Why Do Factories Still Burn Fossil Fuels?

Electricity can make almost any temperature. The hard question is whether it can deliver the right heat, into the right product, every hour, without rebuilding the factory's economics.

Last updated September 2026
Figure 1 · Industrial heat is a temperature-and-process map

The same megawatt-hour does different work

Select a representative load. Temperature narrows the technology set; heat-transfer mode, duty cycle, atmosphere, cleanliness, and product tolerance narrow it again.

Food · Commercial

Wash and pasteurise

Representative need
75°C
Duty
Hot water / batch
Candidate
Heat pump + hot-water storage

Integration boundary: Low lift and recoverable refrigeration waste heat can make COP, not element efficiency, the decisive advantage.

Figure 1: Representative screening points, not universal process specifications or equipment limits. A facility assessment must use the product's actual temperature, heat flux, atmosphere, cycle, contamination tolerance, and uptime requirement. Maturity is an editorial classification informed by DOE and IEA technology maps.

The answer in one paragraph

Industrial heat is not one substitution. Below roughly 150°C, heat pumps can turn one electrical unit into several thermal units when a usable heat source exists. Electric boilers and resistance can cover steam and broad temperature ranges with near-unity conversion but inherit the full electricity price. Induction, infrared, microwave, arcs, and plasma can outperform combustion by placing energy closer to the work. At high temperature, device efficiency stops being the useful headline: heat flux, atmosphere, refractory life, furnace geometry, continuous uptime, product quality, and the grid connection set the result.

  • Process heat accounts for roughly half of US manufacturing onsite energy; DOE estimates about one-third of process-heating energy is ultimately lost as waste heat.
  • IEA finds industrial heat pumps are entering commercial operation below 160°C, while systems above 200°C remain demonstrations, prototypes, or concepts.
  • At 2025 US average industrial prices, resistance heat buys electricity at about $86/MWh before losses while gas enters near $18/MWh of fuel energy. Efficiency alone cannot close that spread.
  • Thermal storage can buy power at a different hour and reduce connection peaks. It does not create cheap energy or eliminate the process heat exchanger.
Part I: The temperature map

Start with the load, not the technology

Temperature is the first filter. The process medium, heat-transfer path, operating schedule, and product specification decide whether a nominally capable heater is useful.

Useful process heat≈temperature × flux × uniformity × availability÷energy + integration + disruption cost
RouteUseful domainAdvantageBinding constraint
Industrial heat pump / MVRUsually <160°C commercial; higher systems emergingCOP >1 by upgrading waste or ambient heatTemperature lift, working fluid, heat-source availability
Electric boiler / resistanceSteam and broad direct-heating rangeSimple, precise, near-unity device conversionElectricity price, connection size, element and refractory life
InductionConductive workpieces; melting and heat treatmentRapid, local, high power densityPart geometry, coupling, frequency, coil design
InfraredSurface heating, drying, curingFast response; avoids heating some furnace massLine of sight, surface absorption, penetration depth
Microwave / radio frequencyDielectric and wet materialsVolumetric and selective heatingField uniformity, material response, scale-up
Electric arc / plasmaMelting and very high-temperature dutyExtreme flux and direct energy deliveryElectrodes, refractories, power quality, product chemistry
Thermal batterySteam to high-temperature process heatMoves electricity purchase in timeCycles, discharge rate, heat exchanger, utilisation

Heat dominates the factory energy balance

51%DOE estimate of US manufacturing onsite energy used by process heating in 2018.
~⅓Share of process-heating energy DOE says is ultimately lost as waste heat.
30%Share of process heat served as steam in DOE's 2018 MECS account.

The first project is often not a new heater. It is insulation, air-leak reduction, condensate recovery, heat integration, or using waste heat as the source for a heat pump.

Part II: Where electricity already wins

Low temperature rewards moving heat

A resistance element cannot convert more than one electrical unit into one thermal unit. A heat pump can deliver several because most output is upgraded heat recovered from air, water, refrigeration, exhaust, or another process stream.

Carnot sets the direction, integration sets the COP

The ideal heat-pump ceiling falls as the lift between source and sink widens. Real machines also pay for finite heat-exchanger approaches, compressor losses, pumps, and pressure drop. A plant with warm wastewater and a 90°C hot-water demand is a different proposition from one making 150°C steam from winter air.

IEA's 2026 monitor places commercial industrial heat pumps primarily below 160°C and estimates that currently available systems could technically supply around 20% of global industrial heat demand. Technical potential is not adoption: industrial heat pumps supplied under 1% of light-industry heat in 2025.

