The Energy Generator Queue

Power generation is now limited by manufacturing capacity, not by capital — and that is the opening.

1. The queue

Three companies build a large-frame gas turbine: GE Vernova, Siemens Energy and Mitsubishi Heavy Industries.

GE Vernova closed its second quarter of 2026 with 116 GW of gas equipment backlog and slot reservations, up from 83 GW at the end of 2025, and expects at least 125 GW under contract by December. It is taking reservations for 2031 delivery. Siemens Energy holds a 69 GW turbine backlog inside a record €136 billion company-wide book. Mitsubishi Heavy Industries is sold out into 2028 while doubling capacity.

Wait times across the H, J and HA classes run five to seven years depending on frame and location. Turbine prices have roughly tripled since 2019. A turbine ordered today arrives around 2030 or 2031. The turbines that will run data centres in 2027 were ordered in 2023 and 2024.

1.1 What is actually scarce

Not steel, and not capital.

The hot section — the blades and vanes that sit in a gas path a few hundred degrees above the melting point of the metal they are made from, spinning under enormous centrifugal load — comes from a thin bench of specialist foundries. Casting capacity cannot be conjured. A new line is measured in years.

The second constraint is people. The welders, machinists and commissioning contractors were let go a decade ago when nobody was building thermal plant, and were never replaced. Where a delivery slot exists, there is not always a crew to install the machine.

This is why money does not clear the queue. GE Vernova has committed nearly US$600 million to its United States factories over two years, taking HA-class output at Greenville, South Carolina from 37 units in 2025 to 62 in 2026 and 74 in 2027, with a further US$41 million at Schenectady. That is a serious response, and it still does not arrive before the end of the decade.

1.2 The rest of the plant is queued too

A power station is a system, and the electrical equipment is now as constraining as the prime mover. This part of the queue applies to every technology, including solar and batteries.

Large power transformers ran 7 to 14 months before the pandemic. They now average 128 weeks, with generator step-up units at 144 weeks. Demand for generator step-up transformers has grown 274 per cent since 2019. High-voltage circuit breakers run about 125 weeks. Medium-voltage switchgear runs beyond 90 weeks against a schedule that once assumed 24.

The upstream cause has the same shape as the turbine problem. Grain-oriented electrical steel and the copper that goes with it come from a handful of producers, and new capacity is measured in years.

1.3 Equipment lead time and time to power

Two things get conflated in public debate: how long the machinery takes to arrive, and how long the plant then takes to build. The total is what a country experiences.

The public argument about generation is conducted in dollars per megawatt-hour. That argument assumes the machine can be bought. It cannot be bought before the early 2030s at any price, because the constraint is casting capacity and skilled labour.

The relevant question is no longer which generator is cheapest per unit. It is which generator can be built inside the queue.

1.4 The workaround, and what it costs

The market has already found a way around the turbine queue, and it is worth stating plainly because it is the strongest argument against everything that follows.

Reciprocating engines — gas-fired piston engines of 4 to 23 megawatts, assembled in banks into plants of 450 megawatts and beyond — deploy in 12 to 24 months against 36 to 84 months for heavy-duty turbines. Data centre developers have moved them from backup duty to primary power at scale. INNIO booked its largest ever order, 2.3 GW, with VoltaGrid. Wärtsilä has accumulated about 2 GW of United States orders since 2025, including 790 MW in Texas and a 412 MW plant in Ohio built from 40 engines and running off-grid as primary power. Caterpillar engines are being used for 2 GW of on-site generation at a West Virginia campus. Caterpillar’s own backlog has grown more than three and a half times since early 2024. Taken together, the reciprocating engine buildout is approaching the same scale as GE Vernova and Siemens Energy combined.

The engineering case is real. A bank of engines fails one unit at a time rather than all at once. Individual units can be shut down instead of running the whole plant at inefficient part load, and a modern gas engine reaches full load in about two minutes. Medium-speed engines achieve roughly 46 per cent electrical efficiency in simple cycle — the highest simple-cycle efficiency of any prime mover.

What it costs is efficiency and fuel. A combined-cycle plant converts nearly 60 per cent of its fuel to electricity; an engine plant does not. The industry is knowingly trading long-run efficiency for speed to energisation, because months of delay strand billions of dollars of computing hardware. And the workaround is filling up behind itself: engine lead times have stretched to 18 to 36 months as the orders above have landed.

The relevant point for this memo is narrower. An engine plant is a gas plant. It moves the constraint from the machine to the fuel — which in Australia is the subject of section 4.1, and is not the safe ground it is in Texas.

2. The other queue: transmission and HVDC

Three European firms — NKT, Prysmian and Nexans — hold about 75 per cent of the high-voltage direct current cable market.

NKT’s high-voltage backlog stood at €13.5 billion in the first quarter of 2026, and roughly 95 per cent of it is contracted to European transmission operators. Factory slots for 525 kV-class subsea HVDC are booked tightly enough that award timing now reflects manufacturing windows rather than developer schedules. Prysmian is enlarging its Karlskrona plant in Sweden into the largest high-voltage offshore cable factory in the world, with a further expansion at Cologne tied to awards from the German grid operator.

Converter stations are the same shape. Hitachi Energy, Siemens Energy and GE Vernova took roughly 60 to 65 per cent of 2025 order intake between them. Hitachi disclosed a US$4.2 billion HVDC backlog and Siemens US$3.8 billion. Converter transformers account for about a quarter of station cost and carry their own long backlogs.

