The industry’s first hydrogen-powered NVIDIA GB300 NVL72 systems have arrived at ECL’s Mountain View facility. Lambda, a cloud computing provider, has deployed this technology. The systems are built by Supermicro (NASDAQ:SMCI), and in fact, a fuel cell used to convert the explosive gas into electrons generates water as a byproduct, which, according to ECL, the operator of the bit repository that houses the group, is recycled for facility cooling. In other words, quite a technological breakthrough.
“As we move toward gigawatt AI factories, diversified energy is becoming essential infrastructure”
Lambda has doubled its presence at ECL’s Mountain View campus, now occupying 100% of the facility, which operates as an off-grid, water-free, and emissions-free modular data center powered entirely by hydrogen fuel cells. The aforementioned NASDAQ:SMCI is a prominent player in the Technology Hardware industry with a market cap of $27.85 billion and impressive revenue growth of 46.59% over the past twelve months, according to InvestingPro data. “As we move toward gigawatt AI factories, diversified energy is becoming essential infrastructure,” said Ken Patchett, Vice President of Data Center Infrastructure at Lambda.
The system features approximately 20 TB of HBM3e memory and dense FP4 performance of one exaFLOP
The deployment addresses a significant challenge in the industry, as few data centers can handle the power density and cooling requirements of these advanced systems. These dense, rack-scale systems incorporate 72 of Nvidia’s most powerful Blackwell Ultra accelerators. In total, the system features approximately 20 TB of HBM3e memory and dense FP4 performance of one exaFLOP. The consequence of such a dense system is power consumption. The 4,000-pound systems were fully integrated into the data center in two hours, according to the company.
Considering that a fully equipped GB300 Superpod, requires more than a megawatt of power, we can assume it is less
Yuval Bachar, founder and CEO of ECL, said the collaboration “demonstrates that high-performance, off-grid, zero-emissions data centers are not just aspirational, but operational, at scale.” Lambda did not reveal how many of these systems it is deploying at the facility. However, considering that a fully equipped GB300 Superpod, which includes eight NVL72 rack systems, requires more than a megawatt of power, we can assume it is less.
Occupying space in the cloud, which, as we know, involves the effort of supercomputers that, in turn, require electrical resources
This news is interesting and comes at a time when AI and technology are experimenting in entirely new fields. Not only that, but it would also be interesting to explain how they intend to protect the planet with this super cloud that can store so many thousands and millions of pieces of data. Lambda, founded in 2012, describes itself as “The Superintelligence Cloud” that builds infrastructure for AI training and inference. The company continues to explore hydrogen energy as part of its commitment to developing what it calls “gigawatt-scale AI factories.” That is, occupying space in the cloud, which, as we know, involves the effort of supercomputers that, in turn, require electrical resources to function properly.
Regarding the latter, according to the companies themselves, hydrogen will be piped to the facility through three separate pipelines that converge on the site and will reportedly provide enough capacity for up to 2 GW of computing in the future. It remains to be seen what proportion of this fuel will come from sustainable sources, which is one of the main challenges these facilities may face in the future.
