Setting Up Large-Scale Battery Production for AI Data Centres: SoftBank's Blueprint

Introduction

As artificial intelligence (AI) workloads surge, data centres require reliable, large-scale energy storage to manage intermittent power demands and ensure uptime. SoftBank Group’s mobile-services subsidiary aims to address this by manufacturing large-scale battery cells at a former Sharp plant in Sakai, Osaka. With a target output of one gigawatt-hour (GWh) per year and partnerships with South Korea’s Cosmos Lab and DeltaX, the company is creating a blueprint that others can follow. This how-to guide breaks down the process into actionable steps, from site selection to next-generation chemistry adoption.

What You Need

• Former manufacturing facility – A large, repurposable site like the Sharp plant in Sakai, Osaka (ideally with existing clean rooms and power infrastructure).
• Technology partners – Experts in battery chemistry and production, such as Cosmos Lab and DeltaX (specialized in next-gen solutions).
• Investment capital – Funding for retrofitting, equipment, and R&D (SoftBank’s mobile unit has allocated significant resources).
• Regulatory approvals – Environmental permits, zoning clearance, and safety certifications.
• Supply chain agreements – Contracts for raw materials (zinc, halide compounds, and initial chemistry components).
• AI data centre clients – Offtake agreements to guarantee demand for the batteries.

Step-by-Step Process

1. Step 1: Secure a Suitable Manufacturing Facility

   Identify an existing industrial site with the right footprint and infrastructure. SoftBank chose the former Sharp plant in Sakai, Osaka – a facility with clean rooms, high-voltage power, and logistics connections. Renegotiate leases or purchase the site. Conduct environmental audits and prepare for retrofit.

2. Step 2: Forge Strategic Technology Partnerships

   Collaborate with specialized firms to accelerate production. SoftBank partnered with South Korea’s Cosmos Lab (battery chemistry expertise) and DeltaX (manufacturing automation). Formalize agreements covering IP, exclusivity, and milestone deliverables.

3. Step 3: Define Initial Battery Chemistry and Production Capacity

   Determine the first generation of cells. SoftBank’s initial chemistry (likely lithium-ion or another established type) targets a production capacity of 1 GWh per year. This output is enough to support a mid-sized AI data centre’s backup and peak-shaving needs. Plan line layouts and procure machinery.

4. Step 4: Set Up Production Lines and Begin Manufacturing

   Retrofit the facility with electrode coating, assembly, and testing lines. Hire skilled operators. SoftBank aims to start production in the fiscal year beginning next April. Implement rigorous quality control to achieve high cycle life and safety for data centre applications.

   

5. Step 5: Transition to Next-Generation Chemistry

   SoftBank’s roadmap includes switching to zinc-halide chemistry by 2027. This technology offers higher energy density, longer lifespan, and lower cost. Phase it in by adjusting manufacturing processes – e.g., new electrolyte handling and cell design. Run parallel lines during the transition to maintain supply.

6. Step 6: Scale Up and Support AI Data Centre Demand

   After proving the initial 1 GWh line, expand capacity based on demand from AI operators. Use learnings from the first plant to replicate the model elsewhere. Establish recycling programs for end-of-life batteries to ensure sustainability.

Tips for Success

• Start with proven chemistry – Begin manufacturing with a stable technology (e.g., lithium iron phosphate) to de-risk the facility before introducing novel zinc-halide.
• Plan for flexibility – Design production lines that can adapt to different cell formats and chemical processes as battery tech evolves.
• Engage early with regulators – Pre-approve environmental and safety plans to avoid delays.
• Secure long-term raw material contracts – Zinc and halides may become competitive; lock in supply with Cosmos Lab and other partners.
• Build a skilled workforce – Retrain former Sharp employees or partner with local tech institutes for battery manufacturing expertise.
• Leverage AI for production optimization – Use machine learning to monitor cell quality and reduce defects – fitting for a data centre-focused venture.