One component narrows the path further. Only Hitachi Energy and Siemens Energy have proven 400 kV, 9 kA hybrid direct-current circuit breakers — enough on its own to move meshed project schedules by six to twelve months.

Installation is queued as well. Cable-laying vessels are as constrained as the factories: Prysmian operates seven and takes delivery of an eighth in early 2027.

Chinese suppliers — NR Electric, C-EPRI and Xi’an XD among them — are winning international work at 20 to 30 per cent below Western bids, with compliance and security requirements limiting their traction in European markets. That is a real alternative and a real risk, and it is named here because a memo about supply chains that ignores the largest low-cost supplier is not being honest about the market.

The conclusion of the first two sections is a single sentence. It is not just the generators. The wires, the converters and the ships that lay them are queued too.

3. Why it does not clear

The order books above were built in about eighteen months, and the conditions that produced them are structural.

Three demands arrived at once. Artificial intelligence and data-centre load forecasts landed on top of the electrification of transport and heat, which landed on top of the replacement of thermal plant reaching end of life. Each of the three would have been a normal decade of work. Together they exceed what the world can currently manufacture.

Supply cannot answer at the same speed, for reasons that are physical rather than commercial. A turbine plant is not a factory that can be run harder; it is a chain that ends at a foundry with a finite number of investment-casting furnaces and a workforce that takes years to train. The same is true of an electrical-steel mill and of a cable plant with a single vertical extrusion tower. The industry consolidated to three or four suppliers per product during the decade when nobody was ordering, and the capacity that was closed in that period is not sitting idle waiting to be restarted. It is gone.

A delivery slot has consequently become an asset in its own right. Reservation agreements, deposits and slot resale are now part of how the market works, which means the queue rewards whoever can commit capital earliest — not whoever needs the power most.

3.1 What a seven-year wait does to a project

A long lead time is not a delay. It changes the asset.

Interest accrues across the whole build. Prices index across a queue measured in years, so the contract signed is not the cost incurred. Demand forecasts made in 2026 govern plant delivered in 2033, and the forecast error compounds over the interval. Technology risk runs the same way: what was ordered may be superseded by what exists when it lands.

The distributional effect matters more than any of these. A hyperscaler can pay for a slot years ahead of need, from its balance sheet, at a scale a public utility cannot match. When generation is allocated by who can commit capital first, public-interest generation loses to private compute, and the public pays for the difference through the price of everything else.

4. Australia’s exposure

Australia has no domestic manufacture of large gas turbines, no HVDC cable plant and no converter-station capability. Every machine in the replacement of the coal fleet must be bought from the same firms every other country is queuing for.

That is the first exposure. The second is that the country has not joined either queue in a way that secures its position. The largest HVDC cable book in the world is 95 per cent committed to European transmission operators. Australia is not at the back of that queue. It has not entered it.

The scale of what is being planned makes the gap concrete. Anthropic was reported in August 2026 to have approached the New South Wales government about siting up to five gigawatts of data centre capacity in the state. Australia has about 160 operating data centres, ranked second in the world for data-centre investment attractiveness in 2024, and saw national occupancy rise roughly fortyfold between 2005 and 2025, two-thirds of that since 2020. In March 2026 the Investor Development Agency endorsed 15 data centre projects worth A$51.9 billion for prioritised support.

Five gigawatts is a demand a turbine order book delivers in 2031.

4.1 The fuel is not there either

At National Cabinet on 26 August 2026 the Commonwealth sought agreement that new data centres be powered by additional renewable generation. That position did not survive into the outcome. A flexible approach was adopted on the basis that the states are not all the same, and Queensland and the Northern Territory secured the ability to use coal and gas. The rules are not settled: a further leaders’ meeting is scheduled for later this year and Commonwealth legislation is not expected until 2027.

Set the gas half of that against the Australian Competition and Consumer Commission’s published position on the east coast market.

• Domestic east coast gas supply is in structural decline, with structural shortfalls projected from 2027 unless supply increases or demand falls.
• The southern states — Victoria, New South Wales, South Australia, Tasmania and the Australian Capital Territory — required additional gas every month between April and September 2026, and depend on gas transported from Queensland as their own reserves deplete.
• The July 2026 report expects supply to tighten again in winter 2027, with shortfall risk in the second and third quarters if LNG producers export all their uncontracted gas.
• Queensland LNG producers and their associates control 84 per cent of commercially viable east coast gas resources.

A carve-out permitting gas-fired data centres is therefore being written over a market that is already short, in the state the southern states rely on to get through winter. The turbines to burn that gas arrive in the 2030s and the gas itself is contested from 2027.

This is also where the reciprocating engine workaround of section 1.4 stops travelling. Engines are quick to deliver, and every one of them burns gas. In Texas that is the point, because the gas is abundant and already contracted. In Australia the fast machine arrives into a market the regulator expects to be short, in the state the southern states depend on. The workaround solves the constraint Australia has least of a problem with, and worsens the one it has most.

Australia is not facing one bottleneck. It is facing three: the machine, the fuel and the wire.

5. Eight buyers, no buyer

Australia does not procure power infrastructure as a country. It procures as eight jurisdictions, each with its own scheme, targets, tenders and balance sheet, with a Commonwealth scheme running across the top. Western Australia is not even on the same grid: the South West Interconnected System is not physically joined to the eastern market, so a national plan covers part of the nation.

