Natural gas supply and fuel strategy

Lesson 7 covered the technologies for on-site generation. Most of those plants burn natural gas, and a data centre that generates its own power, whether as a bridge until the grid connection arrives or permanently, becomes in effect a gas-fired power producer. Its reliability then depends on gas supply, its costs on gas prices, and its emissions on how much gas it burns. This lesson explains how natural gas markets and pipelines work, how much gas a data centre plant needs, the choices between firm and interruptible supply and between utility and direct pipeline service, gas price risk and hedging, winter reliability and backup fuels, emissions and carbon pricing, and options such as renewable natural gas, hydrogen blending and carbon capture.

Natural gas markets

In Canada, most gas is produced in Alberta and British Columbia and traded at the AECO hub (NOVA Inventory Transfer) in Alberta, priced in dollars per gigajoule; eastern Canadian markets also reference the Dawn hub in southwestern Ontario, closely linked to US markets. AECO prices have often been well below US prices because of limited pipeline export capacity, though new LNG export capacity on the BC coast (LNG Canada began operating in 2025) is strengthening the link to international prices. Gas distribution to customers is regulated by provincial utility commissions; interprovincial and international pipelines are regulated by the Canada Energy Regulator.

Gas prices are seasonal, higher in winter when heating demand peaks, and volatile: annual averages have ranged from under $2 to over $8/MMBtu at Henry Hub in the past decade, and short-term regional prices can rise far higher during cold snaps. Units: 1 MMBtu equals 1.055 GJ, and 1,000 cubic feet (Mcf) of typical pipeline gas contains about 1.037 MMBtu.

How much gas a data centre plant needs

A generator's gas consumption is its output multiplied by its heat rate, the fuel energy needed per MWh of electricity. Heat rates range from about 6.3 to 7 MMBtu/MWh (higher heating value) for combined-cycle plants and solid oxide fuel cells, about 8 to 9 for reciprocating engines and aeroderivative turbines, and 10 or more for older simple-cycle turbines.

# Gas consumption, fuel cost and CO2 for 100 MW of on-site generation at a 90% capacity factor
MWh = 100 * 8760 * 0.90
options = {"combined-cycle turbine": 6.6, "reciprocating engines": 8.3, "aeroderivative simple cycle": 9.3, "fuel cells (solid oxide)": 6.5}
gas_price = 3.50                     # $ per MMBtu delivered (commodity plus transport)
for name, hr in options.items():     # heat rate in MMBtu per MWh (higher heating value)
    mmbtu = MWh * hr
    gj = mmbtu * 1.055
    co2 = mmbtu * 0.0531              # tonnes of CO2 per MMBtu of natural gas
    print(f"{name:28s}: {mmbtu / 1e6:5.2f} million MMBtu ({gj / 1e6:5.2f} PJ, {mmbtu / 1.037e6:4.1f} Bcf); fuel ${hr * gas_price:5.2f}/MWh; {co2 / MWh:.2f} t CO2/MWh")
# Pipeline capacity needed for the reciprocating engine plant at full output
peak_mmbtu_day = 100 * 24 * 8.3
print(f"peak daily gas for 100 MW of engines: {peak_mmbtu_day:,.0f} MMBtu/day (about {peak_mmbtu_day / 1.037 / 1000:.1f} MMcf/day)")

combined-cycle turbine      :  5.20 million MMBtu ( 5.49 PJ,  5.0 Bcf); fuel $23.10/MWh; 0.35 t CO2/MWh
reciprocating engines       :  6.54 million MMBtu ( 6.90 PJ,  6.3 Bcf); fuel $29.05/MWh; 0.44 t CO2/MWh
aeroderivative simple cycle :  7.33 million MMBtu ( 7.74 PJ,  7.1 Bcf); fuel $32.55/MWh; 0.49 t CO2/MWh
fuel cells (solid oxide)    :  5.12 million MMBtu ( 5.41 PJ,  4.9 Bcf); fuel $22.75/MWh; 0.35 t CO2/MWh
peak daily gas for 100 MW of engines: 19,920 MMBtu/day (about 19.2 MMcf/day)

A 100 MW plant running most of the year burns 5 to 7 million MMBtu of gas, comparable to the gas use of 70,000 to 100,000 homes, and needs about 20,000 MMBtu (about 19 million cubic feet) per day at full output. A 1 GW campus powered on site would need ten times that, roughly 0.2 billion cubic feet per day, a significant fraction of a large pipeline's capacity. At $3.50/MMBtu, fuel adds $23 to $33 per MWh of electricity depending on the technology, before capital, operation and maintenance costs.

Pipelines and service arrangements

Gas travels from producing regions through high-pressure transmission pipelines (often operating at 5 to 10 MPa, roughly 700 to 1,400 psi), then to customers through lateral pipelines and local distribution systems at lower pressure. A large data centre plant can be served in two main ways:

• Local distribution company (LDC) service: the utility delivers gas through its distribution system under a tariff, possibly with a new high-pressure lateral built for the customer. This is simpler, but the customer depends on the LDC's capacity and curtailment rules.
• Direct connection (bypass) to a transmission pipeline: a dedicated lateral and metering station connect the plant directly to an interstate or interprovincial pipeline, with transport contracted from the pipeline and gas bought from producers or marketers. This can lower costs and give firmer access for large loads, subject to regulatory rules on bypassing the LDC.

