Lessons/Process economics/Techno-economic analysis

Techno-economic analysis, from an equipment list to a selling price

How a process engineer turns a flowsheet into a number a decision can rest on, with one hydrogen plant worked through from purchased equipment to the minimum selling price.

Alireza Lotfollahzade, EITUpdated 15 November 2026About 30 minutes to read

Take an illustrative hydrogen plant: $400 million of total overnight capital, 30,000 t of hydrogen a year, $18 million a year of fixed operating cost and $65 million a year of variable cost, most of it electricity. Operating cost alone is $2.77 per kilogram. What the hydrogen has to sell for depends on how you charge for the $400 million, and the four honest ways of doing it give anything from $3.21 to $4.68 a kilogram.

The lowest number divides the capital by the plant life and stops. The highest builds a year-by-year after-tax cash flow and asks what price earns a 10 % return. Everything between is a shortcut to that cash flow, and most of this lesson is about where the difference comes from and which number to quote.

I keep doing these estimates by hand and checking them against my own toolkit, so every number on this page is computed rather than copied. You can open the same plant in the techno-economic analysis tool and change any of it.

01

What a techno-economic analysis answers

A techno-economic analysis (TEA) joins a process design to a cash flow. It answers three questions in order: what the plant costs to build, what it costs to run, and what the product must sell for to pay back both with a return. A fourth question usually follows: at the market price, is the investment worth making?

The answer is only as good as the least certain input, and that is almost always the capital at the screening stage. AACE International classifies estimates by how much of the engineering is done. A TEA built from cost correlations and factors sits at Class 5 or Class 4, which is why it is a tool for comparing options and finding what matters, not a budget.

Table 1. AACE estimate classes for the process industries
ClassDefinition completeTypical useLow rangeHigh range
50–2 %Concept screening−20 to −50 %+30 to +100 %
41–15 %Study or feasibility−15 to −30 %+20 to +50 %
310–40 %Budget authorisation−10 to −20 %+10 to +30 %
230–75 %Control or bid−5 to −15 %+5 to +20 %
165–100 %Check estimate or bid−3 to −10 %+3 to +15 %
Source: AACE International Recommended Practice 18R-97 (2011 revision). Ranges are at about 80 % confidence; each class spans the values given, depending on the technology and the estimator.

The ranges are asymmetric: early estimates are more often too low than too high, so the high side is wider. Keep that in mind every time a number below has three significant figures: the third figure is arithmetic, not knowledge.

02

Purchased equipment cost

Equipment cost rises more slowly than capacity. A vessel's capacity grows with its volume and its cost with its wall area, so doubling the size does not double the price. The standard correlation is a power law in the size that drives cost, such as heat-transfer area, flow or volume:

C₂ = C₁ (S₂ / S₁)ⁿ(1)

With nothing better, n = 0.6, the six-tenths rule. Individual equipment types scatter from about 0.3 to 1.0, so use a fitted exponent when one exists. The DOE/NETL report by Loh, Lyons and White tabulates cost against size for 31 kinds of equipment on a 1998 US Gulf Coast basis, and fitting those points gives exponents with known quality. For a shell-and-tube exchanger, carbon steel, 150 psig, the fit is $66,278 at 3,455 ft² with n = 0.708, from 26 points over 100–70,000 ft², R² = 0.95.

A usable cost needs three more corrections: the dollar-year, the material and the location.

C = C₁ (S/S₁)ⁿ × (CEPCI_now / CEPCI_base) × F_M × F_L(2)

For a 5,000 ft² exchanger the power law gives $86,094 in 1998 dollars. The Chemical Engineering Plant Cost Index was 389.5 in 1998, and my provisional estimate for 2025 is 810.3, a ratio of 2.08, so the exchanger costs $179,106 in 2025. In 316 stainless, with a material factor of 2.9 for static equipment, it is $519,406. Rotating machinery takes a smaller alloy penalty (1.8 for a 316 pump) because less of its cost is wetted metal.

