The headline optimal-development-path cost fell from $122B to $106B — a 13% cut that reads like progress. Underneath, transmission costs per kilometre rose 25–55% in real terms, 45 GW of grid-scale batteries deferred a major interconnector, and household grid demand began a structural decline. This compares the 2024 and 2026 plans, the feedback they trigger, and who gains and loses.
The headline splits across four asset classes — but they are recovered in two completely different ways. Generation and storage are mostly merchant: built by investors, recovered through the market or contracts-for-difference. Transmission is almost entirely regulated: it enters the asset base and is billed back to consumers over 30–50 years. Transmission is the smallest slice and the one that overran — and it is the slice you can't escape.
The cut is largely an artefact. Transmission WACC was re-cut from 7% to ~3% (a regulated asset earns a lower risk premium), projects delivered since 2024 drop out of the forward cost, and solar & batteries got cheaper. In delivery terms the plan is more expensive and harder to build than it was two years ago — the opposite of what the press number implies.
Overhead lines are up 25–55% real; substations 10–35%. Drivers: materials & labour shortages, every NEM project bidding for the same crews at once, and social licence — re-routes and mitigation after inadequate prior consultation. AEMO now re-tests previously "actionable" projects against this database; some fall below the benefit threshold and get deferred.
The single biggest structural shift in the plan. Batteries become the largest source of dispatchable capacity through the 2030s, ahead of coordinated household CER and supplemented by Snowy 2.0 and Borumba. This is the force that defers transmission (see row 8) and rewires the whole least-cost calculus.
No precedent in 25 years of the NEM. 36% of suitable homes already have rooftop solar; ~600,000 have batteries; on 4 Oct 2025 CER met >60% of demand for the day. Customers keep connecting (+5.8%) while energy through the network falls — transmission-to-distribution volume is already down ~10% (≈14 TWh) since FY19. The revenue base is decoupling from the asset base.
Queensland's LNP Energy Roadmap keeps coal running, pushing the NEM-wide exit out to ~2049. QLD consumers now pay twice: for ageing, less-reliable coal and for the infrastructure meant to replace it. A delayed exit also raises system risk if replacement capacity arrives late.
~1,435 km less new transmission than 2024 projected. Less wire is needed because batteries firm within regions instead of moving power between them. Fewer kilometres, but each one costs much more — the net is a flatter, battery-heavier topology.
Actionable + future projects save consumers ~$30B versus a no-transmission counterfactual — rising to $61B under a Higher-Demand (data-centre) sensitivity. On about $6B of transmission investment — roughly 6% of the $106B ODP — that is the headline case for building. The benefit grows precisely because the coal-exit and data-centre scenarios are higher-stakes.
The clearest causal signal in the plan: a $5B+ interconnector, previously actionable, was demoted to Future ISP because 45 GW of grid-scale BESS now commoditises inter-regional firming more cheaply. If this repeats, TNSPs face stranded-asset risk on 30-year regulated investments. Storage didn't just enter the merit order — it started rewriting the transmission map.
~600,000 households now have batteries — more than half installed in the last year alone under the federal rebate. This is the accelerant behind the demand collapse (row 4) and the death-spiral loop. Yet AEMO actually revised down coordinated-CER (VPP) assumptions — the AER queried the inconsistency — because consumer batteries can't be cycled like merchant BESS.
A stylised model of the plan's own logic. Move household CER, the battery pipeline, and the real cost of wires, and watch the outputs and feedback loops respond. Directions are faithful to the ISP; magnitudes are illustrative, not forecasts.
The plan's deepest tension isn't technical — it's that four actors' incentive gradients point in contradictory directions. Each is behaving rationally inside its own revenue model. The conflict is structural, not behavioural.
45 GW queued; now the central dispatchable technology. They deferred an interconnector.
Outsized bill cuts vs their network-cost contribution — protected by a 10-year reform glide-path starting ~2030.
