The 2026 ISP’s Hidden Assumptions: Why AEMO’s Least-Cost Claim Deserves Greater Scrutiny

High-voltage electricity transmission tower carrying multiple power lines against a blue sky with scattered clouds.

A closer examination of the modelling choices, excluded costs and predetermined assumptions behind AEMO’s preferred energy pathway

The 2026 ISP is not a clean-sheet, whole-of-system assessment of the cheapest way to supply electricity. It is a policy-constrained optimisation of the remaining decisions, after governments, AEMO and project proponents have already placed a large body of infrastructure outside the decision boundary.

That does not mean the model is fabricated. It means AEMO can make statements that are technically correct while presenting a picture that is materially incomplete.

The central problem is this:

AEMO describes the ISP as a whole-of-system least-cost roadmap, but its headline cost-benefit results are marginal calculations made after tens of billions of dollars of projects, policies and consumer investments have been assumed rather than optimised.

I would characterise the report as internally sophisticated, but strategically framed. The weaknesses lie in the counterfactual, project classifications, imposed policy constraints, incomplete system boundary, financing assumptions and sensitivities that mostly confirm rather than seriously threaten the preferred development path.

I cannot establish deliberate deception from the ISP alone. I can establish extensive selective framing, path dependence and insufficient transparency.

1. The headline “least-cost” claim answers a much narrower question than readers are led to believe

AEMO says the ISP identifies the least-cost system. In reality, it identifies the lowest-cost candidate among a shortlist that:

must meet all incorporated government targets;
contains all committed and anticipated projects;
takes consumer investment and energy efficiency as external assumptions;
assumes particular gas developments and fuel availability;
and optimises only the remaining grid-scale generation, storage and network choices.

AEMO openly says all relevant government policies are incorporated into all three scenarios and that it does not assess the merits or feasibility of those policies.

The accurate description would therefore be:

The lowest-cost modelled pathway among the candidates considered, conditional on all incorporated government policies, existing decisions, project classifications, demand assumptions and technology assumptions.

That is very different from:

The least-cost future electricity system for consumers.

This is perhaps the ISP’s most fundamental act of framing. The conclusion is not necessarily wrong within its boundaries. The boundaries are doing much of the work.

2. The project-status ratchet prevents major investments from being reconsidered

AEMO explicitly says that existing, committed and anticipated investments are not costed or re-evaluated. Their capital costs are common to every development path and therefore disappear from the comparison.

This creates a powerful ratchet:

A project appears in an ISP.
Governments or proponents advance planning, approvals or finance.
It becomes committed, anticipated or policy-supported.
It is inserted into every subsequent scenario.
Its current cost and continuing justification are no longer tested.
Later modelling optimises the system around it.

AEMO confirms that $48 billion of generation, storage and transmission progressed into committed, anticipated or policy-supported categories and was consequently excluded from the 2026 ODP cost comparison.

This is not conventional sunk-cost treatment in the ordinary public sense. Much of the future expenditure has not yet been incurred. Consumers may still have to pay it. Routes, scopes and costs may still change substantially.

A project can therefore be treated as economically irreversible before it is physically built or fully paid for.

That is how HumeLink and Western Renewables Link disappear from the incremental transmission cost, even though they are integral to the future network presented by the ISP.

3. The $6 billion transmission headline is technically true but profoundly misleading

AEMO’s $6 billion figure is not the cost of the approximately 6,000 kilometres of transmission depicted in the ISP.

The 6,000 kilometres includes:

3,500 kilometres of committed and anticipated projects;
1,660 kilometres of actionable projects;
and up to 876 kilometres of future projects.

The $6 billion is the annualised present-value cost to 2050 of only the actionable and future ISP projects. AEMO says their total upfront present-value cost is approximately $16 billion, but only $6 billion is recognised within the ISP horizon after costs are spread over assumed 50-year asset lives.

It excludes the committed and anticipated transmission projects embedded in the 6,000-kilometre narrative.

Consequently, the public is presented with:

6,000 kilometres of transmission and a $6 billion transmission cost.

But those figures do not describe the same portfolio.