Two things in Figure 2 are worth noticing. The six-tenths rule, drawn through the same 5,000 ft² point, underestimates a 20,000 ft² exchanger by 14 % because the real exponent is 0.708. And past 70,000 ft² the curve bends upward: 150,000 ft² is costed as three parallel exchangers of 50,000 ft² each, $2.74 million, rather than one imaginary exchanger at the extrapolated price. Quietly extrapolating a power law past the largest unit a vendor can build is one of the commonest ways to underestimate a large plant.

03

From purchased to installed

A purchased exchanger sitting on a truck is a small part of what it costs to have it working. Foundations, steel, piping, instruments, electrical, insulation, paint and the labour to install all of them typically make purchased equipment only 15–30 % of the finished plant. There are three ways to add them, in increasing order of effort.

One factor for the whole plant

Lang's 1948 factors multiply delivered equipment cost by 3.10 for a solids plant, 3.63 for solids and fluids, and 4.74 for fluid processing. They are quick and blunt, good to perhaps ±30–50 %.

Distributive factors, item by item

The DOE/NETL method builds the installation from bulk materials, each a fraction of the bare equipment cost, and the labour to install each bulk as a fraction of that material. The fractions depend on service: gas or liquid, temperature above or below 400 °F, pressure above or below 150 psig. Table 2 reproduces the worked example in the report, a gas-to-gas exchanger with a bare cost of $62,000.

Table 2. Installed cost of a $62,000 exchanger by distributive factors
Gas service above 400 °F and below 150 psig; setting labour 20 % of equipment for a heat exchanger. Factors and the $150,245 result from Loh, Lyons and White (2002), Tables 5–6 and Appendix A.

The installed cost is 2.42 times the bare cost. Piping is the largest single bulk, which is typical of fluid plants and is why the distributive factors for a gas plant run higher than for solids handling. The same method can be applied with one service class for the whole plant or with a class per item; the second is more work and gives a pump and a crusher different factors, which they deserve.

A factor from a similar plant

If you have a completed project of similar type, its installed-to-purchased ratio beats both. Record where it came from.

04

The capital ladder

NETL's cost methodology defines five capital levels, each built on the last. The names matter, because published capital numbers are quoted at different rungs and the gap between them is large.

EPCC = BEC × (1 + f_EPC) TPC = EPCC + process contingency + project contingency TOC = TPC + owner’s costs TASC = TOC × (TASC/TOC)(3)
  • EPC cost adds the engineering, procurement and construction contractor's services, 15–20 % of BEC.
  • Total plant cost adds contingency. Process contingency covers an immature technology (0–10 % when commercial, 20–35 % from pilot data, more from the bench). Project contingency covers what a budget estimate leaves out, 15–30 % of BEC plus EPC plus process contingency.
  • Total overnight cost adds owner's costs: pre-production 2 % of TPC, spare parts 0.5 %, financing fees 2.7 %, other owner's costs 15 % (studies, permitting, legal, owner's engineering), plus land.
  • Total as-spent capital adds escalation and the cost of carrying the money through construction. It is in mixed, as-spent dollars; the other four are overnight costs in the dollars of one base year.

The TASC/TOC factor comes from the spend profile. Money committed in year n of a y-year build sits in the project for y − n + 1 years before the plant earns anything, accruing simple interest at the pre-tax weighted cost of capital:

TASC/TOC = Σₙ (1+e)ⁿ⁻¹ pₙ + Σₙ WACC (y−n+1)(1+e)ⁿ⁻¹ pₙ(4)

For NETL's three-year gas-plant profile (10, 60 and 30 % of TOC by year), a 5.14 % real WACC and no real escalation, the factor is 1 + 0.0514 × (3 × 0.1 + 2 × 0.6 + 1 × 0.3) = 1.0925.

Table 3. TASC/TOC factors under NETL assumptions
BuildBasisWACCEscalationTASC/TOC
3 yearsReal5.14 %0 %1.093
5 yearsReal5.14 %0 %1.154
3 yearsNominal7.25 %3 %1.242
5 yearsNominal7.25 %3 %1.289
Computed from equation (4) with NETL-PUB-22580 spend profiles; the nominal three-year case starts its spend two years after the base year, as NETL does, which is why it carries extra escalation. All four match NETL's published values.

