Western Canada's gas benchmark has traded at a punishing discount to Henry Hub for a decade, and the accepted cure was always the same: build export terminals, pull the surplus to tidewater, and the discount closes. LNG Canada was the first dose. It has been running for over a year.

The discount narrowed by roughly eighty cents. Then it stopped and began widening again.

The easy conclusion is that the cure doesn't work. It's the wrong conclusion. Canada has close to 6 Bcf/d of new export demand in construction or permitting, plus a gigawatt-scale power build nobody was modelling two years ago. The demand is real and it is very large. What determines the basis is not whether it arrives, but whether it arrives faster than a basin that has added supply every year for five years — and whether the pipes exist to move it when it does.

Three clocks, running at different speeds. The price sits wherever they happen to be relative to each other.

The short version

  • LNG Canada narrowed AECO basis from about −$2.59 to −$1.77/MMBtu — then it re-widened to −$1.912 for the balance of summer 2026.
  • Not a failure of demand. Phase 1 consumes 1.8 Bcf/d; the basin added roughly 1.5 Bcf/d of new supply across 2025–26.
  • The full stack is far larger: Woodfibre, Cedar, Ksi Lisims and Phase 2 take Canadian LNG toward ~6 Bcf/d by the early 2030s — about a third of current production.
  • Run the years and it oscillates. Supply leads through 2027–28, Ksi Lisims nearly closes the gap in 2029, drilling reopens it, Phase 2 closes it again around 2031.
  • The overlooked clock is in-basin demand. Alberta's data-centre queue is 41 projects, ~19,565 MW — more than 1.5x the province's entire peak load. Converted to gas, half of it is 1.44 Bcf/d, or 80% of Phase 2.
  • And it bids for the same molecules without paying a pipeline toll. Phase 2 is being sanctioned into a basin where 19.6 GW of load is queuing for its feedgas.
  • The market has already voted: with all four terminals known, the forward curve prices AECO at $2.316/MMBtu through 2027–28.
Basis figures are NGI Forward Look assessments in US$/MMBtu · project capacities per Canada Energy Regulator · current to September 2026

Eighty Cents, Then a Stall

The convergence was real and deserves credit before it is qualified. NGI's Forward Look put AECO basis to Henry Hub at −$2.59/MMBtu for summer 2025 and −$2.336 for winter 2025/26, reaching −$1.772 by May 2026. Then it turned: −$1.856 in June, −$1.912 for the balance of summer. Spot NOVA/AECO C averaged C$1.365/GJ on 1 April, the lowest since early November and down C$1.265 year on year.Verified

Fig 1The Discount Narrowed, Then Reversed
AECO forward basis to Henry Hub, US$/MMBtu — less negative is better for Canadian producers
$0 −$1.50 −$3.00 −2.59 −2.336 −1.772 −1.856 −1.912 Summer 2025 Winter 25/26 May 2026 June 2026 Bal. summer ~80¢ delivered then given back
Source: NGI Forward Look basis assessments reported by Natural Gas Intelligence, April and June 2026. Separate forward assessments for the periods shown, not a continuous traded series. Verified

The reason is arithmetic. Phase 1 consumes about 1.8 Bcf/d at full rate. The Canada Energy Regulator records 2025 Western Canadian production averaging 19.0 Bcf/d, up 3.8 per cent, with a 20.0 Bcf/d monthly record in November. RBN Energy puts the 2025 gain at 0.7 Bcf/d — a fifth consecutive annual rise — and projects 0.8 Bcf/d for 2026. Two years of drilling delivered roughly 1.5 Bcf/d against a terminal that removes 1.8.Verified

TD Securities forecast exactly this in 2025, writing that Phase 1 demand would be largely filled in real time by supply additions already in motion. The analysis existed; the conclusion didn't travel.


What Is Actually Coming

Reading Phase 1 in isolation is the error most commentary makes, including the bearish version. The pipeline of demand is far larger than one terminal.Verified

On the CER's project figures: Woodfibre adds 0.28 Bcf/d from 2027, Cedar 0.39 from late 2028, and Ksi Lisims 1.58 from 2029 — now federally fast-tracked, with offtake signed and its 2 Bcf/d feeder line under construction. Phase 2 would add another 1.8 in the early 2030s. East Daley puts the total at up to 6.3 Bcf/d if everything envisioned in British Columbia proceeds: roughly a third of 2025 production.