A more transparent headline would separate:

total expenditure on all unfinished transmission projects;
committed and anticipated project expenditure;
actionable and future project expenditure;
annualised cost recognised to 2050;
and total nominal expenditure ultimately recovered from consumers.

Without that reconciliation, the $6 billion figure is not useful as a measure of the real transmission burden.

4. The $28 billion benefit is also a marginal benefit, not a validation of the full transmission plan

The claimed $28 billion net market benefit applies only to actionable and future transmission projects. The counterfactual already includes all committed and anticipated projects.

It therefore does not establish that HumeLink, WRL, Project EnergyConnect, Marinus Stage 1 or the other assumed projects remain beneficial at their latest costs.

It asks:

Given that those projects will exist, what is the value of adding the remaining transmission portfolio?

It does not ask:

Is the complete transmission portfolio, at current costs, still the best available system design?

The benefits of the assumed network remain embedded in both cases because its transfer capacity, topology and generation-access effects shape the system being optimised. Its direct costs are also removed from both cases. Mathematically, that can be internally consistent. Strategically, it removes the most expensive and contentious decisions from review.

The resulting $28 billion figure validates only the incremental portfolio around the assumed network, not the network transformation as a whole.

5. Government policy predetermines a substantial part of the “optimal” answer

Every scenario must meet incorporated national and state policies, including:

renewable energy targets;
offshore wind targets;
storage targets;
REZ and transmission policies;
and state-specific development frameworks.

AEMO includes Victoria’s renewable targets, 9 GW offshore wind target, storage targets and coordinated transmission development as constraints or inputs.

This means the model cannot determine whether, for example:

9 GW of Victorian offshore wind is economically preferable to more onshore generation;
a different geographic generation mix would reduce transmission;
a slower or different electrification path would be cheaper;
or stronger distributed generation and efficiency could replace part of the network build.

AEMO itself says offshore wind is approximately two-thirds more expensive than onshore wind to build and connect, yet the model includes 9 GW because that is government policy.

There is nothing inherently improper about modelling government policy. The misrepresentation occurs when the resulting plan is described as though it independently establishes the economic superiority of the policy-driven system.

The ISP proves, at most:

Given the policies, this is AEMO’s preferred way of delivering them.

It does not prove:

The policies and infrastructure they require constitute the cheapest energy transition.

6. Scenario probabilities are expert judgements presented with a degree of numerical authority they do not possess

AEMO assigns:

46% to Step Change;
27% to Slower Growth;
27% to Accelerated Transition.

Those weights were derived through 25 expert stakeholders assessing 18 uncertain parameters.

These are not actuarial probabilities. They are structured expert judgements about political, technological, economic and social developments extending to 2050.

Yet those percentages directly influence the scenario-weighted $28 billion benefit.

A different group of experts, different parameter dependencies or different views of policy durability could materially change the weighting. The apparent precision of 46%, rather than perhaps a broad range such as 30–60%, lends unwarranted confidence to a deeply subjective exercise.

A robust plan should report:

results under each scenario separately;
results without probability weighting;
break-even scenario weights;
and the extent to which the selected ODP changes under reasonable alternative weighting schemes.

The main report does not give the reader enough information to determine whether the ODP is genuinely robust or merely favoured by AEMO’s chosen weighting.

7. Jurisdictional frameworks create circularity and reduce independent testing

Projects in Victoria, New South Wales and Queensland can progress through state frameworks rather than the standard ISP and RIT-T process. VicGrid coordinates Victorian projects through the Victorian Transmission Investment Framework.

That produces a circular planning structure:

state policy establishes targets or favoured projects;
the ISP incorporates those policies;
jurisdictional planners provide project options, costs and timing;
AEMO models the resulting projects;
and the ISP is then cited as evidence that the projects are required.

The New England REZ project demonstrates the mechanism particularly clearly. AEMO reclassified it as anticipated following advice from the NSW Government and consequently did not retest its actionability in the 2026 ISP.

A project’s progress can therefore become the reason it no longer requires economic reconsideration.

That is governance-driven path dependence, not independent least-cost planning.

8. The counterfactual does not represent a fully optimised decentralised alternative

AEMO’s counterfactual allows generation, storage and distribution development but no new transmission beyond what is already committed or anticipated.