05

Operating cost

Split operating cost the way an operator would. Variable cost scales with output: feeds, catalysts and chemicals, utilities, waste disposal. Fixed cost is incurred whether or not the plant runs: labour, maintenance, insurance and local tax, overhead. For many chemical plants the levelised cost is dominated by one or two variable lines, and a capital estimate good to −30/+50 % on top of a feed price guessed to a factor of two is not a Class 4 estimate of anything.

Variable lines are rate × hours × capacity factor × price. Put consumption at full rate and let the capacity factor scale it, so changing the operating time moves every line together.

Labour

Turton's correlation gives operators needed at any moment from the process configuration, where P counts steps handling particulate solids and Nnp counts compressors, towers, reactors, heaters and exchangers:

N_OL = (6.29 + 31.7 P² + 0.23 N_np)^0.5(5)

Twelve non-particulate steps need 3.0 operators per shift; add two solids-handling steps and it rises to 11.7, which says something true about solids. A seat staffed around the clock takes 4.5 to 5 people once holidays, leave and training are counted.

The double-counting trap

Fixed-cost conventions come as sets, and each set assumes what is inside its lines. NREL's biorefinery reports charge a 90 % labour burden on salaries, and that 90 % already covers benefits and payroll overhead. Peters and Timmerhaus's plant overhead includes payroll overhead too. If you add a 35 % benefits burden to salaries first and then apply either overhead, you count benefits twice and take the overhead on the inflated figure.

Four operators a shift, $78,000 salary, 25 % supervision

Salaries: 4 × 4.8 × 78,000 × 1.25 = $1,872,000. NREL: × 1.90 = $3,556,800. Burden added first, then the 90 %: 1,872,000 + 524,160 burden, × 1.90 = $4,552,704, which is 28 % too high.

I found exactly this in an earlier version of my own toolkit, which is why I mention it. Pick one convention, read what its lines contain, and do not add anything it already includes.

Table 4. Three fixed-cost conventions
LineNRELPeters & TimmerhausNETL
Maintenance3 % of installed cost5 % of FCI2 % of TPC
Insurance and local tax0.7 % of FCI2 % of FCI2 % of TPC
Labour charged asSalariesSalariesLoaded (with burden)
Overhead90 % of salaries60 % of labour + maintenance30 % of labour
Administrationin the 90 %20 % of labour + maintenance25 % of labour
Laboratory, supplies—15 % of labour; 15 % of maintenance—
NREL/TP-5100-47764 §5; Peters, Timmerhaus and West, 5th ed., ch. 6 midpoints; NETL-PUB-22580 §4. FCI here is fixed capital investment, taken as TOC.

An independent check

Turton's one-line cost of manufacture absorbs maintenance, overhead and administration into three multipliers. It is a good check on a line-by-line build-up; if the two disagree by more than about 20 %, one of them is wrong.

COM_d = 0.180 FCI + 2.73 C_OL + 1.23 (C_UT + C_WT + C_RM)(6)

For a $100 million plant with $2.5 million of operating labour and $45 million of utilities, waste and raw materials, COMd is $80.2 million a year.

06

The cost of money

Capital is spent before the product is sold, and someone has to be paid for waiting. The capital recovery factor (CRF) is the annual payment that repays one dollar over N years at rate r, the same arithmetic as a mortgage:

CRF = r (1+r)ᴺ / [ (1+r)ᴺ − 1 ](7)

At 10 % over 30 years CRF is 0.1061, so the hydrogen plant's $400 million costs $42.4 million a year to own before tax, against $13.3 million if you divide by the life. That gap is the whole difference between $3.21 and $4.18 per kilogram in Figure 1.

Income tax adds to it. Revenue that repays capital is taxed, but depreciation is deductible. The fixed charge rate (FCR) grosses the capital recovery up for tax and gives back the present value of the depreciation shield, where dₙ is the depreciation fraction in year n:

FCR = CRF/(1−t) − t · CRF · Σₙ dₙ/(1+r)ⁿ / (1−t)(8)

With NETL's after-tax WACC of 4.73 % real, 30 years, a 25.74 % tax rate and 20-year MACRS depreciation, FCR = 0.0707; at 6.54 % nominal it is 0.0885. NETL publishes 0.0707 and 0.0886, the same to rounding. NETL charges the FCR against TASC, not TOC.

