Crusoe Energy Systems is collaborating with Vast Data to offer high-performance storage for artificial intelligence (AI) to Crusoe Cloud customers.
Crusoe is also teaming up with SES AI Corp. to work on material discovery using Crusoe's AI cloud.
Crusoe goes Vast
Crusoe and Vast have collaborated to develop "Shared Disks," a petabyte-scale file system capable of reads up to 200Mbps per TiB per node with hundreds of GBps in aggregate read and write bandwidth per cluster.
The Shared Disks also offer a 99.5 percent SLA that is offered by Crusoe Cloud GPU VMs with data being protected both from hardware and component failures.
In addition, Crusoe Cloud customers will have access to the Vast Data Platform and Vast Data Store.
"Powered by the Vast Data Platform, Crusoe's Shared Disks offering delivers the modern AI cloud infrastructure today's enterprises need to address the challenges of scaling data-intensive AI workloads," said Chris Morgan, vice president of solutions at Vast Data. "Together, we're delivering AI-driven solutions that provide the speed, security, and operational efficiency required for organizations looking to transform their data landscapes and accelerate their AI deployments to drive innovation and discovery."
Patrick McGregor, Crusoe's chief product officer, added: "Crusoe chose the Vast Data Platform because of its exceptional ability to deliver the reliable file storage that our customers need without any depletion of performance as AI models are scaled."
Crusoe Cloud was launched in 2022 for AI training, inference, and high-performance computing (HPC) workloads. The platform is powered by clear, stranded, and renewable energy by colocating data centers with energy resources. The company was previously known for mining cryptocurrencies via flared gas but pivoted to AI data centers in July 2024.
Crusoe works with SES AI on batteries
The company is also working with SES AI Corporation, a company that develops and manufactures li-metal batteries.
The two are collaborating to "accelerate material discovery." Using Crusoe's AI cloud platform, the companies are mapping small molecules using a supercomputer optimized for AI with the aim of better-understanding battery chemistry, improving energy storage solutions, and energy-efficient technology.
According to the companies, the battery industry has thus far studied 1,000 organic molecules and achieved a coulombic efficiency of over 99.6 percent, but there are trillions of organic molecules under 20 and 30 atoms that could be studied.
The compute resources will be used to map these potential molecules that could be used for future energy technologies, as well as to analyze SES AI's manufacturing data, and help to predict battery incidents.
SES AI primarily develops lithium batteries for use in electric vehicles and urban air mobility.
The models will use Crusoe Cloud's Nvidia H100 virtualized instances. Eventually, this will involve the creation of a multimodal large language model and AI agent, training a 70-billion-parameter Llama 3 model on approximately 200 billion tokens from various battery-related texts and literature.
Once molecules are found that show promise, they will undergo laboratory testing in SES AI's Electrolyte Foundry.
“With access to more Nvidia GPUs, we expect to map a large enough molecular universe that our AI model training will reach sufficient accuracy. Once we have this map, we believe we can accelerate material discovery for any battery problem. This includes not just Li-Metal for EVs and UAMs, but also Li-ion batteries for consumer electronics, grid storage, automotive, and other applications,” said Qichao Hu, CEO of SES AI.
Chase Lochmiller, CEO of Crusoe, added: “Crusoe is dedicated to aligning the future of computing with the future of the climate. By providing a sustainable and cost-effective AI cloud platform, we are supporting SES AI’s vision to map the molecular universe and create ground-breaking AI models that could accelerate a sustainable future. This collaboration showcases how advanced computing can drive scientific progress while minimizing environmental impact.”
Earlier this year, Microsoft teamed up with the Pacific Northwest National Laboratory (PNNL) to find new materials for batteries using Microsoft's Azure Quantum Elements. Using AQE, the researchers at PNNL looked at 32 million inorganic materials to arrive at 18 candidates for their battery project. AI models first cut the number down to around 500,000, and HPC techniques dealt with the remaining 499,988 options. One of those 18 is now being tested.