Artificial Intelligence has seen a huge pick-up in demand in the past several years – and the bright lights of New York are attracting data centers.
They are being leased out of Manhattan from 60 Hudson Street to 32 Avenue of the Americas as co-location centers, according to Tony Gulotta, principal in charge at NYC office for global tax firm, Ryan LLC. He has worked with data center providers in the city.
Gulotta said that finding a location that has enough power is a challenge in the industry, but facilities on 111th 8th Avenue in addition to 60 Hudson Street and 32 Avenue of the Americas – have that luxury, which is why the city is seeing increased demand.
“They have capacity, and they also have the ability to provide them the power, which is another big thing that the AI servers need,” he told GlobeSt.
“The electronics need power, and they’re ready to go. So if you’re the biggest AI in the world of Microsoft, AWS, or you’re a startup, you still really benefit from going into one of those facilities.”
Gulotta added that the most important aspects of data centers are interconnecting and maintaining low latency on connection, which is prevalent in the Manhattan locations.
Right now, New York City is the hot commodity – but Gulotta can see demand growing in Georgia, Texas, and Orangeburg, New York.
“Orangeburg has a payment in lieu of a pilot program, where they’ve attracted a number of these larger data centers,”
“So they’re kind of creating the market with those incentives. And so those programs typically will give them a sales tax abatement. They also generally will give them a property tax abatement, and they’re in areas like Orangeburg, where it’s a lot easier to lay new fiber, it’s a lot easier to construct new buildings.”
In the short term, Gulotta sees the demand for data centers and AI continuing. But one of the challenges to growing the industry, in general, will be getting access to power.
“A smaller jurisdiction may not have the ability. So I think the thing that’s going to limit it would be the ability to provide the power.”
Plus, it’s not cheap. AI Severs cost about 40 times more than the amount of traditional data centers, according to a BlackRock’s iShares report. In figures, that could mean up to $12 billion per gigawatt for every extra data center power that’s needed, it said.