Intro

AEMO's Integrated System Plan, the ISP, is one of the most thorough documents in the Australian energy market. The ISP is a least-cost capacity expansion plan, covering the next ~30 years. It works out the cheapest mix of generation and transmission required to both keep the lights on and hit emissions targets. It lays out when and where to build transmission capacity and supply, while trying to find the cheapest way to do so, for a given scenario and sensitivity.

A least-cost central plan is the truth if one coordinator builds the entire grid. The NEM isn't built that way - it's mostly privatised; so every project has to find its own financing and clear its own hurdle rate. A plan that's optimal for the system as a whole isn't automatically one the market will agree upon.

It's important to flag - this is not an attack on the ISP. Rather, it is meant as an additional commercial layer to support the understanding of the challenges of the energy transition.

TLDR

• The ISP is a least-cost plan, built with certain objectives and constraints.
• We share some considerations around key ISP scenarios and assumptions: Demand, Supply, Network, Fuel.
• We'd love your feedback.

Prices: nothin' to worry about

Smooth sailing ahead: the AEMO scenarios suggest we're in for sustained low prices. That's probably the first warning sign.

Again, these are not Varial's view of future prices. They're what our market model spits out when you plug in AEMO's 2026 Final ISP scenarios and assumptions. But the results are what you'd expect to see from a long-term equilibrium model aimed at hitting both emissions targets and reliability requirements. In this world, new build always arrives on time, always clears its cost, and competes prices back to the cost of the marginal entrant. It's a tidy story, but we don't think it's a useful one for anyone carrying real market risk. The rest of this report is really one argument for why that flat line is too smooth to trust.

Demand: that looks deep

Everything starts with the load the grid has to meet. At the quarterly level the ISP's demand looks fairly reasonable. Although, are data centres the new rooftop PV and we'll continue to underforecast these for years? Or are they the new green hydrogen?

But hang on, let's zoom in. Look inside the daily profile and the assumed shape shifts load in a way that's hard to reconcile with actuals. To be fair to AEMO, the 2026 final assumptions pulled this back from the 2026 draft, the trough isn't cut as deep as it was, but a big gap remains that we can't reconcile - even with more distributed battery uptake. Worth noting, of course, that 2026 is a partial year, and as sunshine picks up again, it may even out.

The issue with this is that it sets up daytime prices for collapse. With demand that low during the day, you can expect a lot of negative or $0/MWh prices during the day, with solar and coal min-gen setting prices.

Supply: glass half empty

Supply is where a plan and a market butt heads, because every megawatt is a commercial decision, not made by a central planner.

Can we even build it this fast?

Is the sheer build out required deliverable? Line up what the grid has commissioned per year through history against what the plan needs commissioned per year from here, the committed and anticipated pipeline plus the ISP's optimised new build, and it looks staggering. Perhaps more worrying is the rushed and subsidised build out of solar capacity that isn't actually lowest cost for the system in the short term, just to try to hit emissions targets. Pair that with the next section and the picture gets harder still.

Optimism - a forecaster's downfall

You can see the slippage already in AEMO's own inputs: compare the commissioning dates in the IASR against the more recent GenInfo, and 22 of the 30 committed and anticipated projects we can match have already moved back, most by six months or more. Optimistic dates feed straight into the supply build, and the supply build sets the price, so the optimism compounds downstream. Note: we use the latest GenInfo in our modelling.

And it may not pay

Say the build does arrive on time. The next question is whether it earns its keep. As more solar comes on, the middle of the day fills with cheap energy and the price each solar farm actually captures falls, so the more you build the less each one earns. The chart below tracks that captured price, wind and solar, across every scenario and weather year.

