Batteries now set power prices, and data centres want more

Grid batteries have moved from supporting players to price-setters in the markets where they are most widely deployed. The Australian Energy Market Operator (AEMO) says grid-scale batteries set prices in almost half of all dispatch intervals in the September 2026 quarter on the National Electricity Market (NEM), up from 18 per cent a year earlier. In the United States, the Solar Energy Industries Association (SEIA) reports sharply lower summer wholesale prices in Texas and California after years of solar and storage build-out.

The same week brought a different kind of storage news: proposals for data centres that pair gigawatt-scale computing loads with multi-gigawatt-hour batteries, and a 1 GW virtual power plant (VPP) funded by Google. This piece sets out what the numbers show, who benefits, and where the evidence is thinner than the headlines.

What the data shows

AEMO’s Quarterly Energy Dynamics report for July to September 2026 describes a market with record demand and falling prices. Average underlying demand across the NEM reached a new high in the September quarter, up 1.3 per cent from the previous Q3 record set in 2025, yet average wholesale prices fell to $63 per megawatt-hour, the lowest September quarter average since 2021. That is 27 per cent below the same months of 2025.

For the first time in any quarter, grid-scale batteries supplied 8% of evening peak generation between 4 p.m. and 9 p.m., ahead of gas-powered generation at 5.1%. Gas output fell 34 per cent to its lowest third-quarter average since 2000, while renewables reached a record 44.3% of generation. AEMO chief executive Daniel Westerman said batteries are increasingly determining the wholesale price, both when they charge and when they discharge, reducing evening peak prices and the occurrence of negative prices during the day.

The share of intervals in which batteries set the price has climbed steadily. AEMO’s first-quarter 2026 report put it at around 32% of trading intervals, and Westerman cited 36 per cent in the June quarter.

The US picture comes from SEIA, a solar and storage trade group, so its analysis is an industry claim built on public market data. It compares June to August 2026 with the same months of 2021, using inflation-adjusted 2026 dollars and day-ahead prices in the two grids.

Average summer wholesale price, inflation-adjusted 2026 dollars
Average summer wholesale price, inflation-adjusted 2026 dollars
Market and period Headline figure Context Source
Australia NEM, Q3 2026 Average price $63/MWh, down 27% year on year Batteries set price in almost half of intervals AEMO
Texas (ERCOT), summer 2026 vs 2021 Average price down 42% Load up 30.1%; storage supplied over 1.3% of energy SEIA
California (CAISO), summer 2026 vs 2021 Average price down 51% Consumption up 2.5%; batteries provide 6% to 7% of daily energy use SEIA
CAISO peak hourly price, $/MWh 141.75 to 59.69 Summer 2021 to summer 2026 SEIA
ERCOT peak hourly price, $/MWh 110.96 to 57.30 Summer 2021 to summer 2026 SEIA

SEIA’s figures, as reported by the trade press, show the peak-hour prices falling by more than half in both states. SEIA estimates the lower prices saved Texans about $3.55 billion and Californians about $2.5 billion in summer 2026 compared with 2021. SEIA also reports that battery storage capacity grew more than 26-fold in Texas and nearly sevenfold in California over the five years.

Why storage compresses prices

The mechanism is easy to describe. Midday solar is abundant and cheap; batteries absorb it and release it in the evening, when gas peakers have historically set the price. AEMO’s own framing is that grid-scale batteries are absorbing excess renewable energy during the day and shifting it into the market during evening peaks, helping moderate prices during high-demand periods.

Home batteries add to the effect. Westerman says that homes with batteries draw an average of 70-75% less power from the grid at peak times than homes with rooftop solar alone, and he credits the federal home battery rebate and the “solar sharer” requirement for shifting behaviour.

A likely reading is that the biggest price effect sits in the shoulder of the day, between sunset and the late evening, and that it grows as battery fleets reach the size of the evening peak. Neither AEMO nor SEIA claims storage is the only cause. SEIA attributes the drop to solar and storage together, and the Australian price fall also reflects wind output and lower gas use.

Who gains, and who does not

Consumers with exposure to wholesale prices gain most directly, as do households that pair rooftop solar with batteries. Developers of solar-plus-storage have a stronger political argument at a time when electricity affordability is a campaign issue in the US.

Gas-fired generators lose volume and, in the evening peak, pricing power. Another group is less obvious: battery owners themselves. If batteries set the price almost half the time and peak prices fall by more than half, the spread between charging and discharging narrows. The data does not yet show how much this has cut battery revenues, and neither source reports it. A reasonable hypothesis is that later projects earn less per MWh of capacity than early ones, which would push the sector towards contracted revenue and longer durations.

Retail customers get a more qualified benefit. SEIA says residential generation costs fell 4 percent in Texas and 10 percent in California between the two summers, but transmission and distribution costs pushed overall residential bills up 3 percent in Texas and 22 percent in California. Wholesale savings, in other words, are being partly absorbed by network spending.

Not every region has followed the national pattern. In the June quarter, power costs in Western Australia’s Wholesale Electricity Market rose by 30 per cent, even though more than 1,000 MW of battery capacity had been added there in the previous year (the AEMO quarter cited is the first of 2026). Local fuel, network and market design factors clearly matter alongside battery volumes.

