AI cloud provider Lambda has deployed Nvidia GB300 NVL72 systems at EdgeCloudLink's (ECL) data center in Mountain View, California.

The ECL data center campus is fully powered by hydrogen, and Lambda claims its deployment is "zero water," cooled using direct-to-chip liquid systems fed by centralized CDUs that recycle water generated as a byproduct of power production.

The GB300 NVL72 system was built by Supermicro, with each rack receiving 142kW of compute power. It features 72 Nvidia Blackwell Ultra GPUs and 36 Arm-based CPUs, NVLink bandwidth of 130TBps, and up to 40TB of fast memory. Compared to the Blackwell GPUs, Blackwell Ultra GPUs have 2x the attention-layer acceleration and 1.5x more AI compute floating-point operations per second (FLOPS). At FP64, the whole GB300 NVL72 rack offers 100 teraflops of compute.

"As we move toward gigawatt AI factories, diversified power is becoming essential infrastructure," said Ken Patchett, VP data center infrastructure, Lambda. "These Nvidia GB300 NVL72 systems represent the building blocks for training and deploying tomorrow's foundation models. Hydrogen-based energy ensures we can power that superintelligence-class compute responsibly."

According to Lambda, the racks were deployed in just two hours. The AI cloud company has increased its capacity at the ECL campus from 50 percent of the facility to 100 percent of its capacity. The data center has 1MW of IT capacity and was launched in May 2024. The data center is powered entirely by a mix of grey, blue, and green hydrogen. Lambda is also set to be the anchor tenant at ECL's upcoming 1GW TerraSite-TX1 data center campus in Houston, Texas, which will also be powered by hydrogen.

“Supermicro is proud to deliver to Lambda the latest Nvidia GB300 NVL72 systems powered by hydrogen-based energy,” said Vik Malyala, SVP of Technology & AI at Supermicro. “By combining Supermicro’s advanced liquid-cooled servers with ECL’s zero-emission power, we’re enabling Lambda's SuperIntelligence Cloud to scale rapidly and sustainably.”

“ECL's work with Lambda sets a new bar for sustainable AI factory power and proves that off-grid, zero-emission, high-performance data centers are not just aspirational, but operational, at scale,” added Yuval Bachar, Founder and CEO of ECL. “Hydrogen power and advanced cooling let AI organizations stay on the cutting edge of AI infrastructure while meeting the highest standards for energy stewardship and flexibility.”

Lambda hasn't shared how many of the racks have been deployed.

The deployment of the latest Nvidia chips follows Lambda's installation of the Nvidia GB200 NVL72 at ECL's data center in January of this year. At the time, ECL's Bachar noted that the facility could offer up to 150kW per rack with "just a simple modification."

Founded in 2012, Lambda rents out colocation space in data centers in San Francisco, California, and Allen, Texas, and offers comparatively low-cost GPU-based cloud compute. Alongside the cloud, it provides colocation space and sells GPU desktops.

Earlier this month, Lambda secured a $1.5bn deal from Nvidia for the chipmaker to lease back its own GPUs. The AI cloud company is reportedly preparing for an initial public offering.

We previously spoke to Yuval Bachar about how the company is powering its data center with hydrogen. Read more here.