At first glance, that sounds reasonable. The problem is that major demand-side alternatives are not co-optimised.

AEMO treats:

rooftop solar;
household batteries;
EV charging;
VPP participation;
energy efficiency;
demand response;
and electrification behaviour

as forecast inputs rather than investment choices competing directly against transmission.

AEMO simultaneously calculates that CER coordination and energy efficiency reduce grid-scale costs by billions of dollars, including:

$5.2 billion from coordinated consumer batteries and EVs;
$9.5 billion from continuing energy-efficiency policies;
and potentially another $7.2 billion from stronger efficiency.

That is a major contradiction in planning architecture.

The ISP acknowledges demand-side alternatives are enormously valuable but does not allow the optimisation to choose between:

one more dollar of transmission, one more dollar of distributed storage, one more dollar of building efficiency, or one more dollar of flexible demand.

Because grid-scale investments are optimised while demand-side investments are imposed, the model cannot legitimately claim to have found the least-cost whole-system mix.

9. Consumer expenditure produces modelled benefits but sits outside the headline cost

By 2050, Step Change assumes:

87 GW of rooftop and small-scale solar;
35 GW/78 GWh of consumer batteries;
more than half those batteries participating in VPPs;
80% of vehicles being electric;
and more than 10% of private EVs participating in vehicle-to-grid programs.

Those assets reduce the need for utility-scale generation, storage and networks.

But AEMO says consumer investments are not optimised and their costs do not directly drive ODP selection.

That gives the ISP a serious system-boundary problem:

the benefits of consumer expenditure reduce the modelled grid cost;
the consumer capital expenditure enabling those benefits is outside the ODP cost;
and the resulting grid cost is then described as the cost of the least-cost path.

AEMO may argue households purchase solar, batteries, appliances and EVs partly for private benefits. That is true. It does not make the expenditure disappear from the economic cost of the overall transition.

The ISP should distinguish clearly between:

regulated grid expenditure;
utility-scale market investment;
government-supported investment;
and consumer-funded enabling investment.

It currently blends the benefits while separating the costs.

10. Demand assumptions create much of the infrastructure “need”

Underlying consumption is forecast to increase from approximately 205 TWh to 390 TWh by 2050. The Step Change assumptions include:

117 TWh from electrification;
61 TWh for road transport;
34 TWh for data centres;
and 35 TWh for hydrogen production.

These forecasts are not neutral background information. They determine the amount and location of generation, storage and transmission.

Data centres and hydrogen are especially uncertain, location-flexible or potentially price-responsive loads. The model can create transmission need by assuming large loads emerge in particular regions, and then calculate benefits from building transmission to serve them.

AEMO gives prominent treatment to a Higher Demand sensitivity. Adding another 39 TWh causes claimed transmission benefits to rise from nearly $30 billion to $61 billion.

That is unsurprising. More large remote or concentrated load makes network capacity more valuable.

The more revealing tests would include:

lower industrial demand;
delayed or failed hydrogen development;
data centres required to bring firm supply or locate near available energy;
substantially stronger efficiency;
local generation and storage around major industrial loads;
and dynamic industrial curtailment.

Some demand and efficiency sensitivities exist, but they are not developed into equally prominent alternative system architectures.

11. The delivery sensitivity is a recovery scenario, not a genuine failure scenario

AEMO acknowledges that both the base plan and its constrained-delivery sensitivity require construction rates above those achieved historically. The sensitivity merely assumes delivery is slower than the ODP before returning to its trajectory by approximately 2035 or 2036.

For transmission, it assumes:

committed projects delayed six months;
anticipated projects delayed 12 months;
actionable and future projects delayed about two years;
costs increase 30%;
and the build trajectory subsequently catches up.

This does not test the harder possibilities:

five-to-ten-year delays;
project cancellation;
continuing route disputes;
compounding cost escalation;
workforce bottlenecks that do not resolve;
multiple concurrent project failures;
or inability to regain the proposed build trajectory.

It is fundamentally a temporary-delay sensitivity, not a structural-delivery failure case.

There is also an important modelling interaction. AEMO delays and increases the cost of generation and storage as well as transmission. When local generation and storage alternatives are constrained, transmission naturally becomes relatively more valuable.

