Hyperscalers, such as Google, Meta, and Amazon, are scrambling to build data center capacity to keep up with AI’s growing demand for computing power. But gaining access to enough reliable electricity is a significant hurdle. Hence, the term speed to power now permeates the industry’s lexicon and has become the metric by which all data center development is measured.
Speed to power refers to the rate at which new or expanding data centers can secure electric power, considering the commercial, technical, regulatory, and economic factors required to obtain a data center’s permits and utility interconnection approval, build the project, and finally receive permission to operate from the utility. In their quest for speed to power, hyperscalers are getting a crash course on the electric utility industry, its varied regulatory red tape and institutional quirks, and its hallmark slow and measured pace.
Achieving speed to power at scale
A recent estimate from Deloitte projects that power demand from U.S. data centers will reach 176 GW by 2035, which is five times what data centers use today. Providing this power is a Herculean effort for a utility industry that has seen little growth in recent decades and is operating on infrastructure well beyond its intended lifespan.
As chip technology advances, far more computing power can be packed into each server rack. Where racks once drew just 3–15 kW, they are now scaling rapidly to 50 kW and 100 kW per rack and more. With hundreds or thousands of racks per facility, total demand per data center may climb to 2 GW and beyond.
For perspective, the average U.S. power plant produces roughly 0.2 GW A large nuclear plant can generate 1 GW, enough to power one million homes. But unlike homes, data centers require hospital-grade reliability. In the electric industry, this is defined as “five nines” of firm supply, meaning power is available around the clock at full capacity. Achieving five nines is difficult for a grid dealing with increasing outages from aging infrastructure and severe weather.
For hyperscalers, the challenge is not only securing enough reliable power but securing it fast enough to keep up with AI demand. And that is where speed to power comes in. With utility power delivery timelines running seven to nine years — a nonstarter for hyperscalers — the data center industry is exploring every imaginable alternative
Some of these solutions verge on the extraordinary as hyperscalers draw on their innovative roots to leave no stone unturned in their quest for fast and reliable power. Google, for example, recently signed an offtake agreement for fusion energy. While fusion holds long-term promise, it is a long way from commercial viability and therefore won’t solve the speed portion of speed to power.
Meta is setting up temporary tents and power to get additional data center capacity online as soon as possible, announcing that it will spend $60 billion on AI data center capex this year.
Geothermal has become a favorite among energy investors, bolstered by supportive federal policy. It is a proven technology with clear potential but development is constrained by geography and project timelines. In a stabilization wedges approach, geothermal deserves a role in the broader generation mix, but it cannot deliver the required scale or speed on its own.
How EVs could deliver
While hyperscalers are throwing everything against the wall and seeing what sticks, they may want to take a closer look at electric vehicles (EVs) to figuratively and literally deliver on speed.
EVs represent a vast and distributed pool of storage. Politics aside, the economics are already compelling. EVs are cheaper to operate on a total-cost-of-ownership basis than their combustion engine counterparts. Fleet adoption, from school buses to delivery vans, is accelerating, creating a resource that could be harnessed at scale.
Vehicle to Grid (V2G) technology enables these batteries to discharge power back into the grid when needed, transforming EVs into mobile storage assets. At scale, aggregated EV fleets could rival the capacity of traditional, utility scale plants, providing the kind of dispatchable power that hyperscalers urgently need.
Skeptics often point to the ‘mobile’ nature of EVs, given that, unlike other power supplies, cars can drive away. But with a fleet large enough, availability can be modeled and guaranteed providing power that is as firm and reliable as stationary batteries. Just as the aggregation of controlled hot water heaters has provided hundreds of megawatts of curtailable power to utilities, EV V2G fleets can commit firm capacity at the aggregate level, even though individual cars ‘drive away.’
While there are still challenges, the largest barrier may be a lack of imagination. Hyperscalers are accustomed to gigawatt-size power purchase agreements and utility-scale infrastructure. But the future grid will be increasingly decentralized.
The potential scale is enormous. A recent Union of Concerned Scientists study found that managed charging and V2G can create meaningful grid value even with modest participation. The sheer scale of storage embedded in millions of EVs is more than enough to offset the kinds of power deficits now stalling new data center projects.
On the business side, imagine a leasing program where consumers enjoy lower EV lease rates in return for enrolling in a V2G contract. In this scenario, the hyperscaler might guarantee each driver a reliable minimum range of 100 miles while reserving the remaining battery capacity for grid support. Drivers would then plug in at hyperscaler charging stations located throughout the utility’s service territory. Preliminary numbers show a windfall of electricity savings for hyperscalers.
While there are still challenges, the largest barrier may be a lack of imagination. Hyperscalers are accustomed to gigawatt-size power purchase agreements and utility-scale infrastructure. But the future grid will be increasingly decentralized. Just as cloud computing pooled server capacity, V2G offers a way to pool EV batteries into an orchestrated storage resource.
For consumers, it means cheaper EVs, a dependable charging network, and resilience during outages. For hyperscalers, it means faster pathways to growth and a hedge against interconnection delays.
Data centers are poised to become one of the defining infrastructure challenges of the 21st century. Meeting their voracious appetite for electricity will require not just building new supply, but building it faster and smarter. Vehicle-to-grid integration represents a promising “speed to power” solution, leveraging assets already on the road to deliver capacity where and when it is needed most.
Anna Demeo is the managing partner at CTSA, advising investors, startups, and corporations shaping the future of energy, infrastructure, and AI.
