ByteDance & BYD: AI Accelerates the Megawatt Charging Era

In an era increasingly defined by efficiency imperatives, artificial intelligence is fundamentally transforming the new energy sector.

On June 18, ByteDance Seed announced that following revelations at the Volcano Engine Force conference on June 11, ByteDance Seed and Volcano Engine will deepen their collaboration with BYD's lithium battery division through joint laboratories and other mechanisms to explore AI for Science combined with high-throughput experimentation, accelerating lithium battery R&D.

This past March 17, BYD unveiled its next-generation Blade Battery featuring "Megawatt Flash Charging" technology. According to BYD, this battery achieves a 50% reduction in internal resistance, supports charging currents of 1,000A, and delivers a charging rate of 10C—all global firsts.

Under the combined power of ultra-high voltage (1,000V) and ultra-high current (1,000A), the flash-charging battery delivers the world's highest mass-production charging power of 1 megawatt (1,000kW), officially ushering in the "megawatt era" of charging and achieving peak charging speeds of "2 kilometers per second"—another global mass-production record.

BYD uses Seed BAMBOO AI to Screen Electrolyte Formulations Quickly

Behind this megawatt flash-charging breakthrough lies the technical contribution of ByteDance Seed.Sources indicate that BYD's development team integrated ByteDance Seed's BAMBOO electrolyte AI modeling framework from the project's inception, enabling rapid screening of electrolyte formulations. Compared to traditional trial-and-error approaches, BAMBOO leverages experimental data to accurately predict critical electrolyte properties including density, viscosity, and ionic conductivity. During the megawatt flash-charging battery development process, it significantly narrowed the experimental formulation space and accelerated the R&D timeline.

According to information released by ByteDance Seed, the BAMBOO framework comprises two core components: the BAMBOO-MLFF machine learning force field and the BAMBOO-Mix formulation property prediction and generation model.

BAMBOO-MLFF represents a machine learning force field designed specifically for liquid-phase organic small molecules, with electrolytes serving as the primary application. Its architecture integrates molecular dynamics physical algorithms with AI model network design and fine-tuning techniques.

source: ByteDance Seed

The BAMBOO-Mix formulation property prediction and generation model addresses the broad screening phase, designed to overcome the time-intensive nature of molecular dynamics methods and the vast chemical formulation search space. This model can directly generate potential formulations based on constraint properties or rapidly predict properties for given formulations, supporting both forward prediction and reverse formulation generation.

source: ByteDance Seed WeChat Platform 

According to the announcement, ByteDance Seed and BYD's lithium battery division will further deepen their collaboration through the establishment of an "AI + High-Throughput Joint Laboratory," targeting breakthrough solutions for power battery fast-charging, longevity, and safety challenges. The partnership will also explore expanded AI applications in battery R&D, accelerating the discovery of new materials and formulations while driving advances in power battery technology.