Hybrid Quantum-Classical AI Cuts Energy Forecast Error 20%

For AI and ML engineers, the technical contribution is hybrid feature engineering: quantum-generated features from Watermelon are combined with classical features to produce richer predictive models without replacing the existing stack. The reported 20% average accuracy gain on real-world next-day forecasting data suggests near-term quantum-enhanced machine learning may emerge not as a separate model class but as a drop-in feature source for production pipelines.

On 1 October 2026, Silicon Quantum Computing (SQC), an Australian quantum hardware company, and Schneider Electric, a global energy technology leader, announced they have progressed to Stage 2 of the Australian Government's Critical Technologies Challenge Program (CTCP). According to the companies, the partnership will receive A$3.6 million (US$2.5 million) in funding, with UNSW Sydney continuing as a research partner. The central technical claim is that during Stage 1, SQC's atomically engineered, quantum-enhanced AI chip, Watermelon, generated quantum features that, when used alongside classical features, improved next-day energy forecasting accuracy by an average of 20% against a classical benchmark, with gains reaching 41% over a 12-month dataset.

“

According to the companies, the partnership will receive A$3.6 million (US$2.5 million) in funding, with UNSW Sydney continuing as a research partner.

The problem the work addresses is becoming more expensive and harder to ignore. As rooftop solar, home batteries, and electric vehicles become widespread, household energy systems have grown more dynamic and difficult to predict. Schneider Electric uses advanced forecasting and optimisation technologies to balance distributed energy resources, including determining when to switch to and from solar energy deployment. Within that operational context, even modest forecasting improvements can increase renewable energy utilisation, improve management of distributed assets, and contribute to lower energy costs for consumers. The Stage 1 results, if independently reproduced, would exceed the threshold of what the companies describe as a modest improvement.

Watermelon's approach is significant because it does not seek to replace classical forecasting models but to augment them. The chip produces quantum features that are combined with classical features to deliver richer models with greater predictive ability. This hybrid architecture lowers the integration barrier for today's quantum computing systems, which remain noisy, limited in scale, and difficult to operate outside specialised facilities. By treating quantum hardware as a feature-engineering component inside an otherwise classical AI pipeline, the partnership offers a more practical path toward production deployment than a fully quantum forecasting system would.

The commercial implications differ for each partner. For Schneider Electric, quantum-enhanced forecasting could become a differentiator in energy management software, grid services, and distributed energy resource orchestration. If the accuracy gains persist across seasons and geographies, Schneider Electric could embed quantum features into its AI workflows for hundreds of homes in Stage 2 and potentially into broader commercial offerings later. For SQC, Stage 2 funding plus a corporate design partner validates Watermelon outside the laboratory and gives the company a credible near-term route to production environments. The Australian government's CTCP backing also signals strategic interest in quantum technologies beyond encryption and computing, framing quantum as a critical infrastructure tool.