Xeal is placing Nvidia computer racks inside metal shipping containers and deploying them near its charging stations. A novel approach to powering AI computing is emerging: using electric vehicle charging stations. This is exactly the concept being pursued by charging station startup Xeal, which plans to raise approximately $2 billion in debt financing over the next 18 months to support the project's rollout.

New York-based Xeal operates a network of thousands of charging stations, mostly located at upscale apartment buildings, high-end retail malls, and office towers across roughly 500 cities in the United States and Canada. Xeal CEO Nikhil Bharadwaj said that although contracts with utility companies allow the charging stations to charge EVs around the clock, the average utilization rate of its charging stations is less than 10%. This means Xeal holds a massive amount of idle power capacity — enough to continuously supply electricity to roughly 130,000 households or power 110,000 Nvidia AI chips.

To that end, the company has launched a new business: supplying idle power to AI chips. In December, Xeal will deploy Nvidia AI server racks next to charging stations in parking lots in Chicago, Houston, and Dallas. The racks will be placed inside custom protective metal containers roughly the size of a single parking space. The power lines connected to the charging stations can directly supply electricity to the racks, with each container housing approximately 48 Nvidia AI chips.

Bharadwaj said Xeal has already secured compute customers for 100 of these compute containers but declined to disclose the customers' names. He said that over the next 18 months, the company aims to raise approximately $2 billion to manufacture and deploy 1,000 containers. Each container holds one set of GPU racks, collectively providing approximately 50,000 GPUs with a total power capacity of 60 megawatts.

Xeal previously completed two rounds of equity financing, raising a cumulative $54 million, with the most recent being a $40 million Series B round in 2022. Bharadwaj said Xeal plans to apply for low-interest asset-backed loans and project financing totaling $2 billion to support its business expansion. He believes the company qualifies for this funding: it built an extensive charging station network using only $54 million in venture capital.

Xeal's plan reflects how even companies far removed from the AI industry are trying every possible way to capture the massive power demand dividend created by the AI data center boom. However, as colleagues at this publication have reported, financing is becoming increasingly difficult for some AI project developers. Even if the first batch of GPU containers operates smoothly, Xeal may not secure its target financing as hoped.

Xeal's core advantage lies in its existing grid access rights. AI data center developers often wait years to complete grid connections. Many companies have turned to considering gas turbines or nuclear power, but such solutions also face lengthy approval and construction timelines. Batteries can only serve as backup power and cannot sustain a full data center's continuous operation over extended periods — they need recharging after just a few hours.

Bharadwaj said: "We only need a few weeks to bring inference and compute services online, rather than waiting years. Capital is the accelerant that helps us build a sustainable compute infrastructure."

Bharadwaj co-founded Xeal in 2019. The company sells charging stations to property owners; if EV usage frequency is high, owners typically recoup their investment within five years. But due to various factors — including the cost of purchasing vehicles — U.S. EV sales have fallen short of expectations, resulting in an average idle rate of 92% for Xeal's charging stations. Many property owners are therefore reluctant to add more charging stations.

Bharadwaj said: "There has always been a chicken-and-egg dilemma here. Real estate clients would say: I don't know if there are EV owners in the building, and I'm not sure whether to install more charging stations."

The AI computing boom has changed the equation. Bharadwaj contacts building owners who use Xeal charging stations, pays them site rent, and obtains permission to place a GPU container in the garage. He tells property owners: "What used to take five years to pay back now takes only two and a half years; you don't need capital expenditure and don't bear operating costs — we cover all electricity expenses." He said compute demand is extremely strong, sufficient to support investment in additional charging stations.

Xeal's existing sites have a total power capacity of 200 megawatts and are expanding at a rate of 10 megawatts per quarter. "Now we can advance charging station projects across the United States without federal subsidies or utility incentive programs. Leveraging already-built infrastructure and generating revenue from the compute business, the entire industry can become self-sustaining."