At a glance

  • “The grid can’t cope” ran against EVs from Australia’s 2019 election to California’s 2022 heatwave, before data centres inherited the line.

  • Engineering studies and Australian trials put a home-charging EV’s diversified evening-peak contribution between about 110W and 2.4kW, against a 7.4kW plug rating.

  • AEMO now models EV charging annually through CSIRO projections: up to 6.5 TWh a year by 2030 and more than 60 TWh by 2050, from roughly 1 TWh today.

  • Data centres are a different load type: concentrated at transmission level, near-constant, and connected through a process the customer funds.

  • AEMO’s 2026 ISP already stress-tests faster data centre growth, in a Higher Demand sensitivity adding about 39 TWh by 2050.

The objection returns at gigawatt scale

Anthropic is running a tender for up to 1.4GW of Australian data centre capacity, a build reported at up to US$15 billion, and our coverage of that tender drew a familiar response. If Australia could never find enough grid for electric vehicles, the objection runs, where are the gigawatts for AI coming from?

Australia has heard “the grid can’t cope” before; the EV version was worked through in trials, tariffs and AEMO’s planning documents. The parallel also has limits. EV charging and data centre demand are different problems for a grid planner, in size, shape and connection point, and AEMO’s 2026 Integrated System Plan treats them that way.

“End the weekend”: the EV grid scare, 2019 to 2022

When Labor took a 50% EV sales target to the 2019 election, Scott Morrison told voters an electric vehicle would not tow their trailer or their boat, and that Bill Shorten wanted to “end the weekend”. Michaelia Cash promised to “save their utes”. Those attack lines were about towing, price and practicality; grid capacity was not among them. The capacity worry ran alongside in commentary, and University of Melbourne power engineers addressed it mid-campaign, finding that uncoordinated charging could overload local networks at around 10% EV uptake, while coordinated charging could sustain more than 80%.

The question kept running for years. As late as August 2024, CarExpert published a three-part investigation titled “Can the Australian power grid handle EVs?”, and its concluding part answered that it can, provided charging is managed.

The fear peaked in California in September 2022, on 25 August 2022 the state’s Air Resources Board adopted a rule ending petrol-only car sales by 2035. Six days later, as a record heatwave built, the grid operator CAISO issued the first of a run of consecutive Flex Alerts asking residents to charge EVs before 4pm and avoid charging between 4pm and 9pm. Newsweek’s headline ran “Californians Told Not to Charge Electric Cars Days After Gas Car Sales Ban”, and versions of it circled the world. The advisory was voluntary, and it asked owners to shift charging earlier in the day. On 6 September demand hit an all-time record above 51GW, an emergency alert cut load by more than 2,100MW within minutes, and California avoided rolling blackouts. EV charging at the time was less than 1% of California’s peak load, a point the corrective coverage from Scientific American and Axios made within a fortnight.

Engineering put the figure near 250W per car at evening peak

A 7.4kW home charger is one of the larger single loads a household can add, and the distribution-edge worry was concrete: if every car on a street charged at 6pm, local transformers would feel it. Between 2020 and 2023, ARENA co-funded trials with Jemena, AGL and Origin Energy to test whether drivers actually charge that way.

Jemena’s EV Grid trial recorded individual maximum demand of 5.9kW to 6.7kW, close to the plug rating, but a diversified maximum across the fleet of 1.06kW to 2.44kW per vehicle, with cars charging on about one day in three. AGL’s orchestration trial measured an average uncontrolled contribution of about 450W per charger, falling to roughly 50W when charging was orchestrated, and participants overrode the controls an average of three times a day across 200 chargers. Origin’s trial cut charging in the 3pm to 9pm window by 79%. ARENA’s 2023 synthesis across the trials reported no observable increase in load between 3pm and 9pm in unmanaged charging profiles, with solar households charging in the middle of the day.

The Electric Vehicle Council’s engineering study put the diversified figure at about 250W per vehicle during the 4pm to 9pm peak, and estimated 1.5 million EVs by 2030 would add roughly 375MW, about 1% of combined peak demand across the NEM and Western Australia’s main grid. Its December 2024 update, with measured trial data folded in, found real-world contributions of 110W to 240W per vehicle, well below the 2kW figures earlier research had assumed.

EV load moved inside AEMO’s planning cycle

The 2018 ISP mentioned EVs only to forecast “a small impact” on grid demand, with no consumption figures. By the 2022 ISP they were central to a forecast near-doubling of NEM consumption by 2050. CSIRO now writes AEMO a yearly EV projections report, and AEMO’s April 2025 data paper expects charging to grow from about 1 TWh today to up to 6.5 TWh a year by 2030 and more than 60 TWh by 2050, the 2050 figure comparable to NSW’s current annual generation.

