Between 26 and 30 October 2026, four Australian data centre figures were published: 3.9 GW from Data Centres Australia and DC Byte, 7 GW from the Commonwealth Bank, A$57bn from CBRE and A$40bn from the Smart Energy Council. Between them they count IT load, rated capacity and two kinds of dollars. This note puts each on inashanu’s unit, IT load at the end of 2030, shows how each converts and gives our reading.
IT load is the power available to servers in fitted data halls. Facility power adds cooling and electrical losses, and AEMO’s rated capacity is the connection capacity a site holds, which covers IT and facility load (AEMO 2026 ESOO, 25 Aug 2026). Where a source states only total power, the inashanu register divides it by 1.43, the median across the 21 campuses that publish both, interquartile range 1.21 to 1.48 (inashanu register, 21 Oct 2026), and the conversions below use that ratio and give the range its quartiles imply.
1The four figures on one basis
| Source | Date | Measure | Basis | Horizon | Figure | inashanu equivalent |
|---|---|---|---|---|---|---|
| Data Centres Australia and DC Byte, Forecast Report Issue 2 | 30 Oct 2026, data at 30 Sep 2026 | Operational capacity, total | IT load of fitted-out halls | End-2030 | 3.9 GW; 1.6 GW now, in 163 facilities | 4.36 GW of operating IT, 3.44 to 5.58 GW from P10 to P90; 1.55 GW at Oct 2026 (inashanu market model, 21 Oct 2026) |
| Commonwealth Bank | 27 Oct 2026 | Added capacity; capacity announced | Rated capacity, undefined; read as facility power | 2030; mid-2030s | About 4 GW by 2030 and 7 GW by the mid-2030s; about 23 GW announced, 76% at planning or earlier | 4 GW is about 2.8 GW of IT (2.7 to 3.3); the central case adds 2.81 GW of IT from the register’s pipeline, 1.89 to 4.03 GW from P10 to P90 (inashanu market model, 21 Oct 2026) |
| CBRE | 28 Oct 2026 | Investable universe | Dollars; scope unpublished | 2030 | A$57bn, up about 90% | None on a stated basis |
| Smart Energy Council, via PV Tech (27 Oct 2026) | 26 Oct 2026 | Added wholesale cost, NSW, Qld, SA and Vic | Dollars; 5 GW of extra grid demand from July 2028 at a 90% load factor, above AEMO’s forecast | 2029-30 to 2034-35 | A$40bn | About 3.5 GW of IT (3.4 to 4.1); inashanu does not model wholesale prices |
Sources: Data Centres Australia and DC Byte, Forecast Report Issue 2, 30 October 2026; Commonwealth Bank, 27 October 2026; CBRE, 28 October 2026; Smart Energy Council, 26 October 2026, via PV Tech, 27 October 2026; inashanu register and market model, 21 October 2026.
2Why they differ
Two of the figures are in dollars: CBRE’s sizes what it calls the investable universe, and the Council’s is a cost to electricity users. Neither gives a price per MW that would turn it into capacity. CBRE does not say how its universe is measured.
The bank and the Council measure on the grid side. The bank does not define rated capacity. Its 9.2 GW for the mid-2030s less its 7 GW of additions implies a base near 2.2 GW, which matches the 2.22 GW of rated capacity AEMO records for operating NEM data centres (AEMO 2026 ESOO, 25 Aug 2026). AEMO is also the one data source the article names. Divided by 1.43, that base is about 1.5 GW of IT, in line with DC Byte’s 1.6 GW at 30 September 2026, and the bank’s 2030 total of about 6.2 GW rated becomes about 4.3 GW of IT. The reading is ours and it does not reconcile every line: the bank’s 3.6 GW under construction or being expanded converts to about 2.5 GW of IT, below the 3.3 GW of IT the register records as under construction where a figure is published (inashanu register, 21 Oct 2026).
DC Byte and inashanu give the total operating at the end of 2030, including what runs today. The bank’s 4 GW and 7 GW and the Council’s 5 GW are additions. The bank’s 23 GW and the register’s 29.3 GW of pipeline IT load, where a figure is published, are proposals, and most lack approval: the bank puts 76% at planning or earlier, and 82% of the register’s pipeline has no planning consent yet (inashanu register, 21 Oct 2026). At 1.43 the bank’s 23 GW is about 16 GW of IT, and with no project list published the gap to the register cannot be traced.
DC Byte and inashanu share the unit and the 1.4 GW opening stock at the end of 2025, so the 0.46 GW between them in 2030 comes from method. DC Byte uses a statistical model fitted to market variables such as sold capacity (Forecast Report Issue 2, p. 9, 30 Oct 2026). The inashanu model follows each tranche through its stage and its operator’s disclosed schedule. The two paths are close at the end of 2027, at 2.3 GW and 2.22 GW, and part in 2028, when DC Byte adds 0.4 GW and the inashanu median adds 0.98 GW as contracted schedules fall due (inashanu market model, evidence to 21 Oct 2026).
3Our reading
To 2030 the capacity forecasts are close once they share a unit: 3.9 GW from DC Byte, about 4.3 GW of IT from the bank on our reading of its basis, and 4.36 GW in the inashanu central case. Each is Australia’s operating total, including what runs today. What separates them is whether capacity already contracted is built by 2030, chiefly CDC’s 555 MW due by the end of FY29 and NEXTDC’s 452 MW of binding new build by mid-2030 (inashanu market model, evidence to 21 Oct 2026). A year’s slip in every disclosed schedule moves the inashanu median only to 4.29 GW. Setting the schedules aside takes it to 3.50 GW, below DC Byte. The central case keeps the contracted capacity because the delivery record supports it: 94% of the phases that began construction reached operation, in a median of 17 months (inashanu delivery record, 21 Oct 2026).
The larger numbers describe proposals and scenarios. In the central case about 10% of the register’s pipeline is operating by the end of 2030: 2.81 GW of IT at the median, and 1.89 to 4.03 GW from P10 to P90 (inashanu market model, evidence to 21 Oct 2026). The Council’s 5 GW of extra demand, about 3.5 GW of IT, exceeds that central addition before AEMO’s own forecast growth is counted, so for 2030 we read it as an upper-range case. The model stops at 2030 and takes no view of the years after. DC Byte’s next six-monthly issue is the first test of the gap that opens in 2028.
This note is research and analysis. It is general information only, it is not investment advice, and it does not take account of any reader’s objectives or circumstances.