National Cabinet on 26 August demonstrated the consequence rather than describing it. The Commonwealth could not hold a single energy standard across the federation, two jurisdictions took a carve-out, and the legislation is a year away while the equipment queue runs to 2031.

What the fragmentation costs, in the terms the first four sections have established:

• No single order is ever large enough to build a factory. A manufacturer sizing a production line needs a decade of guaranteed volume. Eight staggered state tenders, each for one project, is not that. It is the reason no supplier has built a plant here for it.
• Australian states compete against each other for the same scarce slots, in the same queues, from the same three or four firms, bidding up a price they could have negotiated once, together.
• Eight balance sheets borrow separately for infrastructure serving one continent, at a time when state debt is the fastest-growing government debt in the country.
• The schedules do not align, so the aggregate demand a manufacturer would need to see is never visible as a single number, even where it exists.

Federation is constitutional. The states own their networks and their planning powers, and a national approach cannot mean abolishing either. It does not have to. What is proposed here is a single aggregated forward order across jurisdictions — one order book and one counterparty for the manufacturer — leaving each state’s projects and planning intact.

The reason Australia cannot get a position in these queues is not size. Australia is large enough. It is that Australia never arrives as one buyer.

6. What Australia already has on order

The first five sections describe a global constraint. This one asks the narrower question: what has Australia actually ordered, from whom, and when will it arrive.

6.1 The machines

Snowy 2.0. The electrical and mechanical plant — six reversible Francis pump-turbines of 333 MW each, three of them variable-speed, six motor-generators, the auxiliary systems and the whole plant automation — was contracted to Voith Hydro of Heidenheim, Germany, in April 2019, after a tendering process of almost two years. Voith’s own project documentation records that purchasing for the order was executed through its German and Chinese purchasing departments across 2019 to 2026. The units total 2,000 MW with 175 hours of storage.

Marinus Link. The 750 MW HVDC cable system for Stage 1 — 255 kilometres under Bass Strait and 90 kilometres underground in Gippsland — was contracted to Prysmian of Italy in August 2024 for about €600 million, including the booking of the company’s most advanced cable-laying vessel. The converter stations at Heybridge and Hazelwood were contracted to Hitachi Energy of Switzerland in May 2024. The civil works and equipment installation went in December 2025 to TasVic GreenLink, a joint venture of DT Infrastructure — part of Malaysia’s Gamuda group — and Samsung C&T of Korea, for $994 million. The converter stations themselves contain 220/320 kV transformers, gas-insulated switchgear, filter banks and reactors, all imported.

Delivery has been under way for years and is not finished. Power station components — turbine guard valves, spiral cases and draft tube liners — began arriving at Lobs Hole from early 2024, while the runners were still in manufacture and under quality-assurance testing; each is 4.2 metres across and weighs 31 tonnes. Voith’s registered project scope has purchasing activities running from 2019 to 2026. Installation follows the civil works rather than leading them: excavation of the machine hall and transformer hall was completed only in February 2024, and the transformer hall has still to receive its six generator step-up transformers. An order placed in 2019 was therefore still being procured in 2026, for a station expected to operate in 2029.

Not one of those packages is manufactured in Australia. The Australian content in both projects is civil works, installation, survey and subcontracting. Marinus Link’s own supplier programme invited about a thousand local businesses to compete for supply-chain opportunities, which is real work and is not the same as making the equipment.

6.2 The lead times inside those orders

Read the order dates against the delivery dates and the memo’s central figure appears in Australian procurement records rather than in a manufacturer’s investor presentation.

The Snowy 2.0 electro-mechanical order was placed in April 2019, its components have been delivered progressively since 2024, its purchasing ran to 2026, and the auditor-general’s June 2026 report points to commercial operation in September 2029. That is a decade between placing the order and generating power, and seven years between placing it and finishing the buying. Marinus Link’s cable was ordered in 2024 for delivery in 2030, and its converters in May 2024 for the same date — six years, on a project whose construction has not yet properly begun.

Neither of those waits was caused by the current shortage. Both orders were placed before the order books described in sections 1 and 2 filled. A project scoped today faces a longer wait than either, from suppliers with less capacity to spare.

6.3 Meeting the demand curve

Set that against what is required. The coal fleet is scheduled to close through the 2030s. Data centre proposals now before governments include a reported approach for up to five gigawatts in a single state. The Commonwealth has endorsed fifteen data centre projects worth $51.9 billion and has said it will require large facilities to underwrite new generation.

Against that demand, Australia’s firm equipment position is two projects: a 2,000 MW pumped hydro station ordered in 2019 and arriving at the end of the decade, and a 750 MW interconnector ordered in 2024 and arriving in 2030. Everything else that has been announced is either not yet ordered, or is photovoltaic and battery capacity, which is the one category that can still be bought at short notice.

The demand curve is being met by placing orders in other countries’ queues, at the prices and delivery dates those queues set. That is a policy choice, and it has been made by default rather than in public.