The data centre pull

The next wave of storage demand may come from computing rather than from the wholesale market. Three recent proposals pair a data centre of around 1 GW with batteries sized in gigawatt-hours. All three are announced or proposed, not built.

Project Data centre load Battery, GWh Generation Status
Project Meridien, Gingerah Energy (Western Australia) Up to 1,000 MW IT 5 About 5,200 MW solar, 1,500 MW wind EPBC referral; not approved
Project Ares, Energy North (Northern Territory) 1 GW IT, two phases About 16 About 3,000 MWp solar EPBC referral
Pantheon Atlas campus (Croatia) 800 MW usable IT 8 500 MW solar Developer claim
Battery size in proposed data centre projects (not built)
Battery size in proposed data centre projects (not built)

Gingerah Energy has referred Project Meridien to the federal government, about 100 km south of Broome. A referral is an early step and does not mean a project has been approved, and the referral does not describe a connection to an external grid. The site was previously listed as a hydrogen and ammonia export project in the CSIRO HyResource database.

Energy North’s Project Ares sits on a pastoral lease in the Northern Territory. The company says the facility would operate off-grid and draw no power from the NT network, and the referral ties the design to the Australian government’s National Data Centre Expectations of March 2026, which require new data centres to invest in new renewable generation alongside construction. It gives no power rating for the battery.

In Croatia, the investment group Pantheon Atlas has claimed that power for a planned 1 GW campus will be supplied entirely by renewables, with four 400 kV lines also enabling an additional 5.2 GW of renewables onto the grid, per the developer.

A likely reading is that developers see batteries as the firming layer that lets a large load meet a renewables requirement, or avoid a long grid queue, without gas. Whether a 5 GWh battery can deliver the uptime a data centre needs through multi-day low-output spells is an engineering question the referrals do not answer.

Virtual power plants as a second route

In the US, Google is buying flexibility rather than building storage. On 22 October 2026 it and Voltus added 1,070 MW to an initial 100-MW virtual power plant agreement signed in June in the PJM Interconnection. Voltus pools flexible capacity from businesses and homes, such as smart thermostats, batteries and industrial assets, and participants receive bill savings and cash payments.

The headline number needs care. The two figures imply 1,170 MW in total, although both companies describe the project as 1 GW, and the announcement does not reconcile them. Neither company disclosed a price, a cost per megawatt-hour or a payment schedule, and the megawatts are flexible capacity rather than stored energy. Batteries are one of several resource types, so this is a demand-flexibility story with storage inside it.

The context is PJM’s attempt to make large loads bring their own supply. Facilities Dive notes PJM proposed a “bring your own new capacity” framework earlier this year. A VPP is one way for a hyperscaler to claim it is adding capacity without waiting for new plants.

The supply chain question

If data centres and utilities want tens of gigawatt-hours of batteries, the cells have to come from somewhere. The Chinese-dominated lithium iron phosphate (LFP) supply chain is the default, and a US challenger is trying a different chemistry. Peak Energy closed a $108 million Series B led by GM Ventures on 21 October 2026. It makes sodium-ion systems that it says avoid the active cooling LFP systems need, cutting the cost of energy by up to 20%.

Those are company claims; the reporting on the round notes the announcement cites no independent test behind them. Peak currently sources cells from manufacturers in China, built to its specifications. Its roughly $1.1 billion order book, according to Bloomberg, is mostly a 4.75 GWh agreement with Jupiter Power and a 1.5 GWh deal with Energy Vault. The planned 4 GWh factory in Sacramento is due to start production in the first quarter of 2027.

BloombergNEF forecasts sodium-ion demand will reach 257 GWh by 2035, from just over 1 GWh in 2025. The same coverage cautions that the cells still cost more than lithium-ion and China dominates announced capacity. Sodium-ion cells also store less energy for their size and weight, which suits stationary use better than vehicles.

Competing views and open questions

SEIA’s view is that solar and storage are the reason prices fell. Its CEO Tim Pawlenty said that affordable, reliable power transcends party lines and Texas and California are proof of the price effect. As a trade association, SEIA has an interest in that conclusion, and its comparison of two summers five years apart does not isolate storage from fuel prices, weather or demand growth. The Australian evidence comes from the market operator, which has no such stake, and it points the same way for the evening peak.

The counter-argument, visible in the SEIA data itself, is that lower wholesale prices have not produced lower bills. Network costs are rising, and storage does little to reduce those unless it defers grid upgrades.

Several questions remain open:

  • How far battery revenues fall as the fleet starts to set prices most of the time.
  • Whether off-grid data centre designs in remote Australia can meet firm-power standards, and whether any will clear federal and native title processes.
  • What Google actually pays for VPP capacity in PJM, and how much of the 1 GW is battery-backed.
  • Whether sodium-ion systems reach the claimed cost advantage at scale and with US-made cells.

What to watch

Westerman said the full September quarter report was due the following week; AEMO’s next quarterly report will show whether battery price-setting passes 50% and how a hot summer affects evening peaks. In the US, watch for SEIA-style analyses covering other markets, and for any PJM decision on how data centres must secure capacity.

For the data centre proposals, the signals are EPBC decisions and any disclosure of grid connections or battery power ratings. For Peak Energy, the milestone is first production at Sacramento in the first quarter of 2027, followed by independent performance data.

Sources

Leave a Comment