Transportation service is either firm or interruptible:

• Firm transportation reserves a fixed daily capacity on the pipeline, paid as a reservation charge whether used or not. It is curtailed only in true emergencies.
• Interruptible transportation is cheaper but can be cut whenever firm customers need the capacity, which is exactly during cold weather peaks, when gas-fired generation may be most needed.

Gas turbines need gas at a high and stable pressure (often 2 to 4 MPa for large industrial and aeroderivative units), which may require on-site compression if the delivery pressure is low; reciprocating engines and fuel cells generally accept lower pressures. New laterals and pipeline expansions need regulatory approval (DCE-410, Lesson 9) and can take two to four years, so gas supply is often on the critical path for on-site generation.

Gas price risk

For a data centre generating its own power, gas price risk replaces electricity price risk. The same hedging tools apply: fixed-price supply contracts, financial swaps at the relevant hub, options and a mix of fixed and index purchases (Lesson 2). Basis, the difference between the hub price and the local delivery price, must be hedged too.

import numpy as np

rng = np.random.default_rng(8)
years = 5000
heat_rate, MWh = 6.6, 100 * 8760 * 0.9
# Annual average hub price: lognormal around $3.50/MMBtu; some years have a winter crisis adding a spike
base = 3.5 * np.exp(rng.normal(0, 0.35, years))
crisis = rng.random(years) < 0.08
spike_days = rng.integers(3, 15, years)
spike_price = rng.uniform(20, 120, years)
annual_cost = MWh * heat_rate * base + crisis * spike_days * 24 * 100 * heat_rate * (spike_price - base)
per_mwh = annual_cost / MWh
print(f"fuel cost per MWh: median ${np.median(per_mwh):.1f}, 90th percentile ${np.percentile(per_mwh, 90):.1f}, 99th percentile ${np.percentile(per_mwh, 99):.1f}")
fixed = 4.0                                            # a fixed-price gas contract at $4.00/MMBtu for the whole volume
print(f"with a fixed $4.00/MMBtu supply contract: ${fixed * heat_rate:.1f}/MWh every year")
print(f"probability the index cost exceeds the fixed contract: {np.mean(per_mwh > fixed * heat_rate):.0%}")

fuel cost per MWh: median $23.7, 90th percentile $38.0, 99th percentile $56.6
with a fixed $4.00/MMBtu supply contract: $26.4/MWh every year
probability the index cost exceeds the fixed contract: 38%

Most years the index cost is close to or below the fixed price, but a few years with a winter crisis push the cost far higher. A fixed contract costs a little more on average and removes that tail. Physical supply security matters as much as price: a contract that fixes the price is worthless if the gas cannot be delivered because transport is interruptible or a supplier fails to perform.

Winter reliability and backup fuel

Extreme cold stresses gas supply at the same time as it raises electricity demand. Wellheads and gathering lines freeze, pipeline pressure falls as heating demand peaks, and interruptible customers are cut off. During Winter Storm Uri in Texas (February 2021) and Winter Storm Elliott in the eastern US (December 2022), gas supply failures were a leading cause of generator outages. A data centre that depends on on-site gas generation for its primary power must design for these events:

• Firm transportation on pipelines with strong winter performance, and contracts with suppliers who have firm supply.
• Dual-fuel capability: many turbines and some engines can run on liquid fuel (ultra-low-sulphur diesel) as well as gas, with on-site storage for several days (DCE-340).
• On-site or nearby gas storage, such as compressed or liquefied natural gas, for short interruptions.
• Retaining a grid connection and battery storage as additional sources.
• Winterization of the plant and fuel systems, following lessons from past events and regulatory requirements introduced since.

Emissions and carbon pricing

Natural gas combustion emits about 53 kg of CO2 per MMBtu (about 50 kg/GJ), so on-site generation emits roughly 0.35 to 0.5 tonnes of CO2 per MWh depending on efficiency, compared with near zero for nuclear and renewables and about 0.9 to 1.0 t/MWh for coal. Methane leakage upstream in production and transport adds to the climate impact and is increasingly regulated.

In Canada, large industrial emitters, including gas-fired power plants above size thresholds, are subject to carbon pricing under the federal output-based pricing system or equivalent provincial systems (such as Alberta's TIER), which charge for emissions above an output-based benchmark at the carbon price (scheduled to rise to $170 per tonne by 2030 under the federal benchmark). The federal Clean Electricity Regulations also limit emissions from fossil-fuel generating units connected to the grid from 2035, with provisions for backup and flexibility. These rules significantly affect the long-term economics of gas-fired data centre power.

Lower-carbon gas options

• Renewable natural gas (RNG): biomethane from landfills, farms and wastewater, injected into pipelines and bought through certificates; supply is limited and expensive compared with fossil gas.
• Hydrogen blending: many modern turbines can burn blends with 10% to 30% hydrogen by volume (less by energy), and some are designed for higher fractions; but clean hydrogen is scarce and costly, and blending reduces CO2 only modestly at low fractions.
• Carbon capture and storage: capturing CO2 from turbine exhaust (post-combustion) can remove 90% or more, at significant cost and energy penalty, and requires access to geological storage; several projects pairing data centres with gas plants and carbon capture have been announced.
• Transition planning: designing plants and contracts so that gas generation can later move to backup duty when clean firm power or grid connections become available, reducing stranded asset risk.