6.4 What the delivery record does and does not prove

The two projects also carry a cost and schedule record that any proposal of this kind has to answer. Snowy 2.0 was announced in 2017 at an indicative $2 billion for completion in 2021, reset in 2023 to $12 billion for December 2028, and assessed by the auditor-general in 2026 as heading to roughly $20 billion, 262 days behind schedule as at July 2025, with a line-by-line reassessment still unreported. Seven completion dates have now been announced for it. Marinus Link moved from two cables to one and from $3.86 billion to more than $5 billion. Project EnergyConnect rose from $1.5 billion to $2.3 billion. The market operator lifted its transmission cost estimates by about 30 per cent across the board in 2020, and these projects have exceeded even that.

The documented causes are civil, contractual and political: ground conditions that did not match the design, a tunnel-boring machine stuck for months, a design the operator itself called immature at commencement, easements, route selection and social licence. None of them is a manufacturing failure. Nothing here went wrong because a factory could not produce.

That distinction runs both ways, and both are stated. A programme to manufacture equipment is not more of what went wrong at Snowy 2.0; it is the part of the system Australia has not attempted. But the record is a standing warning that a national programme will not run to time merely because it is national, unless the civil and contractual failures are treated as the first engineering problem rather than the last.

7. The rule: build what can be built inside the queue

A gas turbine is forged, cast and hand-finished, and the world builds roughly 60 gigawatts of them a year. The world produced 728 gigawatts of solar modules in 2024, from an installed manufacturing capacity of about 1.5 terawatts a year, and battery cell output is measured in the thousands of gigawatt-hours.

That is a difference in the shape of the supply curve, not a difference in price. Photovoltaic modules and battery systems can be ordered, manufactured, shipped and commissioned inside the time it takes to reach the front of a turbine queue — and while the queue lengthens, module and cell capacity continues to expand.

Three counterweights belong here rather than in a critic’s reply.

First, the grid-equipment queue from section 1 applies to solar and storage as well. A solar farm needs a generator step-up transformer and a connection, and those run 144 and 128 weeks. This is a real constraint on any generation build in Australia, and it is one of the strongest arguments for making that equipment here.

Second, battery cells have their own supply constraints in materials and cell capacity, and the market is concentrated in a way that carries its own risk.

Third, firming is an engineering problem, not a rounding error. Photovoltaics and batteries alone do not deliver a winter week of low sun without long-duration storage behind them, which is the subject of section 8.

With those stated, the finding stands: on delivery speed, photovoltaics and storage are first, and they are first by a margin measured in years.

8. Long duration, and the machines we could actually make

Long-duration storage is where the queue reshuffles the merit order, because the machines involved are not the ones that are sold out.

8.1 Pumped hydro

Pumped hydro uses Francis pump-turbines, supplied by a different set of manufacturers from the gas-frame market. The binding constraint in Australia is not the machine. It is civil works: tunnelling, excavation, geotechnical risk and the schedule and cost overruns that follow. Snowy 2.0 is the case study, and no memo advocating pumped hydro can wave it away. What the experience argues for is honest geotechnical investigation before commitment, and site selection where the head is high and the works are short — the basis of the Alice Hub proposal in this series.

8.2 Concentrated solar power with molten salt storage

Concentrated solar power stores heat rather than electricity, dispatches on demand, and generates through a conventional steam turbine — an item outside the gas-frame bottleneck.

The honest history is that Australia has no concentrated solar industry. The Aurora project at Port Augusta, which would have been the largest solar thermal plant in the world, was cancelled. The technology has been more expensive per megawatt-hour than photovoltaics with batteries, which is why it lost the argument the first time.

While that argument was being lost here, it was being won somewhere else.

China had 1.14 GW of concentrated solar in operation by mid-2025 across 27 connected projects, with a pipeline beyond 8 GW in Qinghai, Gansu, Inner Mongolia and Xinjiang. Article 25 of its Energy Law, in force from January 2025, directs the active development of the technology, and guidelines issued in December 2025 target 15 GW installed by 2030 at a levelised cost competitive with coal.

The plants are no longer demonstrations. The Golmud 350 MW project, which broke ground in June 2026, carries 15 hours of molten salt storage — 11,747 megawatt-hours of stored heat — across 3.7 million square metres of mirrors, and is expected to generate about a billion kilowatt-hours a year. Fifteen hours of storage on top of a day of collection is what round-the-clock output means in practice, and the operating record supports it: the Delingha tower ran thirteen days continuously, the Dunhuang tower fourteen, and a 100 MW trough plant generated 342 GWh from solar heat alone in a single year, at 3,422 equivalent full-load hours. The Hami hybrid pairs 900 MW of photovoltaics with a 100 MW solar thermal unit that runs eight hours after dark — and it is worth being precise, because it is the eight-hour block from the thermal unit that continues, not the full gigawatt.

Two findings from that record matter to Australia, and neither is about cost.

The first is what the machine does for a grid. A salt plant generates through a synchronous steam turbine, so it supplies inertia and system strength as a by-product of generating — along with ramp rates of 3 to 6 per cent a minute and peak-shaving of up to 80 per cent. Those are the services a departing coal fleet takes with it, and they are not what a battery is built to provide. This is a different job from the one photovoltaics and batteries do, not a more expensive version of the same job.

The second is the supply chain, and it is the reason this section sits in this memo at all. China’s concentrated solar supply chain is now more than 90 per cent localised for critical components, with production capacity targeted at 5 GW a year, and costs have fallen as scale arrived: tower plants moving from about ¥0.85 per kilowatt-hour to under ¥0.60, with an 18 per cent reduction available simply from building at 300 MW instead of 100 MW.