Read longitudinally, that gap is the curve: under 1% deployed (Measured) against roughly 20% technically servable with today's machines (Projected), a diffusion curve, not a physics curve. The conversion floors are already in place, so the distance between them is closed by integration, tariffs, and plant economics rather than by a new physical result.

01

Electricity supply

The cheapest annual average can be unusable if the process cannot stop during expensive hours.

Specify
Energy price, carbon intensity, firmness, tariff, and power quality
Acceptance test
Product quality and throughput at operating duty
02

Grid connection

A large furnace can require a site-scale substation before it requires a new heating element.

Specify
MVA capacity, transformer, switchgear, protection, and lead time
Acceptance test
Product quality and throughput at operating duty
03

Power conversion

Transformers, drives, rectifiers, and power electronics add cost, loss, harmonics, and failure modes.

Specify
Voltage, current, frequency, waveform, and controllability
Acceptance test
Product quality and throughput at operating duty
04

Heat generation

Heat pumps, resistance, arcs, induction, infrared, microwave, and plasma solve different loads.

Specify
COP or conversion efficiency, temperature, flux, and turndown
Acceptance test
Product quality and throughput at operating duty
05

Thermal storage

Cheap storage media still need expensive heaters, insulation, heat exchangers, and process interfaces.

Specify
MWhth, discharge rate, temperature, losses, and cycles
Acceptance test
Product quality and throughput at operating duty
06

Heat transfer

A hot element is not useful unless energy reaches the product at the required rate and uniformity.

Specify
Conduction, convection, radiation, phase change, or volumetric absorption
Acceptance test
Product quality and throughput at operating duty
07

Furnace or reactor

Removing combustion changes gas volume, mixing, moisture, chemistry, and sometimes the product.

Specify
Geometry, refractory, atmosphere, seals, material flow, and residence time
Acceptance test
Product quality and throughput at operating duty
08

Controls and quality

A lower energy bill is irrelevant if colour, strength, purity, cure, or metallurgical structure changes.

Specify
Temperature distribution, recipe, yield, throughput, and rejects
Acceptance test
Product quality and throughput at operating duty
09

Operations

Continuous plants price an hour of lost production differently from a batch oven.

Specify
Availability, maintenance, backup, restart, and retrofit downtime
Acceptance test
Product quality and throughput at operating duty
Part III: Why the furnace is part of the process

A flame does more than provide joules

Combustion establishes gas flow, pressure, moisture, oxygen potential, mixing, and radiative conditions. Replacing it can alter the reaction even at the same measured temperature.

Near-unity conversion is not near-unity system efficiency

Resistance heating is already close to its device-level physical floor: nearly all electrical input becomes heat. The remaining frontier lies outside the element: reducing shell loss, avoiding hot exhaust, shortening warm-up, transferring energy directly into the workpiece, recovering rejected heat, and improving yield.

For an electric furnace, the next efficiency gain is often a better process rather than a better resistor.

Electromagnetic methods can change the boundary. Induction deposits heat inside conductive material; microwave and radio-frequency fields can couple volumetrically to suitable dielectrics; infrared can heat a surface without first heating a large air volume. Each can save energy and time, but only when the product's geometry and electromagnetic properties cooperate.

Radiation rises as absolute temperature to the fourth power

High temperature increases radiative heat-transfer potential rapidly, which is useful for moving intense heat into a load. The same law increases radiation through openings and to furnace walls. Refractories, seals, electrode penetrations, insulation, and view factors become system components, not incidental construction.

Part IV: Thermal storage as grid interface

Store heat if the product needs heat

Converting electricity to heat, storing it in brick, particles, salt, water, or another medium, and delivering heat directly avoids paying for a second conversion back to electricity.

Storage is a power-and-energy system

$15/kWhthDOE subsystem cost target for concentrating-solar thermal storage, not a current market average.
>1,300°CTarget temperature in a 1 MWh DOE-funded Rondo subsystem demonstration.
850°CHot-air output reached by Sandia's 2026 prototype solar-thermal test.

These figures have different labels: target, planned subsystem demonstration, and measured prototype outlet. None alone establishes delivered industrial-heat cost.

Four quantities must be priced separately

Charging power

Heaters and the grid connection set how quickly cheap electricity can be absorbed.

Stored energy

Media, container, insulation, and allowable temperature swing set duration and loss.

Heat delivery

Fans, steam generators, heat exchangers, and ducts must meet process temperature and flux.