AEMO then concludes that delays prove transmission should proceed more urgently.

That is not a neutral stress test. It is structured in a way that tends to confirm the value of transmission because the competing infrastructure is also made scarce and expensive.

12. “Robust” is being used too loosely

In the Higher Demand sensitivity, the selected ODP is no longer the least-cost candidate. Another candidate is approximately $60 million cheaper, but AEMO calls the ODP robust because the difference is small.

A $60 million difference is certainly immaterial compared with total system costs.

But that also demonstrates how weakly differentiated the leading candidates may be. The network projects themselves carry uncertainties of billions of dollars:

VNI West is estimated at $7.6 billion with a range of minus 30% to plus 50%.
Gippsland Offshore Wind Transmission components carry ranges extending from minus 50% to plus 100%.
Numerous future projects carry ±50% estimates.

When uncertainty is measured in billions, rankings separated by tens or hundreds of millions should not be presented as a stable optimisation result.

The correct conclusion is that several development paths may be economically indistinguishable within the accuracy of the inputs.

That should lead to preference for:

staged investment;
modularity;
preserving options;
maximising existing assets;
and postponing irreversible commitments.

Instead, AEMO repeatedly concludes that actionable projects should proceed urgently.

13. Transmission financing assumptions materially favour regulated network investment

The 2026 ISP reduces the assumed WACC for transmission from 7% to 3%, while applying rates between 7% and 12% to generation and storage. AEMO says this reflects the lower risk of regulated transmission revenues.

There is an economic basis for differentiating risk. But this assumption substantially changes the competition between:

regulated transmission;
merchant generation;
and merchant or contracted storage.

The underlying reason transmission investors face lower risk is that consumers generally bear more of the revenue and cost-recovery risk through regulation. A lower investor risk does not necessarily mean a lower total risk to consumers.

The ISP should show clearly how the ODP changes under:

a common financing rate;
higher regulated-network WACC;
different cost-overrun allocation;
and actual project financing structures.

AEMO then annualises capital costs, adds asset-specific financing costs and discounts future annual amounts at a uniform 7%.

That may comply with the methodology, but the combination is sufficiently complex that headline figures cannot be interpreted without a full reconciliation. A $6 billion annualised present value is not remotely equivalent to the amount consumers will ultimately finance through regulated charges.

14. Transmission cost estimates remain too dependent on project proponents

AEMO says it updates network options, costs and representations through joint planning with transmission businesses and jurisdictional project proponents.

Project timing in the ISP is frequently “advised by proponent”, and the document acknowledges that subsequent proponent advice can differ from the timing modelled in the final ISP.

This creates an obvious information imbalance:

proponents possess the detailed engineering information;
proponents benefit from projects proceeding;
AEMO relies on their cost, timing and scope information;
and the model then recommends that those projects advance.

It does not mean the information is false. It means independent challenge is essential.

For projects with uncertainty ranges of plus 50% or plus 100%, a single central estimate should not determine actionability. AEMO should publish:

independent reference-class forecasts;
historical cost-overrun distributions;
probabilistic P50/P90 outcomes;
and the ODP under upper-range costs.

Without that, the cost uncertainty is disclosed but not meaningfully allowed to threaten the decision.

15. The ISP exhibits continuity bias

AEMO says stopping and restarting project engagement can disrupt communities, erode trust and make social licence harder to maintain. It cites this as one reason consistency between ISPs matters.

There is some truth in that.

But it can become an institutional argument for continuing a project because it has already been announced. That is another form of sunk-cost thinking.

Continuing an unnecessary, poorly routed or increasingly uneconomic project can erode trust far more seriously than pausing it.

Community-engagement continuity should never become a substitute for:

reconsidering the identified need;
testing different routes;
revisiting the scale or voltage;
or considering non-network alternatives.

Social licence is not maintained by refusing to reopen decisions. It is maintained by demonstrating that decisions genuinely remain open when material circumstances change.

16. Demand-side alternatives are treated less rigorously than transmission projects

AEMO states that transmission planners maximise the existing network before considering new projects.

The main report does not provide enough evidence to verify that assertion.