That is this memo’s argument, executed. A country identified a generation technology whose components it could manufacture, mandated a forward order large enough to justify the factories, localised the supply chain, and drove the cost down by building at scale. Australia has better direct sunlight than Qinghai across most of its interior, it has the salt, and it has the steel and heavy fabrication industries the plants consume. What it has not had is the order.

The case is worth reopening on delivery, on system strength and on sovereignty. It is not yet won on cost against photovoltaics with batteries, and this memo does not claim otherwise — but the cost gap closed in China through scale, which is precisely the variable a twenty-year order book changes.

8.3 The point that carries forward

A steam turbine, a Francis pump-turbine and a high-voltage cable are conventional heavy manufacture. A gas turbine hot section is not. The technologies that avoid the queue are, with few exceptions, the same technologies Australia could build for itself.

9. What Australia manufactures, and what it does not

A proposal that claims Australia can make everything gets dismissed. This section states both halves.

Realistic within the programme: photovoltaic module assembly and, at scale, cells; battery packs and cell manufacture; large power transformers and generator step-up units; switchgear and high-voltage circuit breakers; high-voltage and HVDC cable; steam turbines and Francis pump-turbines; molten-salt loops, receivers and heliostat fields, which China has localised past 90 per cent and Australia could; towers, structural steel, foundations and the civil supply chain around all of it.

Realistic under licence, and worth pursuing: medium-speed reciprocating engines. Australia will not displace Wärtsilä, INNIO or Caterpillar, and does not need to. An engine is cast iron and forged steel, machined to tolerance and assembled — heavy engineering of a kind this country has done before, and metallurgically ordinary beside a turbine hot section. Block casting, cylinder heads, machining, assembly, test and the whole lifecycle service business are achievable here under a licensing arrangement with an established manufacturer. Large forged crankshafts come from a small number of specialist forges worldwide and would be imported at first. The case for building them here is the same as the case for the rest of this list: engines are the fastest generation available, they are the technology every developer is now reaching for, and being able to supply them domestically converts a global shortage into an Australian industry.

Not realistic, and it should be said plainly: forged and investment-cast gas turbine hot sections. That capability rests on decades of metallurgy, a trained workforce and a customer base Australia does not have, and pretending otherwise would discredit the rest of this list. Converter-valve electronics are achievable only with a licensed partner, and should be pursued that way rather than from a standing start.

The list Australia can build is the list of what the technologies in sections 7 and 8 actually require. That is not a coincidence. It is the argument.

9.1 What is already being attempted, and why it stalls

Australia is already attempting most of the list above. It is doing so one company and one grant at a time, and the results show what that produces.

The Solar Sunshot programme commits $1 billion to photovoltaic manufacturing. Against that, Tindo Solar at Mawson Lakes in South Australia remains the country’s only operating panel manufacturer, and is scaling from 20 MW to 180 MW a year on $34.5 million of it, with a feasibility study for a one-gigawatt plant. The Hunter Valley Solar Foundry drew a $171 million federal and New South Wales co-investment for a commercial-scale module plant. SunDrive holds $25 million from the renewable energy agency for its copper metallisation process and a joint venture with a Chinese manufacturer. Upstream of the module, nothing is under construction at all: the ingot, wafer and polysilicon proposals near Townsville hold feasibility funding of $4.7 million and $5 million respectively.

The cable case is the instructive one, because it is the closest Australia has come to building the thing this memo says the world is short of.

In November 2023 SunCable named Bell Bay in Tasmania as its preferred site for an advanced high-voltage subsea cable factory: roughly $2 billion of investment, 1,200 to 1,600 kilometres of cable a year, a two-hundred-metre vertical extrusion tower, 25 to 40 megawatts of power draw, and about 800 construction jobs followed by 400 permanent manufacturing roles. It would have been the first such facility in the southern hemisphere. Tasmania was selected ahead of some thirty sites worldwide, on a deep-water port, rail access and hydro power. Construction was to begin in 2025 and the first cable to be produced in 2029.

By January 2026 the project page had been taken down and the industrial precinct that would have hosted it reported no contact from the company for more than a year. SunCable has not formally abandoned the site. Its stated position is that the extended timeline for the Australia-Asia PowerLink — a final investment decision now expected in 2027 and electricity exports in the mid-2030s — means a dedicated cable factory is no longer necessary to meet the schedule, and that the cable requirement can instead be met from existing global capacity.

That is the argument of section 10, demonstrated in Australia and in reverse. Prysmian is enlarging Karlskrona because a firm order exists. Bell Bay stalled because an order moved. A deep-water port, hydro power, a supportive state government and a billionaire backer were not sufficient, because none of them is an order book.

The pattern across every project named here is the same. Each is a separate company pursuing a separate grant against demand that may or may not appear. A grant reduces the cost of building a factory. It does not tell the factory what it will sell, or to whom, or for how long — and that, not capital, is what a manufacturer requires before it commits a production line.

10. The order book is the policy

None of the above happens project by project. No manufacturer builds a production line for one purchase. A line is built against a guaranteed forward order.

The evidence is in the same industry, and on both sides of the ledger. Prysmian is expanding Karlskrona into the world’s largest high-voltage offshore cable factory, with a second expansion at Cologne tied to awards from the German grid operator. LS Cable is building an HVDC subsea plant in Virginia. Those factories exist because someone signed a long forward order first. Bell Bay did not proceed because the order moved. The order comes before the factory, in both directions.