Utilisation

A low-cost store cycled rarely can add more per delivered MWh than an expensive store used every day.

Part V: High-temperature frontiers

The hotter the process, the less temperature alone tells you

Arcs and plasma can exceed the temperatures of industrial flames. The frontier is delivering that intensity over the required volume and residence time without destroying electrodes, refractories, or product quality.

ESTABLISHED

Electric arc steelmaking

Electricity already melts scrap at industrial scale. Feedstock quality, not attainable temperature, limits which steel products the route can make.

Measured industry operation
PROCESS-SPECIFIC

Induction and resistance

Widely used in melting, heat treatment, silicon, glass boosting, and specialty manufacture where material coupling and geometry fit.

Commercial, not universal
EARLY SCALE

High-temperature storage

Commercial projects and public demonstrations are moving from steam toward hot air above 1,000°C.

Vendor targets remain targets
DEVELOPMENT

Electric kilns and crackers

Cement, chemicals, and mineral processes must reproduce reaction atmosphere, solids flow, and continuous throughput, not only peak temperature.

Pilot evidence
Part VI: Retrofit versus greenfield

The cheapest heater can be the expensive project

Equipment cost is visible. Lost production, site works, grid connection, and requalification are often larger and more uncertain.

Downtime

Tie-ins, demolition, refractory cure, commissioning, and product requalification can cost more than the heater. Plan conversion around a major turnaround.

Footprint and access

An electric unit may be smaller, but transformers, storage, ducting, heat exchangers, and construction access require real estate the existing plant may not have.

Steam architecture

A central boiler serves many pressures and users. Electrifying one load may be easy; retiring the shared header requires every remaining user to move.

Product qualification

Food safety, pharmaceutical validation, coating cure, glass colour, metallurgical structure, and cement mineralogy can make the process recipe a regulated asset.

Reliability

A continuous plant needs redundancy, black-start behaviour, spares, trained maintainers, and an answer for grid interruption.

Asset age

A greenfield furnace can be designed around electric heat. A young fuel-fired asset carries unrecovered capital and geometry optimised around a flame.

Greenfield changes the equation

A new plant can place transformers beside the load, match voltage to heaters, eliminate a steam loop, reuse cooling duty as a heat-pump source, design refractory and airflow around electric heat, and reserve room for storage. Retrofitting inherits every old boundary. This is why asset age belongs in an electrification map beside temperature.

Part VII: Electricity-price sensitivity

The energy-price ratio can overwhelm equipment efficiency

Heat pumps divide the electricity price by COP. Resistance does not. Storage substitutes a flexible electricity price but adds capital, loss, and cycling requirements.

Figure 3 · Interactive delivered-heat model

Electricity price sets the curve; COP bends it

Compare four simplified ways to deliver one thermal megawatt-hour. Change energy prices, utilisation, carbon price, and heat-pump performance. Grid upgrades, retrofit downtime, and process modifications sit outside the boundary.

Lowest modelled cost · 2025 U.S. average$25/MWh-thermal

Gas boiler

Gas boiler
$25/MWh
Resistance / electric boiler
$92/MWh
Industrial heat pump
$41/MWh
Stored electric heat
$66/MWh
Calculation, capital assumptions, and evidence labels

Gas includes fuel at 85% boiler efficiency, EPA's 53.06 kg CO₂/MMBtu combustion factor, a carbon price, $110/kWth illustrative capex, and fixed O&M. Resistance uses 98% device conversion and $180/kWth capex. The heat pump uses the selected COP and $900/kWth capex. Stored heat uses the flexible electricity price, 90% round-trip heat efficiency, $25/kWh-thermal storage in the base case, 300 cycles a year, and $250/kWth charging and heat-delivery equipment. Capital is annualised at 10%.

All equipment and storage costs are editorial assumptions intended for sensitivity analysis, not quotations or market averages. The energy-price anchors are measured national averages; DOE's $15/kWh-thermal figure is an R&D target, not the model's claimed installed cost.

Figure 3: Levelised comparison in 2026 US dollars under explicit assumptions. The base electricity input rounds the EIA's measured 2025 US industrial average of $86.20/MWh; gas starts at the measured $5.23/thousand cubic feet and is treated approximately as $5.23/MMBtu. Results are derived, not plant bids. Demand charges, connection works, taxes, heat recovery, backup fuel, and production disruption can reverse the ranking.
Figure 4 · Grid connection model

A gas pipe and a power line are not interchangeable

A continuous heat load becomes a large electrical load. Heat pumps reduce it where temperature lift permits; diversity and headroom increase the connection required.