AEMO models charging behaviour across flat, time-of-use and controlled tariffs, and updates the numbers each year. The 2023 ESOO put residential EV demand at about 11 TWh a year by 2032-33; the 2024 edition cut that to about 8 TWh on slower fleet turnover. Its reliability forecast is more conservative again: the 2025 ESOO’s maximum demand figure “only accounts for customer-controlled battery and EV charging that is not coordinated”. For reliability purposes, AEMO plans as if the smart charging never arrives.

The pricing and engineering were built alongside. SA Power Networks has run a “solar sponge” tariff since July 2020, pricing the 10am to 3pm window at a quarter of the standard rate. The AEMC followed with rule changes: two-way network pricing in 2021, universal smart meters by 2030, and separate settlement of EV charging from household load from late 2026. Networks trialled dynamic operating envelopes through the ARENA-backed evolve and AEMO’s Project EDGE, setting near-real-time limits that keep flexible load within spare local capacity. Ausgrid chief executive Marc England was “not a subscriber to the narrative” that EVs would create problems for the grid in 2023, because “you don’t need to charge a car at 6pm”.

None of this made the load free. AusNet’s 2026-31 proposal, now before the Australian Energy Regulator, includes about A$120 million of low-voltage and SWER upgrades for load it attributes partly to EVs, and expects EVs to make up about 3% of its network peak by 2031. The national fleet passed 410,000 EVs by September 2025. And in February 2024, when storms toppled transmission towers west of Melbourne, 16 vehicle-to-grid EVs in a Canberra trial fed 107kW back into the grid for ten minutes, which the trial’s researchers call the first vehicle-to-grid response to a grid emergency anywhere.

Data centres put gigawatts on single transmission nodes

The EV question was millions of small, movable loads at the distribution edge. An AI data centre is the opposite shape: one customer, one connection point, hundreds of megawatts, drawn near-constantly. AEMO’s forecast for the entire national EV fleet in 2030 is up to 6.5 TWh a year. A 1.4GW campus running continuously would draw about 12 TWh a year.

Characteristic

Home EV charging

AI data centre

Connection point

Distribution network, household level

Transmission network, dedicated connection

Load shape

Millions of small loads, shiftable by tariff

Single large load, near-constant

Peak contribution

110W to 2.4kW per vehicle, diversified

Hundreds of MW per campus at one node

Planning treatment

CSIRO and AEMO annual projections

Connection queue, ISP demand sensitivities

Source: AEMO, Electric Vehicle Council and ARENA trial reports, July 2026.

Concentration cuts both ways: a data centre cannot be tariff-shifted the way a car fleet can, but it is visible to the planner years in advance, and it pays for its own connection. Large loads enter a formal transmission connection process; at the end of the March 2026 quarter, 11 large data centres representing more than 5GW of maximum demand were working through it, on an approximately two-year path from application to energisation and a five-to-ten-year ramp to full load. The queue is screened hard: Oxford Economics, commissioned by AEMO, found six of every seven megawatts of the 44GW requested nationally is phantom demand, and TransGrid is progressing a rule change to charge large users for the capacity they reserve whether or not they draw it.

The scale is also larger. The 2026 ISP has data centres at almost 10% of underlying NEM demand by 2050, up from about 2% today, inside a total forecast to nearly double from 205 TWh to about 390 TWh. EV charging over the same period grows to more than 60 TWh.

What to watch

The transferable part of the EV precedent is procedural. The fear preceded the measurement; the measurement, the tariffs and the annual forecasting cycle then absorbed the load. The same institutions are now doing that work on data centres, and the 2026 ISP already contains the stress test: a Higher Demand sensitivity adding about 39 TWh of large-load demand by 2050, under which the transmission plan’s net market benefit rises from nearly A$30 billion to A$61 billion and the chosen build path stays within A$60 million of least-cost. AEMO identifies timely delivery of new generation and transmission as the constraint.

The shape does not carry over. EV demand proved movable by price signal; a training campus draws through the evening peak regardless of tariff, and its impact lands on wholesale prices, transmission budgets and household power bills through different channels. The current round of regulatory work is testing exactly that: the AEMC is preparing advice for energy ministers on how data centre growth affects wholesale prices and bills, and TransGrid’s use-it-or-pay-for-it proposal is before the rule maker. Underneath both sits the question of whether Australia’s grid rules reward building over optimising. AEMO’s next Electricity Statement of Opportunities is due in August 2026, and the NSW grid-connection inquiry reports by 30 September 2026.

Sources and further reading

Primary sources: