Facilities sitting on excess energy capacity can monetize that resource by hosting on-site edge computing modules that essentially make the buildings part of a distributed server system for power-hungry data centers, according to Christopher Turner, senior vice president, business development and policy, at clean energy technology company Arcadia.
Earlier this year the company began working with facilities to assess the amount of energy they use relative to the capacity that their local utilities have committed to them for their operations.
If they have headroom available, as Turner describes it, and they have stable access to fiber connectivity and can meet other infrastructure requirements, they can use the power to host the modules in return for rental fees.
“It can unlock attractive revenue opportunities,” Turner told Facilities Dive in an interview.
Inside the modules are graphics processing units that function as part of the distributed computing system. Depending on the scale of the site and the amount of power available, a facility might host mini-split-sized modules that consume about 20 kilowatts of power or shipping-container-sized units that consume a megawatt worth of power, he said. Arcadia reimburses the facility for its electricity costs for powering the modules at the local utility’s prevailing rate and adds the rental fee on top of that.
For facilities that commit to hosting modules for a multiyear term, the arrangement gives them a steady recurring revenue stream that they can leverage for value-added initiatives, like energy efficiency upgrades or decarbonization efforts, Turner said.
“That [annual recurring revenue] is meaningful,” he said. “It’s a regular inflow, so [if] you have an energy manager or a facility manager who says, ‘I don’t know how I would [pay for] XYZ energy-efficiency measure with this tight budget,’ this is a way they can go back and say this [initiative] is actually self-funding; we’re going to use this to deploy energy efficiency [measures] and create more headroom for us.”
Organizations that have their own computing needs can buy the modules and use them to generate their own output, he said.
“If I have a business that wants to … generate their own token, they would cover the cost of that,” he said. Token refers to the computing output they generate.
For those that go the ownership route, Arcadia would offer its expertise to help them procure energy in the most cost-effective way. As an energy adviser and procurement shop for building owners, it uses its visibility into energy rates to help its customers lower their cost of energy.
“We would be able to say, do you want us to also understand the impact on your bill on-site by adding this?” Turner said. “Do you also want us to procure energy from retailers in deregulated markets or optimize your rate schedule to support your ownership and make it as cost-effective as possible for you to own and operate your own modules?”
One early adopter of the distributed AI program is using an under-leased office building to host the modules, he said.
“We’re trying to find ways to maximize the property value and support the property owners with office space that’s been challenging to lease,” he said.
Another organization that’s invested in EV charging stations at multiple locations has excess capacity that it plans to monetize through the program, Turner said. “They have decent utilization on-site for a few of these [charging stations], but they’re not getting the full potential, so we’re using some of that … electrical infrastructure,” said Turner, who described the revenue from the modules as a way to “de-risk” the investment that organization made in the infrastructure.
The revenue opportunity is available to any facility that has excess capacity and otherwise has an infrastructure that can support the modules, but the fee can vary based on where the facilities are located, where they get their power and other factors. Facilities closest to the biggest computing demand would benefit the most, he said.
“Getting the most lucrative rent payment or host payment will be [a facility] in one of the main metropolitan areas where you’re seeing a lot of AI adopted, a lot of need for AI compute,” he said. “Larger sites [in those areas] are easy to service, easy to create that connectedness. So, we’re trying to aim for those Tier 1 metros, [but] anybody who has a deeply remote operation” could benefit as well.