Resistance connection requirement56.7MWe

Direct resistance draws 51.0 MWe at the heater. A heat pump at the selected COP draws 14.3 MWe before auxiliaries. Storage can change when power is drawn, but not the energy that must cross the meter.

Figure 4: Derived steady-state power balance. Resistance assumes 98% device efficiency. The connection value divides resistance input by the selected usable share to represent operational headroom; it is not a utility interconnection study.

Who is building what

Thermal batteries, high-temperature industrial heat pumps, and electric boilers electrify process heat across temperatures from 100°C to over 1500°C. Search the record, or filter by technology.

8 programmes
Rondo EnergyRondo Heat BatteryElectric resistance heating elements warming refractory clay bricks up to 1150°C, delivering continuous high-pressure steam or superheated air
Reported evidence
Commercial 2 MWh unit operating continuously at Calgren Renewable Fuels in California; expanding industrial installations in Europe and the US.
Announced next step
Gigawatt-hour annual production capacity supplying zero-carbon steam to chemical, food, and cement manufacturing.
Unresolved risk
High volumetric footprint inside existing space-constrained factories and electricity price spreads against cheap pipeline natural gas.
Antora EnergySolid Carbon Thermal BatterySolid carbon/graphite blocks heated by resistance to 1500°C+, discharging power via high-efficiency thermophotovoltaics (TPV) and high-temp heat
Reported evidence
Inaugurated commercial manufacturing facility in San Jose, California; commissioned first commercial thermal battery at a Fresno industrial site.
Announced next step
Simultaneous delivery of industrial steam and on-demand firm electricity from a unified modular thermal battery.
Unresolved risk
TPV cell production yield and preventing carbon oxidation through vacuum and inert gas hermetic sealing over decades.
Electrified Thermal SolutionsJoule HiveElectrically conductive refractory ceramic bricks that act as their own heating element via direct Joule heating up to 1800°C
Reported evidence
Demonstrated core brick conductivity and thermal stability; selected for a $35M DOE industrial demonstration grant.
Announced next step
Direct replacement of fossil burners in cement kilns, glass melting tanks, and metal forging furnaces.
Unresolved risk
Ceramic element degradation under rapid cyclic thermal shock and long-term electrical contact stability at ultra-high temperatures.
AtmosZeroBoiler 2.0Modular, drop-in industrial air-source heat pump boiler producing saturated steam up to 165°C with high seasonal coefficient of performance (COP)
Reported evidence
Piloting commercial units with New Belgium Brewing; closed Series A funding to establish domestic manufacturing.
Announced next step
Direct drop-in replacement for standard fossil-fuel package boilers across breweries, pharmaceuticals, and food plants.
Unresolved risk
Refrigerant performance degradation and lift limits when ambient air temperatures drop during severe winter extremes.
Turboden / HeatexLarge-Scale Steam Heat PumpsCentrifugal compressor industrial heat pumps capturing low-grade industrial waste heat (40–80°C) to generate process steam up to 200°C
Reported evidence
Multi-megawatt installations operating in chemical plants and paper mills in Scandinavia and Central Europe.
Announced next step
Replacing fossil boilers with systems achieving COPs between 2.5 and 4.0 using industrial effluent heat.
Unresolved risk
Compressor design constraints for high-temperature working fluids and site-specific engineering requirements.
Babcock & Wilcox / Vapor PowerElectric Electrode BoilersHigh-voltage electrode boilers passing current directly through water to produce high-pressure steam at nearly 100% electrical efficiency
Reported evidence
Operating in district heating and industrial facilities globally with ramp rates from standby to full steam output in under one minute.
Announced next step
Rapid grid-balancing steam services utilizing curtailed renewable power during negative pricing events.
Unresolved risk
Operating cost: pure resistance boilers have a COP of 1.0, requiring electricity to be cheaper per MWh than natural gas to compete.
Boston MetalMolten Oxide ElectrolysisDirect electrochemical reduction of metal oxides at 1600°C using an inert iridium-based anode, bypassing fossil combustion completely
Reported evidence
Inaugurated commercial plant in Minas Gerais, Brazil, for high-value critical metals extraction; testing steel-scale electrolytic cells in Massachusetts.
Announced next step
Licensing commercial MOE cells for multi-million-tonne green ironmaking by 2026–2030.
Unresolved risk
Refractory corrosion and anode lifetime in corrosive molten oxide slag baths at extreme operating temperatures.
US DOE Industrial Heat ShotFederal Decarbonization InitiativeFederal R&D funding targeting an 85% reduction in greenhouse gas emissions from industrial heat by 2035 through electrification and clean fuels
Reported evidence
Disbursed billions under the Bipartisan Infrastructure Law and IRA for advanced industrial heat demonstration pilots.
Announced next step
Cost parity between electrified heat solutions and fossil fuels across low-, medium-, and high-temperature industrial processes.
Unresolved risk
Policy reversals, industrial inertia, and delays in connecting heavy industrial power loads to electrical substations.

