Hydrogen Production Powers First Off-Grid NVIDIA GB300 NVL72 AI Systems in Zero-Emission Data Center

On September 24, 2025, Lambda teamed up with ECL and Supermicro to fire up the very first production-grade NVIDIA GB300 NVL72 AI systems at ECL’s off-grid “MV1” campus in Mountain View, California. Running entirely on hydrogen fuel cells, this modular data center isn’t just a high-performance powerhouse—it’s a showcase for true zero-emission technology and zero external water use. In a world where massive foundation models push traditional power and cooling to the brink, MV1 proves sustainable AI compute isn’t a pipe dream—it’s here, and it works.

Reimagining Hydrogen-Powered AI Compute

By tapping into hydrogen production via fuel cells, ECL’s MV1 stays completely off the grid and out of the municipal water loop. Even better, the cooling circuit gets a second life with the very water produced during power generation—captured, purified, and recycled back into the system. ECL claims this could net a negative local water footprint (pending independent validation). That’s the kind of real-world impact that makes green hydrogen and cutting-edge cooling technologies more than just buzzwords.

• First-ever production-grade NVIDIA GB300 NVL72 AI systems powered entirely by green hydrogen
• Each Supermicro cabinet pulls 142 kW of compute muscle
• Plug-and-play rack integration in just two hours
• No diesel generators humming in the background
• Zero external water draw thanks to recycled byproduct water

Historical Context

Back in the early 2010s, modular data centers started appearing as a nimble, cost-effective alternative to traditional builds. At the same time, hydrogen fuel cells were mainly relegated to telecom backup power—brief stints, nothing more. Then, in mid-2023, ECL demoed a 5 MW green hydrogen module in Mountain View that ran 24/7 without touching the grid. MV1 takes that proof-of-concept a giant leap forward by pairing hydrogen energy with high-density GPU compute for the first time at production scale.

MV1 isn’t just flexing compute power; it’s rewriting the playbook for data centers in remote or resource-constrained sites. By ditching diesel generators and mechanical chillers in favor of fuel cells and closed-loop liquid cooling, it tackles both carbon emissions and water usage head-on. With regulators and customers demanding greener setups, this kind of forward-looking design shows how hydrogen infrastructure can shatter old limits on power density.

Technical Under the Hood

Inside each cabinet, you’ll find 72 NVIDIA GB300 GPUs talking to one another over NVLink for blistering bandwidth and rock-solid low latency. A direct-to-chip liquid cooling system—managed by centralized distribution units (CDUs)—keeps temperatures in check. Here’s the kicker: that coolant is actually the water byproduct from converting hydrogen into electricity. Talk about a closed loop for both power and thermal management.

Each NVL72 rack is built with eight liquid-cooled trays, nine GPUs apiece, all tuned for distributed training frameworks like PyTorch and Horovod. The system even recycles waste heat from the fuel cell stack, pushing PUE below 1.05—on par with the most efficient hyperscalers.

• Hydrogen fuel cells convert H₂ into electricity, with pure water as the only byproduct
• CDUs route that water through microchannel cold plates attached directly to GPU dies
• Superior heat transfer cuts overall cooling power, boosting efficiency
• Racks achieve full operation within two hours, slashing deployment time

Strategic Implications

This landmark rollout underscores Lambda’s vision of gigawatt-scale “AI factories.” By doubling its footprint at ECL’s Mountain View site, Lambda is staking its claim on hydrogen’s future in AI. Going off-grid insulates them from volatile electricity prices and brownout risks—huge benefits as foundation models demand ever more compute. And that two-hour rack integration? It means AI startups and research labs can scale on demand without a massive installation bill. With power costs climbing in tech hubs like Silicon Valley, MV1’s hydrogen-based economics are drawing serious interest from sustainability-focused investors. Up next: TerraSite-TX1, a multi-gigawatt campus near Houston where Lambda plans another hydrogen-fueled showcase. As ESG mandates tighten, MV1’s blueprint could become the gold standard for clean, high-performance compute.

Company Context

Founded in 2012 by a group of hands-on AI engineers, Lambda has become a go-to for scalable GPU cloud compute and colocation—its “Superintelligence Cloud.” ECL, led by CEO Yuval Bachar and backed by Hyperwise Ventures and Molex, specializes in off-grid, modular data centers fueled by green hydrogen. Supermicro, headquartered in San Jose, builds the liquid-cooled racks that house NVIDIA’s powerful GPUs. Together, they’ve woven compute, energy, and hardware expertise into a seamless solution, letting clients tap into dedicated green compute without building their own mini power plants—shortening time to market from years to months.

Environmental and Market Impact

Environmental: MV1’s zero-emission profile wipes out Scope 2 emissions, and by returning purified byproduct water to local aquifers (pending audit), it may even help recharge groundwater.

Industrial: Demonstrates that off-grid, hydrogen-based data centers can rival grid-tied facilities, cutting the need for new substations and transmission upgrades.

Market: Spurs investment in hydrogen infrastructure and advanced cooling, opening fresh funding channels for green AI ventures.

Technical: Validates that liquid-cooled GPU clusters can run reliably on modular hydrogen power plants, paving the way for deployments in remote or resource-scarce environments.

Societal: Offers policymakers a living example of how compute-intensive industries can hit aggressive decarbonization targets, bolstering support for hydrogen infrastructure programs.

Looking Ahead

MV1 is living proof that hydrogen-powered AI compute works—and the real test comes when TerraSite-TX1 rolls out in Houston. If this momentum holds, we could see gigawatt-scale, hydrogen-fueled AI campuses all over North America by late 2027. Advances in electrolysis—whether tied to renewables or next-gen feedstocks—will be key to scaling these zero-emission technology hubs. As hydrogen fuel cells continue to mature, projects like MV1 will set the bar for decarbonizing high-performance compute at scale.

About ECL
Founded by veterans from Facebook, Microsoft, LinkedIn, Cisco, HPE, and Bloom Energy, and led by CEO Yuval Bachar, ECL specializes in off-grid, modular, hydrogen-powered data centers. Backed by Molex and Hyperwise Ventures, the MV1 pilot in Mountain View is ECL’s first full-scale proof-of-concept for zero-emission AI compute facilities.