Silicon Valley loves to talk about saving the planet—usually while plugging into a power grid that’s straining under the weight of AI. Now a company called ECL says it’s built a data center that cuts the cord entirely and runs primarily on hydrogen.
The project, dubbed MV1, is up and running in Mountain View, California. And ECL just pulled in another $10 million in funding led by Hyperwise Ventures to speed up R&D and push the concept beyond one shiny pilot site.
An AI data center built for brute-force computing
ECL is aiming straight at the hottest—and most power-hungry—corner of tech: AI workloads. Training and running models means packing in GPUs, pulling massive electricity, and dumping heat like a furnace. ECL’s pitch is that its facilities are designed from the ground up for that reality, not retrofitted after the fact.
Europe’s 100% Hydrogen Gas Turbine Is Here—and It’s a 560-Megawatt Statement
The company says its modular, “autonomous” sites can operate off-grid and hit a power usage effectiveness (PUE) of 1.1—an efficiency metric where closer to 1.0 is better. It also claims it can support high-density deployments up to 75 kW per rack, the kind of number that makes old-school enterprise server rooms look like museum pieces.
What ECL is selling: speed, density, and hydrogen
Here’s the checklist ECL is putting in front of customers:
Hydrogen as primary power: The company says hydrogen lets it avoid carbon emissions tied to conventional generation. (The big asterisk, of course, is where the hydrogen comes from—“green” hydrogen isn’t the same thing as hydrogen produced from natural gas.)
Dubai Built a $4.3B Solar “Desert” the Size of a Small City—and It Actually Works at Night
Scale in 1-megawatt chunks: ECL says customers can expand capacity in 1 MW increments, basically treating data center growth like snapping on Lego bricks.
Faster build timeline: ECL claims it can design and deliver in under 12 months, compared with the industry’s typical two to three years. If that holds up at scale, it’s a real competitive weapon—AI demand isn’t waiting around for zoning meetings and transformer backlogs.
Cooling tricks: ECL says it uses water produced during hydrogen power generation plus “backdoor heat exchanger” tech to cool high-density racks, reducing dependence on local resources. Translation: less begging the local utility (or municipality) for more water and power.
Spain’s new hydrogen coach claims 620 miles of range—and a 20‑minute refill
Real-time management: Its in-house system, ECL Lightning™, is pitched as a monitoring-and-control layer tuned to optimize AI performance.
The first customer: Cato Digital, betting on dense NVIDIA GPU gear
ECL’s first tenant is Cato Digital, a sustainability-focused cloud provider led by data center veteran Dean Nelson. Cato Digital is offering dedicated GPU servers to platform partners, enterprises, and retailers—hardware meant for inference engines, generative AI, and continuous training of small language models (SLMs).
The company is using NVIDIA DGX systems configured at maximum density to take advantage of MV1’s design. In plain English: cram in as much GPU horsepower as possible, then see if the building can keep it powered and cool without blinking.
Who’s behind ECL—and why hydrogen now?
ECL was founded in 2021 and is led by Yuval Bachar, an engineer with eight U.S. patents and the founder of Open19, a project that pushed an open standard for servers. The company says its team includes data center veterans from places Americans actually recognize: Facebook, Microsoft, LinkedIn, Cisco, HPE, and Bloom Energy.
Bachar’s argument is straightforward: plenty of data center operators already see hydrogen as a future power source because of its safety, efficiency, reliability, and sustainability potential. The problem is money—retrofitting existing facilities is expensive enough that many operators can’t justify the leap.
ECL’s workaround is to build hydrogen-first sites designed for the AI era’s space, power, and cooling demands—rather than trying to bolt hydrogen onto yesterday’s infrastructure.
The catch: hydrogen hype meets real-world economics
ECL is right about one thing: AI is turning electricity into destiny. But hydrogen isn’t magic fairy dust. The climate math depends heavily on how the hydrogen is produced, and the cost curve still scares off a lot of buyers.
Still, if ECL can actually deliver high-density AI capacity in under a year, off-grid, with a PUE around 1.1, it’s going to get meetings—especially from companies tired of waiting in line for more utility power.
