — leading-edge fab UPW consumption has grown proportionally with process complexity; recycling rate improvement partially offsets gross consumption growth but net withdrawal still increases
NF3 (CVD chamber cleaning)	More CVD and ALD process steps per wafer at advanced nodes (each step requires chamber cleaning); GAA nanosheet architecture at 2nm adds ALD cycle count significantly; chamber cleaning frequency increases with film deposition rate and batch size	Increasing — NF3 consumption per wafer tracks CVD step count directly; 3nm node fabs consume significantly more NF3 per wafer than 7nm era fabs	Increasing — NF3 is the highest-volume high-GWP gas at leading-edge fabs and its consumption is growing; this is the primary driver of the semiconductor industry's Scope 1 GHG growth even as Scope 2 improves through renewable procurement
Process chemicals (photoresist, CMP slurry)	EUV multi-patterning stacks require more resist coat/expose/develop/strip cycles per wafer; more CMP steps per wafer at advanced nodes; EUV resist is more expensive and consumed at lower volumes per coat but at more coat steps total	Mixed — EUV resist volume per coat is lower than ArF resist (thinner films) but coat step count increases; CMP slurry consumption increases with CMP step count; net chemical cost per wafer increases at each node	Increasing cost intensity — chemical cost per wafer start has increased consistently with node advancement; photoresist cost at EUV nodes is 5–10× ArF resist cost per liter, with comparable or higher total volume requirements
Heat rejection	Proportional to power consumption — as fab electrical load increases with node advancement, heat rejection requirement increases in direct proportion; EUV scanner waste heat is a significant new heat source at advanced nodes	Increasing — same driver as electrical power; heat rejection per wafer increases with power per wafer	Increasing — chiller plant capacity requirements have grown proportionally with fab power consumption; cooling tower water consumption has grown accordingly, compounding the UPW water demand growth at water-stressed sites

Resource Infrastructure as a Competitive Moat

The resource intensity of leading-edge semiconductor manufacturing is itself a competitive barrier to entry that reinforces geographic concentration of production capacity. Establishing the supply chain infrastructure for power, UPW, bulk gases, specialty gases, and chemicals at a greenfield site — and qualifying each of those supply chains to the tolerances required for leading-edge process technology — takes 3–5 years and represents billions of dollars of infrastructure investment beyond the fab building and equipment cost. This infrastructure investment is largely non-recoverable if the fab is repurposed or closed: the on-site ASU, UPW plant, and chemical distribution system have limited alternative uses.

Established fab clusters — Taiwan's Hsinchu Science Park, South Korea's Hwaseong-Pyeongtaek corridor, Japan's Kumamoto-Kyushu cluster — benefit from decades of accumulated resource infrastructure: multiple competing gas suppliers with on-site ASUs and cylinder filling stations, established specialty chemical distribution networks with local inventory, grid infrastructure designed for semiconductor loads, and water treatment and discharge infrastructure sized for fab operations. A new fab at an established cluster site can access this existing infrastructure immediately. A greenfield fab in a non-traditional location — central Ohio, Sonoran Desert Arizona, upstate New York — must build the equivalent infrastructure from scratch, in parallel with fab construction, on timelines that become a critical path constraint to production ramp.

Fab Resource Detail Pages

Resource	Detail page	Coverage
Electrical power	Fab Power	Demand benchmarks by fab type; load distribution by tool category; power quality requirements; delivery architecture; node escalation; CHIPS Act grid interconnection status
Power resilience and microgrids	Microgrids	BESS as primary power conditioning asset; microgrid topologies; blackstart, islanding, grid-forming inverter operation; DER integration; load prioritization hierarchy
Ultrapure water	Ultrapure Water (UPW)	UPW purity specification; 6-stage production train; consumption benchmarks; wastewater segregation and treatment; water stress site selection table; supplier ecosystem
Bulk and specialty process gases	Gas Delivery Systems	Gas categories and hazard profiles; three-tier delivery architecture; VMBs and gas cabinets; purification technologies; mass flow controllers; safety systems; on-site ASU and H2 generation
Process chemicals	Chemical Delivery Systems	Chemical categories and process roles; three-tier delivery architecture; photoresist delivery precision; CMP slurry continuous-flow constraint; safety systems; piping materials; supplier ecosystem
Process gas emissions and abatement	Emissions & Abatement	GWP reference table; abatement system engineering (thermal oxidizer, plasma scrubber, wet scrubber, catalytic oxidizer); the CF4 destruction problem; alternative chemistry; regulatory framework
Electrification and decarbonization	Electrification & Decarbonization	Scope 1/2/3 framework; residual fossil fuel electrification targets; hard-to-abate thermal processes; RE100 commitments vs. grid reality; IRA and CHIPS Act incentive alignment; CHIPS Act site decarbonization comparison
Vacuum systems	Vacuum Systems	Pressure regimes by process step; dry pump and TMP technology; tool-dedicated vs. centralized vacuum architecture; sub-fab layout; vacuum-abatement integration; supplier ecosystem

Cross-Network — ElectronsX Coverage

Semiconductor fab resource intensity connects to EX's coverage at multiple points. The power consumption story (1–3 TWh/year per leading-edge fab) connects to EX's grid demand analysis. The water stress story at TSMC Arizona connects to EX's coverage of water as a constraint on electrification buildout — gigafactories, data centers, and fabs all face the same compound challenge of siting large industrial facilities in regions chosen for power and incentives but sometimes constrained by water. The heat rejection story — cooling towers competing with UPW for water at water-stressed sites — is a systemic pattern that EX's industrial facility coverage addresses across the electrification buildout.

EX: Grid Overview | EX: Facility Electrification | EX: Electrification Bottleneck Atlas | EX: Industrial Electrification | EX: BESS Overview

Related Coverage