The Sovereign Build Corporation programme set out across this memo series is exactly such an order: a twenty-year continental build at 95 per cent sovereign content, supporting about 58,000 direct jobs and an estimated 400 to 600 new Tier-2 and Tier-3 businesses. The detail is in The Sovereign Manufacturing Multiplier and the continental plan memos and is not restated here.

What this memo adds is the reason to sign it now. A domestic line takes years to stand up. So does a turbine slot. Australia can spend the same interval either waiting in someone else’s queue or building the capacity that removes the need to queue — and only one of those leaves an industry behind.

11. The rent is the infrastructure

In August 2026 the assistant minister for science, technology and the digital economy put the question publicly: whether Australia will be able to extract rent from these facilities, or whether it will end up as a commodity in someone else’s supply chain. He called it an important unanswered question.

It has an answer, and it has two parts.

The first is section 9. A country that manufactures the equipment captures the value of the buildout rather than importing it.

The second is the system itself. A data centre is mobile and its operator is usually not Australian. Land is not scarce and tax concessions compete with every other jurisdiction’s tax concessions. What an operator cannot bring with them, cannot ship, and cannot build faster than the queue is the power system: firm generation, the transmission that reaches it, and the storage that holds it up.

So Australia sells access rather than land. The programme builds the national asset — the corridor, the transmission backbone, the pumped hydro and the desert generation — and compute connects into it and pays for the privilege of connecting. That is rent in the strict sense: a return on a scarce asset that cannot be replicated elsewhere.

The structure holds for four reasons. It is durable, because the alternative to connecting is a seven-year wait for a turbine. The asset stays Australian, because the wires and the storage outlive any tenant. It compounds with the manufacturing case, because the country earns once by making the equipment and again by owning the system it is installed in. And it fits the Commonwealth’s proper role, which is not to pick a winner or subsidise one in, but to create the opportunity and let Australians hold the asset that collects the return.

Australia does not have to hope for a benefit from this buildout. It can own the thing the world is short of, and charge for access to it.

12. The sequence

The programme phases against the queue rather than against a political term.

Immediately, while the world waits for turbines. Photovoltaic and battery deployment at scale, because it is the only generation that can be delivered inside the interval. Site acquisition, approvals and grid studies for everything that follows. Orders placed now for the transformers and switchgear whose lead times are already measured in years.

Concurrently, while the factories are stood up. Transmission corridors and civil works, which are labour and earthmoving rather than imported machinery. Pumped hydro excavation, where the schedule is governed by tunnelling rather than by a supplier. The first domestic lines: transformers, switchgear and cable, which are the least exotic and the most immediately needed.

From the 2030s, as the lines mature. Domestic steam turbines and pump-turbines, heavy electrical equipment, and the salt-loop and receiver fabrication that a concentrated solar programme would require.

The timing is the argument. Everyone still queuing in the early 2030s will be taking delivery of their first frame. By then Australia is not in the queue.

13. The cost of not doing it

The alternative is not a neutral option. It is a decision with its own price.

It means buying generation and transmission equipment from a three-firm oligopoly, at prices that have tripled, on delivery dates in the 2030s that Australia does not control and has not secured. It means coal closure dates that depend on replacement plant nobody has ordered. It means the manufacturing value of the largest infrastructure buildout in a century accruing to Sweden, Germany, Japan, Korea, China and the United States, while Australia contributes land, approvals and electricity demand. And it means answering the assistant minister’s question in the way he feared — as a commodity in someone else’s supply chain.

The counterfactual is set out in full in Without the SBC. This memo adds one line to it: the window in which the decision can still be made is the length of the queue, and the queue is finite.