And that is only the export half. TC Energy has raised its 2025–2035 North American demand growth forecast twice this year, from 40 to 46 to 51 Bcf/d, with about 70 per cent concentrated in the US heartland, Alberta and Mexico — and reported record data-centre participation in its intra-Alberta capacity offering for 2030–32. Section IV puts a number on that half.

Fig 2The Demand Stack, by Commissioning Year
Cumulative Canadian LNG feedgas requirement, Bcf/d
6.0 3.0 0 1.80 2025 LNGC Ph1 2.08 2027 +Woodfibre 2.47 2028 +Cedar 4.05 2029 +Ksi Lisims 5.85 early 2030s +LNGC Ph2 Plus in-basin power load: ~0.2–0.45 Bcf/d if Alberta's allocated and proposed data-centre capacity is built.
Sources: Canada Energy Regulator project figures (Woodfibre 0.28 Bcf/d, Cedar 0.39, Ksi Lisims 1.58); LNG Canada Phase 1 feedgas of 1.8 Bcf/d and Phase 2 of approximately 1.8 Bcf/d as publicly disclosed; East Daley (up to 6.3 Bcf/d if all BC projects proceed). In-basin figure converts AESO-allocated and proposed data-centre load at combined-cycle heat rates. Estimate

Why the Basis Will Oscillate, Not Converge

Set the demand stack against basin growth and the pattern is neither bullish nor bearish. It is periodic.Interpretation

Demand arrives in lumps, on commissioning dates. Supply arrives continuously, at roughly 0.8 Bcf/d a year. Hold RBN's run rate flat and the gap moves like this: supply leads by about half a Bcf/d in 2027 and nearly a full Bcf/d in 2028; Ksi Lisims lands in 2029 and closes it to almost nothing; two more years of drilling reopen it; Phase 2 closes it again around 2031.

Fig 3Lumpy Demand Against Continuous Supply
Cumulative new demand less cumulative basin growth from 2026, Bcf/d
+0.5 0 −1.0 demand ahead ▲ supply ahead ▼ −0.52 2027 −0.93 2028 −0.15 2029 Ksi Lisims −0.95 2030 +0.05 2031 Phase 2 Each terminal buys a step narrowing. Drilling spends it back before the next one lands.
Insight Stream arithmetic: cumulative CER project feedgas requirements from 2027 less cumulative basin growth held at RBN's projected 0.8 Bcf/d annual rate. Illustrative on published rates; not a production or price forecast, and assumes every project commissions on its stated date. Estimate

That is the honest shape of it. Not a discount that closes, and not one that never closes — a discount that tightens sharply at each commissioning and loosens between them, unless the basin's growth rate falls or egress capacity changes the arithmetic.Interpretation


The Bid Nobody Modelled

Here is the variable left out of every version of this debate, and it is the largest single number in this article.Verified

The AESO's own connection project list for July 2026 contains 41 data-load projects requesting roughly 19,565 MW of demand transmission service. A year earlier it was 29 projects and just over 16,000 MW. For scale, Alberta's entire peak demand is about 12,000 MW, and the City of Edmonton draws around 1,400 MW. The queue is more than one and a half times the province's peak load.

Convert that to gas. At a modern combined-cycle heat rate of about 6.8 MMBtu/MWh and a 90 per cent capacity factor, a gigawatt of continuous data-centre load burns roughly 0.15 Bcf/d. The full queue would therefore require about 2.9 Bcf/d. Nobody expects all of it to be built — but a quarter of it is 0.72 Bcf/d, and half of it is 1.44 Bcf/d.Estimate

