With generative AI, you know it looks like magic. Chatbots and multimedia models make it easy to come up with poems and high-resolution videos with a snap of your fingers.
However, the fast output and sophisticated interfaces of AI models mask the vast amount of physical infrastructure behind them. And as AI continues to grow, the data centers and power plants where it is built are beginning to gain widespread attention outside of the industry.
Earlier this week, I took a train to Orangeburg, New York. luckA newsroom in downtown Manhattan. I was visiting Orangeburg, one of a growing number of AI infrastructure projects popping up across the country, and the future home of data center company DataBank's latest site, LGA3. .
DataBank already operates two data centers in the New York metropolitan area. One is in Newark, New Jersey, and the other is in Chelsea, Manhattan. However, LGA3 will be its largest site. The $250 million, 200,000-square-foot facility will consume up to 45 megawatts of energy to power five giant data halls packed with computer chips.
Although the facility won't open until next year, tenants have already reserved space, most notably New Jersey-based AI startup CoreWeave. The company recently secured a staggering $19 billion valuation and has already reserved almost half of LGA3's capacity.
“The explosive growth of artificial intelligence has required a complete re-evaluation of traditional data centers to meet next-generation computing requirements. Provide some of the technology CoreWeave founder Ben Venturo wrote me a note.
My taxi left the station and took four wrong turns as it drove past farmhouses and office parks on its way to the LGA3 construction site between the substation and the New Jersey state line. My first impression when I got out of the car was the size of the building. LGA3 is a huge, one-story hall the size of a New York City block with high ceilings, so it wouldn't be surprising if commercial jets flew inside.
“Technology addiction”
DataBank CEO Raul Martynek greets me, wearing clear-rimmed glasses and a lavender button-down. Martynek has been involved in the Internet infrastructure industry for decades, since the advent of the commercial Internet in the 1990s. He has been with Databank since 2017, overseeing the company's 69 data centers in the US and UK. Martinek told me that he hasn't seen the kind of explosion in demand for digital infrastructure that AI is creating since the recent dot-com bubble. 90's.
“Humans are dependent on technology. And ultimately, for the data center sector, what we're doing is enabling humans to adopt more technology,” Martinek said. he told me. “As you deploy more technology, you need more fiber, more cell towers, more data centers. The center is the bottleneck.”
Nathan Howard/Bloomberg
The massive increase in demand for AI has propelled data centers to front page headlines. Martinek says that most of the things we do online today, from accessing images on our phones to scrolling through social media to prompting ChatGPT, involve more physical interaction than we realize. He explained that some hardware was involved. Wi-Fi routers and cell towers constantly send signals through underground fiber optic cables to data centers and remote servers to access stored information and keep the Internet running.
“The Internet is a network, right? Information is sent as photons through fiber optic cables, and they travel around the world at near the speed of light,” Martinek said. “I was talking to my network guy last night and said, 'What do you do?' He said, 'We're plumbers, right?'
And these days, plumbing has become a good business. The amount of data generated on and by the internet has increased exponentially over the past 20 years, and with the expectation that it will become even faster with AI, the demand for space to store all that information. is increasing.
“This device didn't exist until 2007,” Martinek told me, pointing to his iPhone. “Now think about how much content and how many applications have been created. [by it.] All that stuff ends up in the data center…that's the physical ecosystem. ”
Provided by: Source code communication
AI increases the need for architectural space
Data centers may not be the sexiest projects, but the surge in demand from AI companies is attracting huge amounts of money, lots of press coverage, and even some of the biggest names in the construction industry are getting on board. DataBank alone has spent approximately $4.5 billion on data center projects since 2016. Tishman Speyer, the real estate company building LGA3, is one of the most well-known companies in the industry. The company worked on the World Trade Center and Chicago's John Hancock Center, and also owns Rockefeller Center. A low-slung data bank next to a suburban Little League baseball stadium may seem an odd addition to the company's portfolio, but the company is betting that data centers will prove to be just as important as high-rise buildings. There is.
When I first received the invitation to visit, I was amazed by the place. new york? Are there some of the highest real estate and energy prices in the country? Wouldn't it be cheaper to build this in the middle of the desert, where land is cheap and the cheapest renewable energy is available?
However, Martynek explained that for many customers, being based thousands of miles away from one of the most important parts of their business is not practical. New York is one of the nation's largest data center markets, with about 800 megawatts of capacity currently online, much of it for financial and technology companies that rely on nearby computing power to build and trade around the clock. It corresponds to
“It's not practical to have a data center in a remote location. The latency is too high,” Martinek said, referring to response time delays between computers and off-site data centers. “Too many things can happen along the way.”
“Data centers tend to be clustered around metropolitan areas,” he continued. “New York has always been a very large data center market. It's really a function of the population, it's a function of the companies. If you were J.P. Morgan, you wouldn't want to put a data center in Omaha.”
Public policy is also a factor. Earlier this year, New York Governor Kathy Hochul announced the state's Empire AI initiative, which earmarked more than $400 million for data center projects and other infrastructure.
I put on a hard hat and reflective vest and walked around a building under construction with two construction managers. They pointed to an area the size of an airplane hangar where the bank of computer chips will eventually be located, as well as ventilation and water cooling to prevent overheating.
Once construction is complete, CoreWeave and DataBank's other customers will begin installing chips, and DataBank expects the facility to be fully operational by early next year. Once online, CoreWeave will begin leasing its computing power to tech startups and other AI companies. Martinek said DataBank has had no trouble finding customers.
“We signed a contract with CoreWeave last year. This building didn't even exist at the time, it was just dirt. There's such a demand for this product,” Martinek said. “There's a frenzy going on.”
As the saying goes, strike while the iron is hot, and Databank is already finalizing plans for another site right next door, LGA4. Next time you're driving around town, take note of the unassuming buildings nearby. The AI data center boom may be closer than you think.