Data centre energy demand is not going to wait for supply chains to untangle tariff logjams and geopolitical complications.
It doesn’t have time.
That demand is also set to rewrite today’s power grid equation, because competition for power is intense today and about to get far more intense tomorrow.
Why?
Ask Peter Sacco.
“We’ve identified that data centres and AI [artificial intelligence] are going to expand electrical consumption,” the president and founder of PTS Data Center Solutions said during an October 30 Energy Central Data Center’s Green Revolution discussion, “[but] we’re not even considering what it’s going to be like when robotics comes online.”
AI-enabled robots, he noted, “are like mini-Teslas, and you know when Elon Musk figures AI will build not one million, not one billion, but 10 billion robots, what does that electrification look like to the grid?”
I’d say unmanageable.
North America’s grid has neither the infrastructure nor the technology today to service an energy-hungry AI industry.
So, Mr. Sacco, how will that AI data centre industry satisfy its energy appetite?
“The answer is that the AI industry is not going to wait. Let’s remember what the AI industry is. [These are] the world’s leading innovators, and they are not going to let supply chain [delays] and lack of power be the thing that interrupts us from becoming the world’s leader in AI.”
Supply chain inefficiencies are serious impediments to an industry in a hurry.
Time, after all, is money. And investment money dislikes delay. Power utility companies and energy infrastructure developers aren’t keen on it either.
Mike Smith, principal of the KLN Group power utility consultancy, offered an anecdotal example of the data centre delay dilemma.
He described a recent conversation he had with an East Coast energy utility executive to illustrate the industry’s frustration with how supply chain delays hurt an energy company’s ability to grow and service its customers.
Smith, the Energy Central data centre panel discussion’s moderator, said the utility’s executive pointed out that company planners for substation installations in new subdivisions for the past two decades could rely on wait times for transformers of “three to six [or] maybe nine months. … Well, that three to nine months has become, maybe two to three years or even longer.”
Those supply chain delays originate in Shanghai/Ningbo, Rotterdam, Qingdao, Singapore, Norfolk, Houston and other major ports, as listed in the most recent Linerlytica port congestion data, and are compounded by 21st-century tariff uncertainty and geopolitical turmoil. Port performance, especially in North America, has suffered.
Regulation and permitting also add years to infrastructure completion timelines.
But, as Sacco pointed out, data centres do not have time for multi-year delays.
The workaround, he said, will be off-grid “behind the meter” options.
So, onsite electricity generation. Call it local energy near-shoring, rather than waiting years for a main power grid connection.
In short, energy decentralization, which Sacco said, “is going to be the key.”
And not just for energy efficiency.
“I think we’re going to see decentralization of the internet, decentralization of the public cloud, decentralization of the energy industry,” Sacco said, “in order to serve the massive amounts of power and distribution of information that is going to be needed in the AI age.”
And what will power local data centre substations while the continent’s utility grid catches up to that energy demand?
Natural gas or hydrocarbon derivatives, Sacco said, “while we build out a nuclear industry, which, by the way, is the inevitability of where we're going to go.”
He added that small nuclear reactors in the U.S. could achieve commercial viability in “less than eight years.”
The research and innovation being invested in batteries, gasification plants, direct chip-to-cooling and the other supporting technologies for small onsite power generation plants, Sacco said, will help service data centre power appetites in the short term until main power grid options can meet the demand explosion.
The focus will also be on data centre quantity: smaller and medium-sized data centres that require less power and face fewer regulatory complications than their larger counterparts.
But Sacco said that will not eliminate the overall utility grid supply challenge.
“Because … if I were to quantify AI as a clock – and I can’t stress this enough – if AI were a 24-hour clock, we’re at 2 a.m. We’re just getting started. We’ve got 22 hours left to go, and by noon, we’re not even going to recognize the world, and we are marching exponentially fast towards it.”
The need for computing power will redouble electricity demand and amplify the need to remove the roadblocks to servicing it.
“So, [data centres] are going to do whatever it takes to get the power in the ground now,” Sacco said, “because they realize that time is money with respect to AI, and so it is going to push the government to figure out how to cut regulation, cut red tape and allow technical innovation to take place, because this is the most fundamental, transformational technology that man has ever wrought.”
www.linkedin.com/in/timothyrenshaw
@timothyrenshaw
