ECL today announced that CSC1, its 35MW FlexGrid-powered data center in Santa Clara, Calif., is fully committed across three tenants, with deployment beginning in Q1 2027 and completion expected by Q3/Q4 2027. The three tenants, including established AI cloud platforms and inference-focused startups backed by leading AI research organizations, are deploying AI inference workloads at CSC1. The company has simultaneously opened three additional sites, CSJ1, CSU1, and CSU2, for tenant inquiries.
CSC1 reached full commitment just eight months after its April 2026 announcement and before construction is complete. This is because ECL’s modular block architecture and behind-the-meter power generation deliver a purchase-order-to-keys timeline of nine to 12 months in metro markets, compared to the two to three years typical of conventional grid-dependent data center development. This new development comes at a critical time. 451 Research currently projects that US data center electricity demand will almost triple in the next five years, reaching about 183.2 GW by the end of 20301, while Bloomberg reports that half of all US data center projects this year face delays or cancellations due to power constraints2.
“Every major AI operator in the Bay Area is doing the same math right now, weighing years-long interconnection queues against inference demand that compounds monthly,” said Yuval Bachar, co-founder and CEO of ECL. “These three tenants chose CSC1 because FlexGrid gets them to production in months, not years. They are deploying next quarter instead of waiting for a utility to approve a substation. CSC1 is the first in a distributed network of ECL inference-scale facilities across the Bay Area, because inference has to be located close to people, data, and applications, often in and around major cities where 50 or 100 spare megawatts rarely exist on the grid.”
CSC1 supports rack densities up to 150kW at launch, with an architecture designed to accommodate future densities up to 250kW and beyond as tenant requirements evolve. The facility is ready for NVIDIA GB300 and Vera Rubin architectures, with direct-to-chip liquid cooling and a sub-1.15 PUE. FlexGrid combines up to 5MW of on-grid electricity, up to 30MW of natural gas generation through microturbines and fuel cells, and up to 5MW of hydrogen power, with battery systems providing transient control and backup. The power mix can shift over time based on operational needs and energy market conditions, and the facility’s architecture requires no redesign when fuel sources change.
ECL’s modular block architecture eliminates several supply chain bottlenecks that slow conventional builds. The design sources many smaller components rather than a few large ones, such as utility-scale transformers, which now carry lead times exceeding 120 weeks, enabling ECL to use a completely different supply chain than traditional data center operators. It builds blocks in parallel and omits conventional UPS systems, further compressing the construction cycle. Tenants can begin running production workloads in early blocks while later phases are still under construction.
When hydrogen is used as a FlexGrid power source, byproduct water from fuel cell generation supplies the cooling loop, eliminating the freshwater draw that increasingly concerns data center operators in drought-prone California. The facility’s cooling architecture combines direct-to-chip, air, and hydrogen byproduct water-based systems, and is designed for Tier III-level reliability.
| 1. | |
| 2. |
Availability
CSJ1, CSU1, and CSU2, the next sites in ECL’s planned distributed inference network across the Bay Area, will follow the same FlexGrid architecture and are designed for inference-dense AI workloads at 25-35MW scale. Prospective tenants can contact ECL directly through www.ecldc.com for capacity planning and early site engagement.
About ECL
ECL is a next-generation data center company focused on building high-density, AI-ready facilities that can be deployed where traditional power infrastructure falls short. ECL’s FlexGrid microgrid architecture and ECL Lightning real-time management platform combine grid power, natural gas, batteries and hydrogen fuel cells to deliver resilient, efficient power tailored to AI and inference workloads. ECL introduced the world’s first operational hydrogen-powered AI data center at its MV-1 site in Mountain View, California, in 2024 and is scaling its FlexGrid architecture across U.S. and international markets. Backed by Molex Ventures and Hyperwise Ventures, ECL enables scalable, low-carbon compute for AI, high-performance and edge workloads. For more information, visit www.ecldc.com.
View source version on businesswire.com: https://www.businesswire.com/news/home/20261103121494/en/
Media gallery