Now put that capture next to what a new plant needs to earn. Take the IASR's own build cost, spread it over the asset's life at the IASR's own cost of capital, and divide by the energy it generates, and you get the price each technology has to capture just to break even. Set the two against each other and it isn't a fair chunk of the plan's new build that fails to clear its cost of capital, it's very nearly all of it. On this run new solar never earns what it needs in any year, capturing roughly a third of its break-even price, and new wind only clears until 2028 before it too runs $10 to $25 a megawatt hour short for the rest of the decade. The red shaded gap is the subsidy, or the carbon price, per megawatt hour that someone has to provide for the plan's build to actually get financed. With the end of the LRET - visibility on this is going to fall off a cliff. Again, worth remembering: this isn't what we expect VWAPs to be in commercial reality, just what we would expect if AEMO's assumptions hold.

There's a subtler supply assumption worth pulling out too. Every new wind and solar project in a renewable energy zone (a REZ, an area earmarked for lots of new build) is effectively given the same smooth, zone-average generation trace. That's a reasonable approximation if you picture the fleet spread evenly across the whole viable area, with one project's lull covered by another's gust. Real build-out doesn't look like that. It comes in a handful of very large projects clustered in the pockets of the zone. A lumpy, concentrated fleet swings harder than the average trace: sharper peaks, deeper troughs, more of the extreme intervals that actually set prices. The smooth trace assumes away exactly the volatility a trader cares about.

Network: the neglected child

The really hard part of capacity expansion is network expansion - not supply. Defining the project, abstracting it to a model, estimating the timeline - all horribly difficult tasks. Again, like supply, human nature is to optimistically assume a better timeline than reality might reflect. Line up some of the flow-path and REZ expansion projects in the 2024 ISP against where the 2026 ISP now puts them, and the dates have moved by years, mostly later.

And the ISP editions only show you the survivors. AEMO keeps a separate register of every transmission project and logs each change to it, and that log is where the churn really shows. Across its last ten releases, a bit over two years, it records roughly sixty timing changes, twenty-five scope revisions and nearly twenty projects removed or downgraded out of the actionable list. Whole flagship items get quietly reclassified: QNI Connect dropped back to a future-ISP project, the Central Queensland to Southern Queensland expansion replaced with a smaller option, Marinus Stage 2's scope cut back. These are the projects hitting the community opposition, landholder access and planning walls that a least-cost optimiser assumes away, and VNI West and the New England links have run into exactly those. When a line slips or disappears, the REZ behind it is stranded, the generation that depended on it slips with it, and the two compound rather than cancel. For a planner that's a revision between drafts; for a position it's the difference between profit and loss.

Fuel: am I upside down?

Coal and gas still set the marginal price often enough that the fuel path does a lot of the work in any forecast, and it's the input where we most openly differ from the plan. This is also where that flat price line from the top of the report comes from: the ISP's fuel trajectories (coal from a consultant in 2024, gas refreshed by the same consultant in 2025 for the 2026 ISP) glide down smoothly and never really move, so the prices they drive glide down smoothly and never really move either. Fuel doesn't behave like that. It gaps on a cold snap, a supply outage, a pipeline constraint, an export shock, a global conflict. A path with the volatility ironed out is a path that has quietly assumed away most of the risk in the market, and we'd expect materially more shocks than these curves carry.

The scenario logic looks inverted to us as well. Take Accelerated Transition: it runs a faster, higher-growth, more electrified economy, and pairs it with softer commodity prices. Those two don't sit together. A world building and electrifying that hard is a world pulling harder on gas for firming and on coal through the overlap, which should carry firmer, spikier fuel, not cheaper. The faster the transition, the more the residual thermal fleet gets leaned on at the worst moments, and that's a fuel-cost story the smooth downward curves don't tell.

Standing on the shoulders of giants

None of this is a knock on the ISP. It is the right process for the job it's trying to achieve - capacity planning for a grid that has policy objectives, and it's an honest starting point for a market view, which is why our free sample runs on its scenarios. What VarialView adds is the read the plan was never meant to give you: price outlooks, deliverability-adjusted timing, spikier and more realistic fuel price assumptions, demand shape sensitivities; custom scenarios that allow you to understand the commercial market rather than the planning one.

If you're carrying risk in this market, the ISP tells you where the system might be heading; VarialView tells you what it's worth to your book. Have a dig through the charts in the other tabs to see the forecast in full, and if you want us to walk you through what it means for a specific deal, get in touch.