Thermal battery efficiencies describe electricity-to-heat recovery under continuous operation. Standby thermal losses (1–3% per day) become significant during prolonged factory shutdowns.

The new bottleneck

Abundant electricity moves scarcity into the factory

Once clean power is cheap, the binding work is delivering its heat at the required flux, cleanliness, continuity, and product quality while a live plant keeps shipping.

First

Recover and upgrade

Map waste heat, lower required temperatures, and use heat pumps where source and sink align.

Then

Electrify the simple loads

Hot water, low-pressure steam, batch ovens, and direct electromagnetic processes establish operating experience.

Finally

Redesign the process

High-temperature continuous assets need furnace, material-flow, grid, storage, and product engineering together.

There is more than one finish line

  1. Temperature-capableThe heater reaches the nominal process temperature.
  2. Flux-capableIt transfers enough heat into the real product to preserve throughput.
  3. Quality-capableAtmosphere, uniformity, residence time, and control preserve yield and specification.
  4. Plant-capableConnection, reliability, maintenance, footprint, and downtime fit the site.
  5. Cost-competitiveDelivered heat wins under the actual tariff, utilisation, financing, and production boundary.

The fuel-price ratio and temperature decide the first conversions

IEA’s 2025 industrial-heat review says commercial large heat pumps can serve up to roughly 150°C and electric boilers can make steam up to about 350°C. Those are equipment capabilities, not proof of integration into every food, paper or chemical site. High-temperature steel and cement process steps need different equipment and have separate chemical emissions. In its 2026 monitor, IEA estimates existing industrial heat pumps could technically serve about 20% of global industrial heat, while Japan’s roughly 650 MW of installed industrial heat-pump capacity in 2024 covered only about 1% of light-industrial demand.

For a resistance boiler, each kWh of heat needs about one kWh of electricity. At a 90% gas-boiler efficiency, the electricity price must be near or below 1.11 times the gas price per kWh of fuel to match the energy bill, before capital and carbon cost. A heat pump at COP 3.5 tolerates an electricity-to-gas price ratio of about 3.89. IEA’s 2024 country comparison reports standalone electric-boiler running costs about four times gas in the United States and twice gas in China on its example assumptions. That is why a heat pump or waste-heat recovery may clear first even when an electric boiler is technically simple.

China’s electricity share of final energy reached 28% in 2024, versus 27% in 2023, though this economy-wide measure is not an industrial-heat share. The meaningful plant-level diffusion metric is fossil heat displaced by temperature band, with measured electricity use, output quality, connection lead time and delivered cost.

The global heat curve is still early

IEA estimates electricity supplied about 4% of global industrial process heat in 2024 and models 12% in 2030, with China representing more than half of the projected growth in renewable electric process heat. The 2030 figure is a scenario, not installed thermal-battery capacity. Rondo reports a 100 MWh heat battery entering operation in 2025; that is company-reported site capacity, with no public audited installed $/kWh-th for a cross-vendor cost curve.

IEA describes industrial heat pumps reaching roughly 200°C in emerging applications, beyond its narrower commercial mainstream range around 150°C. The industrial price test must include tariff demand charges and utilization as well as energy $/MWh; a site with short process runs can lose an apparent COP advantage to capital recovery and connection costs. Cement clinker and virgin steel reduction still require separate chemistry and high-temperature solutions, covered in the cement and steel reports.

Sources, method, and boundaries

Temperature points in the explorer are representative screening values. The cost model keeps measured energy-price anchors separate from editorial equipment assumptions. Technology temperatures describe reported commercial or demonstration boundaries, not universal material limits. Company and funded-project targets are explicitly labelled.

Measured
An observed energy price, operating statistic, or completed test reported by a primary source.
Derived
A calculation whose inputs and boundary are shown in the model.
Projected
A modeled future outcome, such as a technical-potential or scenario figure.
Target
A future cost, temperature, efficiency, or emissions objective.
Editorial inference
A maturity or integration judgement synthesised from the evidence.