It identifies approximately $214 million of voltage-management investment that could unlock almost 4 GW of additional CER export capacity. It also acknowledges broader distribution upgrades, but says the $214 million is additional to those wider investments.

What is missing is a transparent hierarchy showing, region by region:

existing line thermal and stability capability;
dynamic line-rating opportunities;
series compensation and power-flow control;
substation and transformer upgrades;
distribution-level alternatives;
local batteries and demand response;
reconductoring or uprating;
and only then new corridors.

The ISP lists some smaller network alternatives but does not show the avoided kilometres, avoided easements and avoided cost available from a systematically optimised existing-network-first strategy.

In practical network planning, that comparison should be visible, not asserted.

17. Gas is a hidden parallel infrastructure system supporting the electricity plan

The ODP depends on 17 GW of flexible gas generation by 2050. AEMO acknowledges that pipeline constraints and gas shortfalls could threaten delivery of the plan.

Yet gas infrastructure is not co-optimised with electricity infrastructure.

AEMO also says:

gas price effects have not been assessed;
commercial viability has not been tested;
upstream exploration, drilling and extraction costs are excluded;
and gas investment remains the responsibility of the gas industry.

Even the counterfactual assumes gas generation with carbon capture and storage while excluding the infrastructure required to transport and store the captured carbon.

That is a major system-boundary omission.

A plan cannot fully establish electricity-system least cost while depending on a parallel fuel and infrastructure system whose:

prices;
commercial feasibility;
upstream expenditure;
storage requirements;
and some emissions-management infrastructure

are not included in the same optimisation.

The phrase “backed by gas” hides a substantial contingent infrastructure and fuel-security obligation.

18. Reliability is demonstrated, but not comprehensively proved

The ISP illustrates reliability through a 15-day winter period involving intermittent renewable lulls across the southern mainland, supported by storage, gas and imports.

That is useful modelling, but it is not by itself a comprehensive resilience case.

A genuinely forensic reliability assessment should expose outcomes under combinations such as:

prolonged correlated low wind and solar;
low hydro inflows;
gas supply constraint;
major transmission outage;
coincident generator outages;
bushfire or extreme-weather corridor outages;
delayed restoration;
and multiple failures during planned network outages.

The ISP also states that system restoration is not included among the power-system needs considered in the ISP because it is handled through other frameworks.

That may be administratively legitimate, but it means the “secure and reliable least-cost system” headline does not incorporate every service necessary to recover from widespread failure.

Similarly, the main report says projected mainland inertia remains above the system-wide minimum to at least 2042, despite the plan extending to 2050.

This may be addressed in supporting material, but the headline assurance exceeds what the main report demonstrates.

19. Coal-retirement robustness is asserted using sensitivities not reproduced in the final ISP

Step Change projects that almost two-thirds of remaining coal capacity retires by 2035, faster than announced.

Coal flexibility, failure rates and retirement timing materially affect:

replacement generation;
system strength;
storage;
gas requirements;
and transmission timing.

AEMO says it tested faster and slower coal retirement in the Draft ISP, but did not reproduce those sensitivities for the final ISP. It instead says their limited variance “strongly suggests” the final ODP remains robust.

For a plan whose final inputs changed materially after the draft, that is insufficient.

Final inputs should be tested in final sensitivities, particularly where the model depends on accelerated coal exit to establish urgency for replacement infrastructure.

20. The generation pipeline assumption deserves more scepticism

AEMO says 38 GW of new grid-scale wind and solar is required by 2030, with a particularly large shortfall in wind projects. It assumes 90% of connection applications reach output, based on a stated historical 10% non-progression rate.

Even using AEMO’s assumptions, the current pipeline does not meet the ODP requirement, and both the base case and constrained case require delivery rates above recent history.

This is less a forecast than an expression of what must happen for the plan to work.

The ISP should distinguish more clearly between:

modelled requirement;
announced project pipeline;
financially committed projects;
projects with planning approval;
projects under construction;
and credible deliverable capacity.

Treating connection applications as near-deliverable supply risks overstating readiness while using the same pipeline to justify network investment.

21. Environmental and community costs are converted mainly into delay and project expenditure

AEMO says transmission costs include planning, approvals and the estimated time and cost of obtaining social licence.