Real or nominal, never mixed

A real analysis uses constant dollars of one year and a real discount rate. A nominal analysis inflates every cost and uses a nominal rate. The two rates are linked by the Fisher relation:

(1 + r_nominal) = (1 + r_real)(1 + i)(9)

NETL's 7.25 % nominal WACC is 5.14 % real at 2.01 % inflation. Applying a nominal rate to cash flows that do not inflate overstates the cost; applying a real rate to inflating ones understates it. In a nominal formula method, operating cost that escalates at i is levelised with a factor that NETL gives as 1.384 for 3 % escalation over 30 years at 6.54 %:

LF = CRF × [ 1 − ((1+i)/(1+r))ᴺ ] / (r − i)(10)

07

The cash flow and the minimum selling price

The formula methods are approximations of one thing: an after-tax cash flow, year by year, from the first dollar of construction to the end of life, solved for the product price at which the net present value is zero at the required return. NREL calls the result the minimum selling price, and it is what US national-lab studies report.

Σₜ [ P·Qₜ + Bₜ − OPEXₜ − Taxₜ − Capitalₜ ] / (1+r)ᵗ = 0 → solve for P Taxₜ = t × (P·Qₜ + Bₜ − OPEXₜ − Depreciationₜ − Interestₜ)(11)

The cash flow carries what the formulas cannot:

  • Capital spent over the build, here 8, 60 and 32 % of TOC over three years, the NREL profile.
  • A start-up year at half revenue but 75 % of variable cost and all of fixed cost.
  • Depreciation on the actual schedule, and tax losses in early years carried forward against later profit.
  • Working capital spent at start-up and recovered at the end; land bought, never depreciated, and recovered at the end.
  • Debt drawn during construction accruing interest until the plant starts, then repaid in level payments.

At 10 % the hydrogen has to sell for $4.68 a kilogram, of which $2.77 is operating cost and $1.91 is the capital charge, including tax. That is more than the FCR's $4.59 because the cash flow compounds the cost of carrying construction spend where TASC adds it as simple interest, loses half a year of revenue at start-up, and defers tax deductions it cannot use yet. Each two points on the discount rate move the price by 32 to 39 cents. If the hydrogen actually sells for $5.50, the same cash flow gives an NPV of $140 million at 10 %, an IRR of 14.1 % and a discounted payback in operating year 11.4.

Three errors the cash flow makes easy

When I audited my own toolkit against these rules I found three, and they are worth checking in any model:

  1. Land depreciated. NETL puts land inside TOC, so a model that depreciates TOC depreciates land, and if land is also entered as its own outflow it is paid for twice. Depreciate TOC less land.
  2. A nominal rate on real cash flows. Choosing a nominal basis must inflate costs, not only change the discount rate.
  3. Free construction debt. Debt drawn during a three-year build accrues interest before the first repayment. A useful test: with no tax, borrowing at exactly the discount rate must leave the selling price unchanged. If leverage makes the product cheaper in that test, the loan is being given away.

08

Byproducts and allocation

Most plants make more than one thing: export steam, surplus power, oxygen from an electrolyser. The simplest treatment credits byproduct revenue against cost and charges the rest to the main product. It is the right choice when the byproduct is small.

When credits approach half of total cost, the main product's cost becomes a small difference of large numbers, and the headline is really a forecast of the byproduct price. Then allocate cost between co-products by mass, energy content or market value instead, and say which. The choice can move the answer by a factor of two, and a levelised cost that does not name its allocation rule cannot be reproduced.