14. Sources

1. Gas turbine order books (§1) — GE Vernova, Second Quarter 2026 Results and accompanying investor commentary: 116 GW of gas equipment backlog and slot reservations at the close of Q2 2026, up from 83 GW at the end of 2025, with at least 125 GW expected under contract by December and reservations being taken for 2031 delivery; nearly US$600 million committed to United States factories across two years, with HA-class output at Greenville, South Carolina rising from 37 units in 2025 to 62 in 2026 and 74 in 2027, and a further US$41 million at Schenectady.
2. Turbine backlogs and capacity (§1) — Siemens Energy quarterly reporting: a 69 GW gas turbine backlog within a record €136 billion company-wide order book. Mitsubishi Heavy Industries: large-frame capacity sold out into 2028 with capacity being doubled.
3. Lead times and pricing (§1, §3) — industry supply-chain analyses of H, J and HA-class delivery: wait times of five to seven years by frame and location; turbine prices approximately tripled since 2019; combined-cycle construction schedules lengthening from about 3.5 years in 2023 to about five years, and to seven for some heavy frames; investment-casting capacity for hot-section components and the loss of skilled welding, machining and commissioning labour identified as the binding constraints; slot reservations, deposits and resale as an established market practice.
4. Grid equipment lead times (§1, §7) — Wood Mackenzie supply-chain survey data and International Energy Agency transmission analysis: large power transformers averaging 128 weeks against a pre-pandemic 7 to 14 months, generator step-up units at 144 weeks, high-voltage circuit breakers at about 125 weeks and medium-voltage switchgear beyond 90 weeks against a former 24; demand for generator step-up transformers up 274 per cent since 2019; grain-oriented electrical steel and copper supply identified as the upstream constraint.
5. HVDC cable and converter market (§2, §10) — company reporting and market analyses for NKT, Prysmian, Nexans, Hitachi Energy, Siemens Energy and GE Vernova: approximately 75 per cent of the HVDC cable market held by three European suppliers; NKT high-voltage backlog of €13.5 billion at Q1 2026, approximately 95 per cent contracted to European transmission operators; Hitachi Energy HVDC backlog of US$4.2 billion and Siemens Energy US$3.8 billion; 60 to 65 per cent of 2025 converter order intake taken by three suppliers; converter transformers at about a quarter of station cost; 400 kV, 9 kA hybrid DC breakers proven by two suppliers only; Prysmian operating seven cable-laying vessels with an eighth due in early 2027; the Karlskrona and Cologne expansions and the LS Cable HVDC plant in Virginia; Chinese suppliers bidding 20 to 30 per cent below Western competitors.
6. National Cabinet, 26 August 2026 (§4, §5) — Australian Broadcasting Corporation coverage of the leaders’ meeting on data centres and energy: the Commonwealth’s proposal that new data centres be powered by additional renewable generation was not adopted; a flexible approach was agreed on the basis that jurisdictions differ; Queensland and the Northern Territory retained the ability to use coal and gas; rules remain unsettled, with a further leaders’ meeting scheduled and Commonwealth legislation expected in 2027; the assistant minister for science, technology and the digital economy publicly raised whether Australia can extract rent from these facilities or will become a commodity in a global supply chain.
7. Australian AI and data-centre policy (§4) — Australian Government, National AI Plan (December 2025); Expectations of data centres and AI infrastructure developers (23 March 2026), including the expectation that developers invest in Australian supply chains; Investor Development Agency endorsement of 15 data centre projects worth A$51.9 billion (27 March 2026); the Prime Minister’s address of 15 July 2026 setting out a requirement that large AI data centres underwrite new power supply.
8. Data-centre scale and investment (§4) — reported approach by Anthropic to the New South Wales government regarding up to five gigawatts of data-centre capacity (August 2026); approximately 160 operating data centres in Australia; second global ranking for data-centre investment attractiveness in 2024; national occupancy growth of approximately fortyfold between 2005 and 2025, two-thirds of it since 2020.
9. East coast gas market (§4) — Australian Competition and Consumer Commission, Gas Inquiry reports and quarterly releases to July 2026: east coast domestic supply in structural decline with structural shortfalls projected from 2027 unless supply increases or demand decreases; southern states requiring additional gas every month between April and September 2026 and dependent on Queensland supply as local reserves deplete; tightening expected in winter 2027 with shortfall risk in quarters 2 and 3 if LNG producers export all uncontracted gas; Queensland LNG producers and their associates controlling 84 per cent of commercially viable resources; contracted prices of approximately $13 to $15 per gigajoule. Analysed in full in Australia’s Gas Problem — and the Corridor That Fixes It.
10. Coal and nuclear build times (§1) — United States Energy Information Administration construction-duration data for coal-fired generation, showing median construction in excess of 60 months; recent Western nuclear new-build schedules exceeding a decade of construction; the last Australian coal unit, Bluewaters, completed in 2009; the prohibition on nuclear power in the Environment Protection and Biodiversity Conservation Act 1999 and the Australian Radiation Protection and Nuclear Safety Act 1998.
11. Chinese concentrated solar power (§8.2, §9) — SolarPACES, the China Solar Thermal Alliance Blue Book of China’s Concentrating Solar Power Industry, and associated 2026 reporting: 1.14 GW installed by mid-2025 across 27 grid-connected projects with a pipeline exceeding 8 GW; Article 25 of the Energy Law of the People’s Republic of China, in force 1 January 2025, directing active development of the technology; National Development and Reform Commission and National Energy Administration guidelines of December 2025 targeting 15 GW by 2030 at costs competitive with coal; the CGN Golmud 350 MW project, which broke ground 16 June 2026, with 15 hours and 11,747 MWh of molten salt storage, 3.7 million square metres of mirror field and expected annual generation of approximately 1 billion kWh; continuous operation of 13 days at the SUPCON Delingha tower and 14 days at the Shouhang Dunhuang tower; 342.23 GWh of pure solar thermal generation in a year at the CSNP Urat 100 MW trough plant, at 3,422 equivalent full-load hours; the CTG Hami hybrid of 900 MW photovoltaic and 100 MW solar thermal delivering eight hours of post-sunset output from the thermal unit; supply chain localisation above 90 per cent for critical components with capacity targeted at 5 GW a year; tower levelised costs falling from about ¥0.85 to below ¥0.60 per kilowatt-hour with an 18 per cent reduction attributable to scaling from 100 MW to 300 MW.
