In his keynote kicking off Metro Connect 2025, Ganzi warned that the industry isn’t just in the business of digital infrastructure anymore; it’s in the power transmission and distribution business, as AI’s insatiable energy demands push the grid to its limits.
“We have a power distribution and transmission problem,” Ganzi said, adding the industry needs to think about how to “attack power”.
Subscribe today for free

“Here we are in 2025, with around 60 to 65 gigawatts of compute—an incredible jump over the past 11 years,” Ganzi said. “If you think back to the early days of public cloud, in 2011 or 2012, it took a decade to reach this point. And that was a decade when no one was really talking about data centres.
“You didn’t have 100 companies claiming they had a data centre, a piece of land, and adjacent power, ready to go. Now, it’s turned into an incredible game of speculation.”
Ganzi referenced recent comments from Masayoshi Son, the CEO of SoftBank and one of the leading architects behind the Stargate project, who framed AI as a multi-trillion-dollar Capex opportunity, estimating costs at $10 million per megawatt.
Calling Son a “Clintonian” salesman who could sell anything to anyone, Ganzi acknowledged the excitement but focused on a more grounded figure of 171 gigawatts of compute power expected over the next five years, a growth trajectory he sees as both staggering and realistic.
The same day Ganzi took the stage, reports surfaced that Microsoft cancelled multiple data centre leases across the US.
Ganzi told attendees that Microsoft’s decision to cancel “doesn’t matter” as capex on data centres will continue to go up.
“There's going to be more opportunity, not less opportunity,” Ganzi said. “You have to think of this almost a sequential bucket. Each year a bucket fills up. We do a little bit leasing. More water comes into that bucket, we spill it over, and it goes into next year's bucket. The bucket for 2025 is already full.”
Beyond power, the other issue Ganzi highlighted during his keynote was supply, with new data centre projects set to struggle to get their hands on vital equipment like turbines.
Lead times on key components like generators, turbines, and battery storage are stretching further than ever, Ganzi said, warning that even well-funded players will struggle with bottlenecks.
“If you decide to go build a data centre tomorrow and head to Germany to order turbines from Siemens, they’ll tell you, ‘See you in 2029,’” he said. “This is one of the biggest problems we have.”
But while supply chain constraints are slowing deployments, demand continues to surge. DigitalBridge itself has 16.8 gigawatts of power capacity, a figure Ganzi claimed is the largest power bank in the industry, surpassing data centre giants like Digital Realty and Equinix.
The AI boom, he stressed, isn’t just driving investment in data centres – it’s reshaping digital infrastructure as a whole.
The shift to AI inference, which moves workloads closer to end users, will require a new wave of fibre deployment, edge data centres, and low-latency networks to handle the 10x increase in mobile data traffic expected over the next five years.
According to Ganzi, AI workloads will be spread across four critical areas: consumer, enterprise, device-to-device, and government.
“Remember, with data sovereignty, the amount of data that's being stored, modeled, consumed, and shaped is like we've never seen before,” he said, arguing that distributing AI processing with data sovereignty requirements would make middle and last-mile connectivity essential components of the AI infrastructure ecosystem.
“Everything about generative AI is about location and latency,” he said. “Latency is what we do in this industry, and in fibre, we’re going to be very busy for a very long time.”
Ganzi also highlighted the aging infrastructure problem facing the industry. “A lot of people don't talk about the aging backbone of the internet,” he noted, pointing out that some networks are now 25-30 years old.
“Glass dissipates over time. That's just science, folks,' Ganzi explained, emphasising that billions, if not hundreds of billions of dollars will need to be spent rebuilding fibre infrastructure that can no longer meet modern demands.”
And for investors worried about whether AI infrastructure spending will hold up in the face of cost concerns, Ganzi had a simple message: the capex cycle is only accelerating.
He pointed to DigitalBridge’s $26 billion war chest for 2025 – $9 billion in equity and $17 billion in debt as proof that capital is still flowing into the space.
Private credit, in particular, is playing an increasing role in financing digital infrastructure, with $90 billion in non-traditional digital infrastructure debt expected to be placed in the next five years.
“The credit markets are open,” he said. “But the way we finance this sector is changing.”
“Digital credit has become a big part of what we do,” Ganzi said, adding that it was among the fastest-growing verticals at DigitalBridge, with its $4.5 billion business set to double in size this year.
RELATED STORIES
DigitalBridge CEO on the AI revolution’s impact on investments
Metro Connect 2025: Fibre frenzy takes centre stage as tower sector falters
Metro Connect 2025: IBM’s former AI chief on the tech set to replace today’s semiconductors
Metro Connect 2025: Medical pioneer's trillion-dollar AI infrastructure vision