"$4 billion dollar in the ground"—Shark Tank investor Kevin O'Leary breaks down AI’s energy dilemma and the future of data centers

Shark Tank investor Kevin O’Leary recently discussed that artificial intelligence is driving advancements across industries, but its increasing energy demands pose a major challenge. In a Morning Brew Daily interview, a portion of which was posted on Investocracy Markets’ YouTube channel on February 18, 2025, O’Leary underscored the significant investment required for the same, stating:

"What does it cost to build one of these? $2 billion just for the platform and another $2 billion for the stacks of chips. $4 billion dollar in the ground for a one-gigawatt data center."

Kevin O'Leary further shared that with the expansion of AI adoption, finding solutions for energy generation and infrastructure will be critical to sustaining this technological shift.

---

The energy challenge of AI: Shark Tank Kevin O’Leary on data centers and power constraints

The growing demand for AI infrastructure

According to the Shark Tank investor, AI-driven companies such as Google, Microsoft, Amazon, and Tesla require high amounts of computing power to support their operations. This demand has led to a surge in the need for data centers, but constructing them presents logistical and financial challenges. Kevin O’Leary highlighted the scale of this issue in the interview, stating:

"In order to just meet the demands of the seven largest hyperscalers—the Teslas and the Googles and the Microsofts and the Amazons—they have demand for 30 GW of data centers."

The Shark Tank investor noted that traditional data centers were built at 50 megawatts, making current demand far greater than before. The investor also explained the geographical concentration of existing facilities, particularly in Virginia, which houses 45% of the nation’s data centers.

However, new regulations in the state now prohibit backup generators for these facilities, which creates concerns about their ability to maintain uptime. O’Leary explained that modern data centers must meet a reliability standard of 99.999%, meaning they can only experience around 14 seconds of downtime per month.

---

Energy limitations and infrastructure challenges

The expansion of AI infrastructure is being hindered by energy supply constraints. The Shark Tank investor pointed out that securing power for new data centers is becoming difficult.

"I don’t care what state you go to, you go to the power authority there and say, ‘I need, let’s say, just 100 megawatts,’ [and they’ll say] ‘Can’t do it. We have no power in America right now.’"

Beyond supply limitations, public opposition can also be a barrier. O’Leary explained how the construction of a new data center can impact electricity prices for residents, leading to pushback from communities.

"If you call up... ‘Hi Mr. Slabinsky, I want to build a data center next to your house, and your electrical bill is going up 40%.’ No thank you. That’s the problem," the Shark Tank investor said.

He added that resistance from residents often results in political and regulatory challenges, further complicating data center expansion.

---

Potential solutions for AI’s energy needs

To address these challenges, alternative energy sources must be considered. Kevin O’Leary discussed a possible approach, which involves utilizing stranded natural gas in remote locations to generate electricity.

"I think the way we’re going to solve it is we’re going to find stranded gas in places like West Virginia, North Dakota, Texas, bring our own generators in the middle of deserts or in the middle of the Arctic or wherever it’s got to be, and generate electricity from clean natural gas and build data centers there."

He emphasized that resolving the power issue is crucial before launching new data center projects. The Shark Tank investor explained that his data center announcements occur only after securing a reliable power source.

---

Shark Tank airs on Fridays at 8 PM ET on ABC and is available for streaming on Hulu.