Quantum chip improved Schneider's energy forecasts by 20% in tests

Silicon Quantum Computing (SQC) said on 2 October 2026 that its Watermelon quantum chip improved the accuracy of Schneider Electric’s next-day energy forecasts by an average of 20%, and by as much as 41%, across 12 months of data. SQC and Schneider Electric, working with UNSW Sydney, have won A$3.6 million (US$2.5 million) from the Australian government to take the work into hundreds of homes.

The accuracy figures are the companies’ own, from SQC’s announcement. Schneider Electric, an energy technology group with 160,000 employees, uses forecasts like these to balance rooftop solar, home batteries and electric vehicles on the grid, and Watermelon gives its forecasting models extra inputs generated by a quantum processor.

How much better were the forecasts?

Watermelon lifted forecasting accuracy by an average of 20% over a classical benchmark, with a best gain of 41%, when the two companies applied it to 12 months of next-day forecasting data, SQC said on 2 October 2026. The forecasts covered grid consumption, the electricity drawn from the network, during stage 1 of a project funded by the Australian government’s Critical Technologies Challenge Program.

Watermelon works inside a classical forecasting model. The chip generates quantum features from the data, and the model uses them alongside its classical features, which SQC says gives “richer models with greater predictive abilities”.

Watermelon adds quantum features to a classical model

Watermelon is a quantum reservoir: SQC’s chip turns input data into quantum features, extra signals that a classical machine learning model learns from, according to SQC’s product page. SQC says the approach suits time series and sparse data. A quantum reservoir processes information through the natural dynamics of a quantum system, which gives it memory and non-linear behaviour suited to time series forecasting, Telstra and SQC wrote in a joint release in October 2025. Electricity demand is a time series, a run of readings taken one after another.

SQC was founded in 2017 and is led by Michelle Simmons, its founder and chief executive. Its processors use phosphorus atoms placed in pure silicon, according to its technology page, which puts the company among the builders of silicon qubits. The announcement calls Watermelon an “atomically engineered, quantum-enhanced AI chip”.

Watermelon has run a forecasting trial before. In a 12-month project with Telstra, the quantum reservoir forecast network metrics with accuracy on par with a deep learning model, and training and fine-tuning took days against weeks for the deep learning model, Telstra and SQC announced on 13 October 2025.

Why are homes getting harder to predict?

Rooftop solar, home batteries and electric vehicles have made household energy systems more changeable and harder to predict, SQC and Schneider Electric said on 2 October 2026. Schneider Electric uses forecasting and optimisation technology to balance these distributed energy resources, including forecasting when to switch to and from solar power.

Better forecasts can improve how those resources are managed, raise the use of renewable energy and help lower costs for consumers, the announcement says, and even modest gains in accuracy can deliver those benefits. “The energy system is becoming more dynamic with variables like rooftop solar, EVs, and home batteries creating new levels of complexity,” said Colette Munro, Pacific Zone President at Schneider Electric. She said the work with SQC showed advanced energy technology has the potential to manage that complexity better, “lowering costs for consumers and using energy more efficiently across the system.”

Who is paying for stage 2?

The Australian government is paying A$3.6 million (US$2.5 million) for stage 2 of SQC and Schneider Electric’s project, the largest of the six stage 2 grants in round 2 of its Critical Technologies Challenge Program, according to the government’s grant list. The list names Silicon Quantum Computing as the applicant, with UNSW and Schneider Electric’s Australian arm as partners, on a project for quantum-enhanced AI in energy networks.

The programme’s second round set four challenges for quantum technology: biosecurity, health for First Nations peoples, transport and logistics, and energy networks. Stage 1 paid A$100,000 to A$500,000 (about US$69,000 to US$347,000) to test whether an idea is feasible. Only teams that finish stage 1 are invited to apply for stage 2, which pays A$1 million to A$5 million (about US$0.7 million to US$3.5 million) to build working prototypes or demonstrations. Fourteen projects shared A$5.7 million (US$4.0 million) at stage 1. Six went on to stage 2, two of them in energy: SQC’s and a quantum timing project led by QuantX Labs with Siemens.

The US dollar figures use the Reserve Bank of Australia’s daily exchange rate table, F11.1, which gives A$1 to US$0.6933 for 2 October 2026, a rate that matches the companies’ own conversion of the stage 2 grant.

Hundreds of homes come next

Stage 2 of SQC and Schneider Electric’s project expands the modelling to hundreds of homes across Australia and builds it directly into Schneider Electric’s AI workflows, the companies said on 2 October 2026. SQC calls the work “a clear pathway for today’s quantum computing systems to be used in production environments”.

“We have always believed that quantum processors would work alongside CPUs and GPUs to deliver real-world performance gains,” said Michelle Simmons, adding that the results showed how quantum-enhanced AI “can address practical challenges across the energy sector and time-series datasets more broadly”. IBM published a reference architecture on 12 March 2026 that shows how quantum processors can work alongside GPUs and CPUs.

Watermelon launched in 2025 and is sold through the cloud and as hardware, including turnkey installations in data centres. SQC says customers in telecommunications, banking and high-frequency trading already use it.