SITELIVE 
Why deep storage is back on the plan
AEMO's 2026 Integrated System Plan splits the storage task in two. By 2050 the optimal development path needs about 35GW of shallow and medium storage — the four- to twelve-hour batteries now flooding the connections queue — plus 5GW of deep storage delivered by new pumped hydro, on top of the 7GW of existing hydro-electric capacity. The distinction is physical, not financial: batteries move energy within a day, while reservoirs move it across weeks and seasons and cover extended wind and solar lulls. AEMO's illustration is Snowy 2.0, which would deliver up to 350GWh over a week, roughly the average needs of three million households. The plan counts Snowy 2.0 and Kidston as committed, adding 2.5GW by the end of 2028, with Borumba anticipated to add another 2GW. Each of those three numbers currently depends on a tunnel.
Snowy 2.0: a tunnelling programme with turbines attached
The engineering is going better than the headlines suggest. In May 2026 Snowy Hydro confirmed that TBM Lady Eileen Hudson — 2,300 tonnes — broke through into the underground cavern complex, completing a 6km tailrace tunnel to Talbingo Dam after an earlier 2.85km access tunnel, and finishing her work on the project. TBM Florence, the machine that famously bogged in soft ground, is approaching the halfway point of a 15km drive and is now setting its own best rates, while a fourth machine, Monica, custom-built for the geology of the Long Plain Fault Zone, bores the opposite end of the same tunnel; the two will meet mid-tunnel and be dismantled underground. Focus is shifting to the power station cavern itself, almost a kilometre down, with a workforce of around 5,000. This part of the project behaves like heavy civil works because that is exactly what it is.
The commercial picture is the problem. The Australian National Audit Office's June 2026 performance audit reports project expenditure of $11.1 billion at 31 March 2026 against the $12 billion budget set at the 2023 reset, finds Snowy Hydro's management of delivery only “partly effective”, and identifies governance deficiencies that have affected value for money. Two findings matter most to anyone who builds underground for a living: there was no agreed baseline schedule, so the auditors could not determine how far from completion the project actually is; and the incentivised target-cost framework adopted at the reset has not delivered the expected productivity gains or held contractors to milestones. A detailed cost reassessment was due to report by mid-2026, with cost risk sitting primarily with the Commonwealth.
Kidston and Borumba: the private test and the public one
Kidston, in north Queensland, is the first pumped hydro station built in Australia in 40 years — 250MW and about 2,000MWh, using the upper and lower pits of a worked-out gold mine, with a funded construction cost of around A$777 million at financial close in April 2021 and dispatch rights sold to EnergyAustralia. It registered in the NEM in November 2025, and AEMO's medium-term availability data, tracked by WattClarity, has since shown full availability sliding towards July 2027; Genex's chief executive has told trade press there are no technical issues, only the complexity of civil works on a method the country has not attempted in four decades. Borumba is the public test, and it is slower: Queensland formally stopped the far larger Pioneer-Burdekin scheme in November 2024 on viability grounds, transferred Borumba's oversight to QIC, and is running a revised business case alongside exploratory works — geotechnical drilling and a $20 million interim civils package awarded to FKG in May 2026 — which Queensland Hydro states expressly do not represent a decision to construct.
The honest economics, and what they demand of delivery teams
Deep storage is an infrastructure asset priced by an energy market: committed a decade before the scarcity it is built for arrives, costed against ground conditions nobody can fully characterise beforehand, and repaid out of price spreads the battery fleet is actively narrowing. Snowy 2.0's history — a $2 billion announcement, a $4.5 billion feasibility figure, a $6.1 billion approval, a $12 billion reset, now exceeded — is not proof that pumped hydro cannot work; the original Snowy Scheme cost the equivalent of more than $8.5 billion in today's money and still earns. It is proof that estimating tens of kilometres of tunnel through fault zones as if it were a power project is how the number goes wrong. Read plainly, the ANAO report is a contract-administration document: incentivised target-cost contracts, now the default after the collapse of fixed-price appetite, only function with an agreed baseline, trustworthy productivity data and milestone evidence both parties accept. For contractors the lesson is unglamorous — record measured advance rates per shift, ground conditions and support classes at the face, standing time with cause, and the date each milestone's evidence was actually produced.