• Transgrid has more than 10 GW of data centre enquiries within 12 km of Western Sydney and about 1.5 GW signed. It says capacity there is largely exhausted beyond 2033.
• Many developers had already committed to land, design and planning before the constraint was clear. That capital is now exposed to delay, unbudgeted upgrade costs or, in some cases, a site that can’t be energised at the scale it was bought for.
• Spare capacity is uneven. Analysis of UK and Irish networks found 74% of distribution networks under-utilised, rising to 93% in some areas. Knowing where the headroom sits before investing can be the difference between a connectable site and a stranded one.

In June, Transgrid told the market that the transmission network in Western Sydney “is becoming increasingly constrained, with capacity largely exhausted beyond 2033.” More than 8 GW of prospective data centre load is in advanced discussions, against about 1.5 GW of signed connection agreements. In August, Transgrid announced up to 2 GW of new capacity, to be funded by the proponents who need it. Even so, several gigawatts of proposed load has no clear path to connection this decade. Since 14 August, projects of 30 MW or more must also pass six readiness tests within three months of signing a connection agreement, or risk losing their allocation.

For developers already in Western Sydney, this is not a theoretical risk. Many secured land and progressed approvals and design on the assumption that power would follow. Some now face years in the queue, a share of augmentation costs they never budgeted for, or a smaller project than the one they planned. Land holding costs, consultant fees and planning spend don’t stop while a project waits for capacity.

The lesson isn’t that Western Sydney was a bad bet. It’s that grid capacity was a material risk in those investments, and in many cases it wasn’t measured before the money was committed. Investors who understand how much spare capacity actually exists, and where, can price that risk up front. Those who don’t find out in the connection queue.

Spare capacity is local, and mostly invisible to investors

Networks are sized for the worst hour of the worst year, then rated conservatively on top of that. Line ratings are typically set using static assumptions about temperature, wind and sag. The result is that available headroom varies sharply from one substation, line and corridor to the next.

In August 2025, Neara published analysis of UK and Irish distribution networks using LiDAR-based, physics-enabled models and digital line rating. It found 74% of those networks were under-utilised at their current maximum operating temperature, rising to 93% in some areas. It also found that raising the operating temperature by as little as 15°C could potentially double capacity on some lines. As Neara’s Taco Engelaar put it, “by extracting more from existing assets, we can speed up connections.”

Spare capacity isn’t evenly distributed. It sits on particular assets in particular places, and it only becomes useful to an investor when you know where.

Published network data shows broad constraint zones, not whether a specific substation can take 300 MW in 2029. Most operators find out through a connection enquiry, by which time land, design and approvals money may already be spent.

Headroom has a location and a time

The second dimension is time. A network constrained for a few hundred hours a year is not constrained for the other 8,000-plus.

In February 2025, Duke University’s Nicholas Institute modelled the 22 largest US balancing authorities. It found 76 GW of new load could be connected if that load curtailed just 0.25% of the time, averaging about 85 hours a year. At 1% curtailment, the figure rose to 126 GW.

That is a system-level number, and it doesn’t tell an investor where to build. Time-based headroom only turns into connectable megawatts at specific connection points, where the binding limit is a short peak rather than a thermal limit for most of the year. Finding those points means mapping capacity by location and by hour together.

Australian policy is starting to reward this. On 5 August the AEMC advised ministers that data centre connection agreements should encourage demand shifting and co-location with generation and storage. An operator that knows which sites are only constrained at peak can design storage and flexibility around that, and negotiate from a position of knowledge.

Spare capacity belongs among the key investment criteria

Data centre investment cases already weigh land, power price, fibre, latency, planning approvals and tenant demand. Grid capacity usually enters as a single answer from a connection enquiry, often late in the process. I’d argue that where the network has spare capacity deserves a place alongside those established criteria, because it has a direct bearing on several decisions investors are already making.

Before buying land, four questions should be answered for each candidate site:

1. Measured or rated? What can the substations and lines serving this site actually carry, using real asset data, conductor condition and dynamic or digital line ratings?
2. When is it constrained? How many hours a year does the local network actually bind, and what firm capacity is available outside those hours?
3. What would release more? Would re-rating, re-tensioning, targeted upgrades or storage unlock capacity faster and more cheaply than a full augmentation?
4. Is it already claimed? Spare capacity near a site is only useful if competing projects in the queue haven’t already secured it.

An operator that can answer these for three candidate sites is making a very different investment decision from one comparing three connection enquiry responses.

A map of headroom isn’t a guarantee of capacity

There are good reasons to be careful.

First, aggregate headroom isn’t site-level capacity. As energy analyst Brian Handshy has argued about the Duke study, system-level flexibility “is not the same as locational, enforceable, interconnection-creditable capacity at a specific point on the transmission system.” That is the strongest argument for mapping by location, not against it.

Second, time-based headroom has a price. Hyperscale tenants contract for 99.95% to 99.99% uptime. Using capacity that is only available outside peak hours needs on-site storage or generation, or workloads that can genuinely move.

Third, the UK findings are not Australian findings. Our networks, climate and bushfire obligations differ, and higher operating temperatures mean more conductor sag, which has to be checked against clearances line by line.

Finally, network operators have to protect system security. The AEMC’s proposed ride-through standards exist because 60 data centres in Virginia dropped 1,500 MW in a single event in July 2024. Any capacity found has to be capacity the network can operate safely, and the network operator has to agree.

None of this argues against mapping spare capacity. It argues for doing it properly, asset by asset, before the investment decision rather than after it.

The question for boards

Nobody has a reliable crystal ball on exactly how much hidden headroom Australian networks hold, or where. That is the point. Right now most investment committees don’t know either, because they have never asked for it to be mapped.

For data centre boards, the question worth asking isn’t whether the grid has spare capacity. It does, in places. The question is whether you know where those places are before you commit capital, and before your competitors do.

#DataCentres #GridCapacity #EnergyTransition #DigitalInfrastructure #CapitalAllocation

References

• Transgrid, “Update on network capacity for new large-scale loads in Western Sydney,” 17 June 2026
• W.Media, “Transgrid sets six tests for data centres seeking NSW grid capacity,” 19 August 2026
• RenewEconomy, “Network seeks to charge data centres for full grid connection capacity, whether they use it or not,” 2 July 2026
• W.Media, “Data centres could hit 30% of NSW load and drive down network costs,” 5 June 2026
• Envirotec, “Majority of UK energy networks are underutilised, says analysis using AI and digital line rating tools,” 7 August 2025
• Utility Dive, “US grid headroom for flexible load,” 11 February 2025 (reporting Duke University Nicholas Institute, “Rethinking Load Growth“)
• Brian Handshy, “Duke’s Rethinking Load Growth study is being misused in the large load interconnection debate,” Medium
• AEMC, “Data centres to bring clean, firm energy, be flexible and pay their way, AEMC advises Ministers,” 5 August 2026
• AEMC, “AEMC proposes new grid standards for data centre connections,” 12 March 2026
• Transgrid, “New transmission capacity in Sydney to be funded by data centres,” 26 August 2026