Fig 4Alberta's Data-Centre Queue, Priced in Gas
Implied feedgas demand at conversion rates, against the LNG projects it competes with, Bcf/d
0 1.0 2.0 3.0 Bcf/d Full queue, 19,565 MW 2.87 50% converts 1.44 25% converts 0.72 FOR COMPARISON — LNG PROJECTS LNG Canada Phase 2 1.80 Ksi Lisims 1.58 Cedar + Woodfibre 0.67 Half the queue is 80% of Phase 2. None of it needs a pipeline to the coast.
Queue figure from the AESO Connection Project List, July 2026 (41 data-load projects, approximately 19,565 MW of requested demand transmission service). Gas conversion by Insight Stream at 6.8 MMBtu/MWh and a 90% capacity factor; conversion rates are illustrative scenarios, not forecasts. LNG figures per Canada Energy Regulator and publicly disclosed project capacities. Estimate

The grid route is closed, which forces the gas outcome

This demand cannot reach the wires. The AESO capped large-load connections at 1,200 MW through 2028 and allocated all of it by late 2025 — 970 MW to the project near Edmonton that became the Meta facility, 230 MW to Keephills. Everyone else must bring their own generation. Which is why gigawatt-class gas plants are now at hearing: the Indus project went before the Alberta Utilities Commission in August 2026 seeking 1,494 MW to supply co-located data halls, with a Heartland application of similar scale. These are not peakers. They are dedicated, high-utilisation industrial burns sited on top of the resource.Verified

And it bids for the same molecule without paying the toll

A data centre in Sturgeon County and an LNG train at Kitimat want the same Alberta gas. The data centre burns it at the plant gate. The LNG train moves it roughly 670 kilometres across a system the CER describes as already at or above capacity, and pays to do so. For the same delivered economics, the buyer with no transport leg can bid more at the hub. Coastal GasLink's tolls are commercially set and not public, so the wedge cannot be sized from outside — but its direction is not in doubt.Interpretation

Fig 5Two Bidders, One Molecule
Why in-basin demand supports AECO more efficiently than export demand
The marginal Alberta molecule priced at AECO IN-BASIN BIDDER Burns at the plant gate No pipeline, no toll Can bid the full delivered value EXPORT BIDDER ~670 km to tidewater Pays a toll on a full system Bids delivered value minus transport Both lift AECO. Only one of them keeps the whole difference.
Framing by Insight Stream. Coastal GasLink route length as publicly reported; tolls are commercially set and not public, so the wedge is directional rather than quantified. Pipeline utilisation per Canada Energy Regulator market snapshot on WCSB pipeline performance, June 2026. Interpretation

Who this helps, and who it complicates

Alberta and northeast BC producers gain a bidder that pays no toll. In-basin load is the only demand in the stack that lifts AECO without first consuming scarce export capacity, and it is the demand most likely to surprise upward — the queue grew by 3.5 GW in a single year.

LNG offtakers acquire a competitor they did not model. Phase 2 is moving toward a final investment decision into a basin where 19.6 GW of load is queuing for the same molecules. Feedgas assumptions written when Alberta's only marginal buyers were pipelines to the US Midwest are being tested by a bidder that sits upstream of the constraint.

And the build clocks are not comparable. A gas plant runs an AUC proceeding and a roughly three-year construction; the Indus application reached hearing in August 2026. An LNG terminal runs a federal impact assessment and a build closer to six years — Ksi Lisims entered assessment early this decade and received its major regulatory approval only in September 2025. In-basin demand can arrive inside the window in which Phase 2 is still being built.Interpretation


What the Curve Says, and Who It Suits

None of this is unknown to the market. Ksi Lisims has regulatory approval and signed offtake; Woodfibre and Cedar are under construction; Phase 2 is designated a national-interest project. And with all of it visible, NGI's forward curve prices NOVA/AECO C at an average of $2.316/MMBtu through 2027 and 2028, and below $3 until 2030.Verified

The market is not betting the demand fails to arrive. It is pricing the race as close to a draw — which, on the year-by-year arithmetic above, is roughly right.