12. Grid services from solar thermal plant (§8.2) — peer-reviewed review of Chinese concentrating solar thermal power, 2025: peak-shaving of up to 80 per cent, ramp rates of 3 to 6 per cent per minute, and synchronous inertia supplied as a property of the generating plant rather than as a separate service.
13. Pumped hydro and concentrated solar (§8) — Snowy 2.0 cost and schedule revisions as the Australian reference case for tunnelling-driven risk; the cancellation of the Aurora solar thermal project at Port Augusta; Francis pump-turbine supply from manufacturers outside the gas-frame market. The Alice Hub pumped hydro proposal is set out in this memo series.
14. Australian equipment orders (§6) — Voith Hydro contract announcement of April 2019 with the Future Generation Joint Venture: six reversible Francis pump-turbines of 333 MW, three variable-speed, six motor-generators, auxiliary systems and plant automation for a 2,000 MW station with 175 hours of storage, with purchasing executed through Voith’s German and Chinese purchasing departments from 2019 to 2026. Marinus Link Pty Ltd contract announcements: Hitachi Energy contracted in May 2024 for the Heybridge and Hazelwood HVDC converter stations; Prysmian contracted in August 2024 for the 750 MW, 345 kilometre HVDC cable system at approximately €600 million, including the booking of its principal cable-laying vessel; TasVic GreenLink, a joint venture of DT Infrastructure (Gamuda group, Malaysia) and Samsung C&T (Korea), contracted in December 2025 for $994 million of converter station construction, equipment installation and 90 kilometres of land cable civil works; converter station equipment lists including 220/320 kV transformers, gas-insulated switchgear, filter banks and reactors. Snowy Hydro project updates from February 2024 for the progressive delivery of turbine guard valves, spiral cases and draft tube liners to Lobs Hole, runner manufacture and dye-penetrant testing of the 31 tonne, 4.2 metre runners, and completion of machine hall and transformer hall excavation in February 2024. Australian National Audit Office performance audit, June 2026, for the Snowy 2.0 schedule position and the September 2029 commercial operation estimate. Cost and schedule figures for Snowy 2.0, Marinus Link, Project EnergyConnect and the 2020 transmission cost revision as previously cited in this list. The $30 to $40 billion all-in figure for Snowy 2.0 is advanced by project critics, is disputed, and is not adopted in this memo.
15. Manufacturing throughput (§7) — International Energy Agency Photovoltaic Power Systems Programme, Trends in PV Applications 2025: 728 GW of solar modules produced worldwide in 2024, an increase of 18.5 per cent on 2023, against installed module manufacturing capacity of approximately 1.5 TW a year; REN21 Global Status Report 2025 for the capacity figure and the concentration of the manufacturing value chain in China. Global gas turbine output of approximately 60 GW a year is derived from the three manufacturers' disclosed capacity and expansion statements and is stated as approximate. Battery cell output of roughly 1,100 GWh in 2024 is quoted in energy rather than capacity and is not directly comparable with the generating-capacity figures; it is given for scale only.
16. Reciprocating engines (§1.3, §1.4, §4.1, §9) — industry and company reporting through 2026: INNIO’s largest recorded order of 2.3 GW to VoltaGrid; Wärtsilä accumulating approximately 2 GW of United States orders since 2025, including 790 MW in Texas and a 412 MW plant in Ohio comprising 40 engines operating off-grid as primary power; Caterpillar engines supporting 2 GW of on-site generation in West Virginia; Caterpillar backlog growth exceeding 3.5 times since early 2024; the aggregate engine buildout approaching the combined scale of GE Vernova and Siemens Energy; deployment in 12 to 24 months against 36 to 84 months for heavy-duty turbines; unit sizes of 4 to 23 MW and plants beyond 450 MW; simple-cycle electrical efficiency of approximately 46 per cent and full load in about two minutes for current medium and high-speed machines; combined-cycle efficiency near 60 per cent as the comparison; engine order lead times extending to 18 to 36 months. Efficiency and fuel-saving claims made by equipment vendors in their own marketing are not adopted here.
17. Australian manufacturing projects (§9.1) — Australian Renewable Energy Agency and Department of Climate Change, Energy, the Environment and Water announcements under the $1 billion Solar Sunshot programme: $34.5 million to Tindo Solar to expand Mawson Lakes output from 20 MW to 180 MW a year with a feasibility study for a one-gigawatt facility; $171 million in federal and New South Wales co-investment for the Hunter Valley Solar Foundry; $46 million to 5B; $25 million to SunDrive from the agency for copper metallisation, alongside its joint venture with Trinasolar and its agreement with AGL for the Hunter Energy Hub; feasibility funding of $4.7 million to Stellar PV for a 2 GW ingot and wafer proposal and $5 million to Solquartz for solar-grade polysilicon, both near Townsville. SunCable Bell Bay: company and Tasmanian Government announcements of November 2023 — approximately $2 billion of investment, 1,200 to 1,600 kilometres of cable a year, 25 to 40 MW of power demand, about 800 construction and 400 manufacturing jobs, construction from 2025 and first cable in 2029, selected ahead of about thirty international sites. Current status from Tasmanian press reporting of January 2026: project page withdrawn, no contact with the Bell Bay Advanced Manufacturing Zone for more than twelve months, and the company’s stated position that the extended Australia-Asia PowerLink timeline — final investment decision expected in 2027, exports in the mid-2030s — allows its cable requirement to be met from existing global capacity.
18. The programme (§10, §12, §13) — Sovereign Build Corporation programme parameters as set out across the MMA memo series: a twenty-year continental build at 95 per cent sovereign content, approximately 58,000 direct jobs and an estimated 400 to 600 new Tier-2 and Tier-3 businesses; detailed in The Sovereign Manufacturing Multiplier, the continental national plan memos, and the counterfactual in Without the SBC.