Toshiba International and DIMAAG-AI Announce AI Data-Center Energy-Storage Development Collaboration
Редакция ЦОД:Главное
Toshiba International Corporation and DIMAAG-AI have announced a collaboration to develop an energy-storage system for AI data centers. The companies plan to integrate Toshiba’s lithium-titanate SCiB cells into DIMAAG’s ZettaWatt Power Platform to manage rapid load swings, provide backup power and support grid stability.
- The collaboration covers development of an AI data-center storage system, not an announced operating deployment.
- Toshiba SCiB LTO cells are planned for integration into DIMAAG-AI’s ZettaWatt Power Platform.
- The proposed AI configuration uses Zenius modules with liquid-immersion cooling.
- The companies target continuous 10C charge/discharge operation, subject to appropriate cooling, and a service life of more than 10 years.
- No MW rating, MWh capacity, customer, site, cost or commercial deployment schedule has been disclosed.
Toshiba International Corporation (TIC) and DIMAAG-AI announced their collaboration on 9 September 2026. The work is framed as a development effort rather than a commercial deployment or an operating installation.
Toshiba is expected to supply its SCiB lithium-ion cells, which use a lithium-titanate (LTO) anode. DIMAAG-AI is to design and manufacture the battery system and integrate it into its ZettaWatt Power Platform. For AI applications, the companies envisage Zenius modules using liquid-immersion cooling.
The stated design target is continuous charging and discharging at 10C, contingent on appropriately engineered cooling, alongside a target service life of more than 10 years. These targets address a central issue in AI data-center power design: storage may need to absorb and deliver high power repeatedly as compute loads change rapidly, rather than serve solely as conventional standby energy capacity.
The companies also state that ZettaWatt could reduce periodic AI-load fluctuations to less than 1% from the grid’s perspective. That claim cannot yet be assessed because the announcement does not disclose the reference baseline, input load profile, assessment period or test methodology.
No installed system, data-center site, customer, commercial deployment timetable, power rating or energy capacity has been disclosed. As a result, the announcement is best viewed as an early-stage storage-development initiative aimed at AI data-center power conditioning, with its technical and commercial performance still to be demonstrated.
AI workloads combine high electricity demand with fast-changing power profiles. The proposed system could demonstrate how high-power storage, power electronics, cooling and controls can be integrated into AI data-center power architectures to buffer cyclical load fluctuations seen by the grid. Its commercial relevance, however, remains unproven because no power rating, energy capacity, cost, test results or deployment details have been disclosed.
- — / 12 сентября 2026 г. · ЯпонскийПервоисточник →
Материал подготовлен редакцией ЦОД: Главное на основе указанных первоисточников.