That treats community resistance largely as a delivery-cost variable.

The main cost-benefit categories do not visibly monetise:

landscape and visual loss;
biodiversity fragmentation;
agricultural disruption;
bushfire exposure;
cultural heritage impacts;
cumulative effects of transmission and REZ development;
health and wellbeing impacts;
or the distributional inequity between host communities and metropolitan beneficiaries.

Some of those impacts enter planning costs, compensation or route design. That is not the same as valuing the underlying social and environmental loss.

AEMO also says it receives information about complex community considerations from project proponents and incorporates it into modelling.

Again, the party seeking to deliver the project is an important source of information about its own social impacts.

The report later calls for a shared national narrative about the purpose and benefits of the transition.

That risks framing opposition as a communications problem rather than evidence that the project, route, scale or decision process may be deficient.

22. Some benefit and cost categories are either omitted or ambiguously treated

AEMO says the ODP selection does not consider network losses, ancillary-service costs or competition benefits, although its fuel-cost category also refers to the effects of dispatch on electrical losses.

That apparent inconsistency needs reconciliation.

Some omissions may work against transmission, such as excluding competition benefits. Others may work in either direction, including losses and ancillary services.

The broader point is that “net market benefit” is a defined regulatory metric, not a complete social welfare measure. It should not be communicated as though every material cost and benefit to consumers and communities has been captured.

The specific Victorian problem

The ISP does not simply exclude every Victorian project because VicGrid is responsible. It includes some Victorian actionable and future projects in the model.

But the architecture allows substantial Victorian expenditure to escape fresh national scrutiny through three mechanisms:

Anticipated status, as with Western Renewables Link.
Policy constraints, including offshore wind, renewable and storage targets.
Jurisdictional delivery, which can bypass the conventional ISP RIT-T pathway.

At the same time, the model assumes WRL and VNI West unlock 24 GW of new Victorian wind and solar, including 9 GW of offshore wind.

The result is a self-reinforcing system:

Victorian policy creates renewable targets; the targets create generation assumptions; the generation assumptions create network need; jurisdictional planning advances the network; advanced project status removes parts of the cost from later ISP comparisons; and the next ISP optimises around the resulting network.

That is why the full cost of Victoria’s transmission transformation cannot be understood from the ISP’s $6 billion figure.

What AEMO has not got wrong

A fair assessment must acknowledge several things.

AEMO does disclose many of these limitations. It does model distribution opportunities, energy efficiency, CER, renewable lulls, delivery delays, gas dependencies and project uncertainty. It has removed or deferred some transmission projects when batteries, local generation, higher costs or changed policy reduced their benefits.

It is also legally required to incorporate government policies and operate within the National Electricity Rules. Some of the model boundary is imposed on AEMO rather than invented by it.

The problem is therefore not that every calculation is false.

The problem is that:

technical qualifications are buried;
marginal results are presented as system-wide results;
assumed projects are described as part of the optimal path without being re-evaluated;
consumer and fuel-system costs sit outside the headline;
and uncertainty rarely changes the recommended direction.

Final thoughts

The 2026 ISP does not prove that the complete proposed NEM transformation is the least-cost future for consumers.

It establishes that, within AEMO’s:

policy-constrained scenarios;
assumed project baseline;
project classifications;
demand forecasts;
financing assumptions;
technology costs;
restricted counterfactual;
and incomplete whole-system boundary,

the chosen development path performs best or close to best among the candidates tested.

That is useful planning information. It is not the comprehensive economic validation implied by the public narrative.

The greatest flaw is not a single erroneous equation. It is the cumulative effect of decisions about what is fixed, what is optimised, what is excluded, what is annualised, what is assumed to eventuate, and which uncertainties are allowed to challenge the answer.

A genuinely transparent ISP would publish a project-by-project reconciliation of all unfinished infrastructure, rerun the system without each major assumed project, co-optimise demand-side and grid-scale alternatives, show total consumer-funded expenditure, test severe non-recovery delivery cases, and report results without subjective scenario weighting.

Until that exists, the ISP should be read as AEMO’s preferred pathway within an increasingly predetermined policy and infrastructure architecture, not independent proof that the entire architecture is economically optimal.