Colovore Building New Data Center for Extreme Density AI Workloads

Nov. 3, 2022
High-density hosting specialist Colovore is expanding in Santa Clara with a 9-megawatt data center supporting AI workloads with rack densities from 50 kW up to 250 kW.

High-density cooling specialist Colovore will double its footprint with a new data center in Silicon Valley’s largest cloud hub.

Colovore will create a 9-megawatt facility in Santa Clara that will be optimized to host liquid-cooled artificial intelligence (AI) workloads for colocation customers. The new data center will enable tenants to deploy racks with power densities of 50kW using water-chilled rear-door cooling units, with the ability to push densities up to 250 kW per rack using direct liquid cooling (DLC).

As a specialist in high-density hosting, Colovore has served customers running the most power-intensive data center applications, especially AI workloads running on NVIDIA GPUs. As more companies adopt AI, power densities are rising beyond what many traditional data centers can support.

“In the last 3 years, it’s really become more of a tidal wave of power density, driven fundamentally by incredible innovations at the chip and processing layers,” said Ben Coughlin, co-founder and Chief Financial Officer of Colovore. “It is really driving densities through the roof.”

“The AI revolution is here and simply put, these servers demand purpose-built, liquid-cooled, high-density colocation environments,” said Sean Holzknecht, President and Co-Founder of Colovore.

New Capacity in a Supply-Constrained Market

Colovore operates in the heart of Santa Clara’s data center district. Its original data center is on Space Park Drive, next to a Digital Realty campus that includes the region’s primary connectivity hub. Neighbors include Equinix and Cyxtera.

The new facility will be next door at 3060 Raymond Street, enabling Colovore to create a single campus. Colovore will retrofit an existing warehouse at the site and expects to bring the facility online in early 2024.

“When we opened our doors in 2013, touting 20 kW per cabinet in our Phase 1, many thought ‘power density’ wasn’t a big issue at the time,” said Holzknecht. “We now support thousands of AI and GPU systems for Fortune 500 companies down to Silicon Valley startups, in hundreds of cabinets each drawing 15 – 50 kW per cab. The bottom line is that this data, and all these incredible AI and Big Data platforms, are really thirsty.”

The Colovore project will bring new capacity into a supply-constrained Santa Clara market. Data center broker CBRE says data center capacity in Silicon Valley is at the lowest supply in its history, with vacancy rates an extraordinary 1.3%.

Santa Clara has long been the Data Center Capital of Silicon Valley due to competitive power pricing from the municipal utility, Silicon Valley Power. It’s the favored spot to deploy new hardware and services from the Valley’s marquee technology companies, as well as a legion of fast-moving startups. But Santa Clara has almost no open land parcels available, meaning developers must acquire sites with existing structures and knock them down to make way for a purpose-built data center. 

By retrofitting an existing facility, Colovore expects to have a shorter path to market than new projects involving demolition and multi-story projects. The project has already received approval for power delivery from Silicon Valley Power and will include seismic reinforcement and other upgrades in its retrofit.

The building is about 30,000 square feet, but Colovore’s design allows it to maximize its real estate by putting more compute capacity into every square foot.

Hot New AI Hardware Drives Densities Higher

As recently as 2020, rack power densities in enterprise data centers averaged 5 to 9 kW, with hyperscale operators at about 8 to 10 kW. Those numbers have begun trending higher as more companies make AI a part of their strategy, making products and services smarter.

Coughlin says the densities will continue to rise as new, more powerful chips enter the market. The latest hardware from NVIDIA and Cerebras averaging more than 1 kW per rack unit. Flash storage has also become more power-intensive, adding to the density challenge.

“In Silicon Valley, we have a seat in the front row in a market that’s advanced,” said Coughlin. “We’re seeing this becoming more mainstream,” said Coughlin. “The technology has delivered, the hardware is getting cheaper, and the applications are getting easier to use. Companies are now making investments in AI.”

Colovore’s design allows it to adapt. It employs a raised floor with pipes underneath, with a water loop running to a heat exchanger outside the building. The chilled-water rear cooling door is its standard design, but Colovore is already running several megawatts of customer workloads using direct liquid cooling. There are a number of approaches to DLC, and the company says it can support them all.

“Server manufacturers are considering a lot of different ways to do it,” said Coughlin. “One of our advantages is that we have the piping below the floor already.

“Our assumption is that this growth (of high-density workloads) is going to continue, and we are well-positioned to deliver.”

Colovore is one of a small number of high-density specialists, along with ScaleMatrix, which is based in San Diego and supports that region’s active ecosystem of genomics research firms. CirraScale offers cloud services for extreme-density apps, and has a presence in both the ScaleMatrix and Colovore facilities.

About the Author

Rich Miller

I write about the places where the Internet lives, telling the story of data centers and the people who build them. I founded Data Center Knowledge, the data center industry's leading news site. Now I'm exploring the future of cloud computing at Data Center Frontier.

Sponsored Recommendations

Guide to Environmental Sustainability Metrics for Data Centers

Unlock the power of Environmental, Social, and Governance (ESG) reporting in the data center industry with our comprehensive guide, proposing 28 key metrics across five categories...

The AI Disruption: Challenges and Guidance for Data Center Design

From large training clusters to small edge inference servers, AI is becoming a larger percentage of data center workloads. Learn more.

A better approach to boost data center capacity – Supply capacity agreements

Explore a transformative approach to data center capacity planning with insights on supply capacity agreements, addressing the impact of COVID-19, the AI race, and the evolving...

How Modernizing Aging Data Center Infrastructure Improves Sustainability

Explore the path to improved sustainability in data centers by modernizing aging infrastructure, uncovering challenges, three effective approaches, and specific examples outlined...

SeventyFour / Shutterstock.com

Improve Data Center Efficiency with Advanced Monitoring and Calculated Points

Max Hamner, Research and Development Engineer at Modius, explains how using calculated points adds up to a superior experience for the DCIM user.

White Papers

Get the full report

Achieving Energy Efficiency Goals in Data Centers

April 15, 2022
One of the challenges global data centers currently face is the need to meet the increased processing and storage needs of their customers while also making their operations more...