Even +39 TWh of load doesn't break the ODP — a bankability signal. Can even firm the grid by chasing solar.
First Demand Side Factors statement; new mandate over data-centre & large-inverter standards.
Carry a rising share of fixed network cost: up to +$250/yr by 2040 absent reform. An upward, regressive cross-subsidy.
30-year regulated assets deferred as storage commoditises firming. Early works & social-licence spend at risk.
Throughput decline erodes returns; the volumetric revenue model must be redesigned before ~2030.
Pay for ageing coal to 2049 and its eventual replacement.
The 2026 ISP is sound planning meeting a delivery machine that can't keep up — and a revenue model wired for a one-way grid in a two-way world. The cost fell on paper while the system grew more expensive, more battery-shaped, and more politically combustible underneath.
The most useful thing in the 2026 ISP is not its forecast — it is the size of the correction from 2024. In eighteen months, AEMO's own roadmap moved further than the decades-long transition it claims to chart. The size of that correction is the lesson: it points to one root cause and several conclusions that hold beyond this plan.
Each 2024 central estimate is a prior distribution. The 2026 actuals are the likelihood. Drag the dial: at full trust in AEMO's plan the posterior stays near the 2024 number; as you let the evidence speak, it snaps toward what actually happened — and the four parameters update in the same direction, away from the optimistic edge.
The ISP is a least-cost optimiser fed point estimates — and almost every uncertain input was set at its optimistic edge: GenCost cost-decline curves, 53% of home batteries orchestrated into VPPs, social licence treated as a schedule risk rather than a costed line item, ageing-coal reliability held flat, and a 2,000-strong atlas of pumped-hydro sites treated as buildable when Queensland's own process advanced two. Each assumption is individually defensible. Run them through a cost-minimiser and the optimism doesn't average out — it multiplies, producing an "Optimal Development Path" that was never a feasible development path. The headline $122B→$106B fall is the same effect in miniature: a lower discount rate makes a system that became harder and more expensive to build look cheaper.
The ISP optimises a snapshot of 2050 and assumes the world holds still on the way there. But the NEM is a closed-loop control problem — long delays, strong feedback, and behaviour that reacts to the plan itself. Least-cost optimisation is open-loop: it feeds a target forward and assumes delivery. Four things systems thinking sees that the optimiser cannot —
The CER death-spiral, social-licence cost loops, RAB arbitrage and demand destruction are caused by the plan's own prices — not external shocks the model can bracket as "sensitivities."
Household battery coordination isn't a dial you set to 53%. Because most people keep their battery for their own backup, real participation settles near 1 in 5 — they won't hand an operator control just because the model needs them to.
6–12 years from "actionable" to consumer impact, against a rulebook that takes ~15 years to change. Decisions land long after the conditions that justified them have moved.
A controller senses error and corrects each cycle. A roadmap commits capital to one forecast and discovers the error only after it is sunk.
Four 2024 commitments that 2026 walked back. In each, a policy or financing preference was overruled by thermodynamics, economics or consumer behaviour — after capital and corridors had already been committed.
The simplest rule the 2024→2026 record gives us for reading the next plan: don't treat AEMO's central forecast as the middle of the range. Treat it as the optimistic end, and adjust each key number toward what can realistically be built and delivered.
What 2028 will ratify — not reverse: batteries beat wires for intra-regional firming; the binding constraint migrates from transmission to the distribution edge; household grid demand keeps decoupling from connections; and the volumetric RAB / tariff model is formally reopened. The direction of every 2024→2026 swerve was toward the consumer-led, storage-heavy, distribution-constrained system — expect 2028 to make that official.
Plan for feasibility, not optimality, and re-plan on a short cycle. The central forecast will be wrong — the 2024 plan proved that in 18 months — so what matters is how quickly the next plan corrects. A two-yearly roadmap that assumes on-time, on-budget delivery is too slow a feedback loop for a system changing this fast.