That has different consequences depending on where a participant sits. Shell operates 40 per cent of LNG Canada and has acquired ARC Resources for $16.5 billion, adding around 370,000 boe/d of largely Montney production. An integrated participant owning molecules and liquefaction is close to indifferent to the basis: a wider discount simply moves value from one part of the group to another. A non-integrated Montney producer is fully exposed to it. An offtaker without upstream benefits while it lasts.Interpretation

Fig 6Who Is Exposed to the Basis, and Who Is Not
Position by structure if the discount persists
INTEGRATED Owns molecules and liquefaction Indifferent value moves within the group UPSTREAM ONLY Montney producer, no export position Fully exposed sells at the discounted hub OFFTAKE ONLY Buys feedgas, sells cargoes Benefits while the discount lasts The discount is not a national problem. It is a transfer, and structure decides the direction.
Framing by Insight Stream. Shell holds a 40% operated interest in LNG Canada and acquired ARC Resources for $16.5 billion; joint venture participants are Shell 40%, Petronas 25%, PetroChina 15%, Mitsubishi 15%, KOGAS 5%. Interpretation

What Follows

1 · Model the commissioning calendar, not the capacity total. Six Bcf/d arriving smoothly closes the basis. Six Bcf/d arriving in four lumps across six years produces a sawtooth. The dates matter more than the sum, and the tightest windows are the twelve months after each start-up.

2 · Put the data-centre queue in the feedgas model. At a 25 per cent conversion rate it is 0.72 Bcf/d; at 50 per cent it is 1.44, or four-fifths of Phase 2. It bypasses the pipeline constraint, builds in roughly half the time, and bids without a transport deduction. Any Phase 2 or Ksi Lisims feedgas assumption that treats Alberta as having only pipeline buyers at the margin is already out of date.

3 · Watch egress, then drilling, then terminals. The CER already has the pipelines at or above capacity. Whether the discount narrows durably turns on Coastal GasLink Phase 2, the Prince Rupert line's completion, NGTL expansions and mainline tolls — and on whether producers show restraint. Neither is decided at Kitimat.

4 · Read structure before sentiment. Before accepting a company's view on where the basis goes, check which side of it that company sits on.

Our call, stated so it can be checked We expect AECO basis to remain wider than −$1.00/MMBtu against Henry Hub through 2028, to narrow materially in the twelve months following Ksi Lisims start-up, and to re-widen thereafter absent egress growth or a slowdown in basin drilling. We expect the same pattern around Phase 2. In short: step narrowings at commissioning, re-widening between them, and no durable convergence until takeaway capacity grows faster than production. We will revisit this against the same public sources in twelve months. Estimate

What would change our mind

  • Producer restraint. If WCSB growth falls below roughly 0.4 Bcf/d a year while the demand stack lands, the basis closes durably and this argument fails.
  • Egress outpacing drilling. If Coastal GasLink Phase 2, the Prince Rupert line and NGTL expansions deliver faster than production grows, the constraint we identify stops binding.
  • Queue conversion above 50%. If more than half of the 19,565 MW reaches construction, in-basin demand alone approaches the scale of Phase 2 and could close the basis with no new egress at all. Equally, if AUC rejections or capital discipline hold conversion below 10%, this clock barely ticks.
  • Project slippage. Equally, if Ksi Lisims or Phase 2 slip by two years, the demand lumps arrive later and the discount stays wider for longer than we expect.

Three Clocks

The bull case says Canada is building six Bcf/d of new demand into a basin producing nineteen, and the discount cannot survive it. The bear case points at a year of LNG Canada operations and an AECO price that ended lower than it started. Both are reading real evidence.

Both are looking at one clock. Demand arrives in known increments on known dates; supply arrives continuously at three-quarters of a Bcf/d a year; and the pipes between them are, on the regulator's own account, already full. Behind all of it sits a queue of 41 data-centre projects requesting 19,565 MW — a bid for the same molecules that pays no toll to reach them, and that almost nobody has put in the feedgas model.

The price of Western Canadian gas is the running difference between those clocks. It will move in steps, not a trend — tighter for a year after each start-up, looser while the basin catches up. Anyone underwriting a producer, a toll or an offtake on smooth convergence is underwriting the wrong shape.

IS

Insight Stream — LNG/LPG & Waterborne Trade Analysis

LNG contract economics, LPG benchmarks and Pacific trade flows, VLGC freight, China PDH and NGL export infrastructure. Independent market intelligence, built only on sources you can check.

Read More →