Government production credits

A production credit pays a fixed amount per unit made, for a fixed number of years: per kilogram of low-carbon hydrogen, per tonne of CO2 stored. Two things decide what it is worth to the selling price. It stops partway through the life, so it has to be discounted over its own years and then spread over the whole life. And if it arrives as untaxed cash, as a tax credit taken by direct pay or sold to another taxpayer usually does, a dollar of credit is worth more than a dollar of taxable revenue:

K_lev = K × Σ_{y=1..Y} 1/(1+r)ʸ × CRF(r, N) / (1 − t) ÷ Q(12)

Drop the (1 − t) when the credit is taxed as income. For the hydrogen plant, a credit of $3/kg for ten years is worth $2.63/kg in the formula and takes the cash-flow selling price from $4.68 to $2.55/kg as untaxed cash, or to $2.75/kg if it were taxed. The cash flow gives back less than the formula because the lower price pushes the early years into tax losses and defers the depreciation shield.

A credit that large deserves the same warning as a large byproduct: quote the cost with and without it, because the credit can be withdrawn, can expire before the plant does, and usually has eligibility rules that a TEA cannot check.

09

Sensitivity and uncertainty

A single levelised cost is less useful than knowing which three inputs it depends on. A tornado chart moves one input at a time between a low and a high value and ranks the swings. Use ranges that reflect how well each kind of input is known: capital at the class range, commodity prices at ±30 %, consumption from a converged flowsheet at ±10 %, capacity factor from −10 to +3 % because plants rarely beat nameplate. A flat ±20 % on everything is the commonest way to draw a misleading tornado.

For the range of outcomes rather than the ranking, sample every uncertain input at once (Monte Carlo) and report P10, P50 and P90, with the input distributions stated. And to turn the question round, solve for the input that hits a target: what electricity price gives hydrogen at $3 a kilogram? That break-even is often the most useful sentence in the report.

10

Before you quote a number

  1. Name the costing method, the discount rate, real or nominal, the tax rate and the dollar-year in the same sentence as the number.
  2. Say whether capital is TOC or TASC, and charge a fixed charge rate only against TASC.
  3. Check that depreciation and interest appear nowhere in operating cost.
  4. Read what your fixed-cost convention already contains before adding burden, overhead or administration.
  5. Look for equipment sized outside its correlation range, and for installed quotes multiplied by an installation factor.
  6. Run all four methods. If straight-line is close to the cash flow, the discount rate is near zero or the capital is negligible; if the FCR is far from the cash flow, the tax or depreciation assumptions differ.
  7. Report the estimate class and the three inputs that move the answer most.
  8. For a first-of-a-kind plant, say that the result is an nth-plant floor.

Try it on your own plant

The tool runs this whole lesson on any equipment list: the correlations, the capital ladder, the four methods, the cash flow and the tornado, with the working shown and every figure downloadable.

Sources

  1. NETL (2021). Quality Guidelines for Energy System Studies: Cost Estimation Methodology for NETL Assessments of Power Plant Performance, NETL-PUB-22580. PDF
  2. Loh, H.P., Lyons, J. and White, C.W. (2002). Process Equipment Cost Estimation, Final Report, DOE/NETL-2002/1169. OSTI
  3. Humbird, D. et al. (2011). Process Design and Economics for Biochemical Conversion of Lignocellulosic Biomass to Ethanol, NREL/TP-5100-47764. OSTI
  4. AACE International (2011). Recommended Practice 18R-97, Cost Estimate Classification System as Applied in Engineering, Procurement, and Construction for the Process Industries.
  5. Turton, R. et al. Analysis, Synthesis and Design of Chemical Processes, chapter 8.
  6. Peters, M.S., Timmerhaus, K.D. and West, R.E. Plant Design and Economics for Chemical Engineers, 5th ed., chapter 6.
  7. Lang, H.J. (1948). Simplified approach to preliminary cost estimates. Chemical Engineering 55(6), 112–113.
  8. IRS Publication 946, How to Depreciate Property, Table A-1.
  9. Kirkland & Ellis (2025). One Big Beautiful Bill Act brings big changes to green energy tax credits (45V and 45Q after the July 2025 act). Kirkland
  10. Lotfollahzade, A. (2026). teakit: techno-economic analysis of chemical process plants, version 2.0.1. GitHub