A multi-billion-dollar data center campus development underway in Grayslake, known as T5 @ Chicago IV, has made big promises for the area, but one Chicago artificial intelligence expert questions whether it and similar projects across the country are building toward the wrong future.
Data centers are facilities that house physical equipment, such as servers and other kinds of computers, to store digital data and provide computation power. They’re for the basic computing, processing and storing of “anything and everything related to the internet,” as Dan Diorio, vice president of state policy for the Data Center Coalition, put it.
Diorio said there has been a construction boom in recent years, with the seasonally adjusted annual rate of construction spending jumping from $19.5 billion in July of 2023 to nearly $41.2 billion in July of 2025, according to U.S. Census Bureau data.
The Grayslake project is part of this growing trend, and if fully built out, it would have over 10 million square feet of data center space, bringing thousands of jobs, and costing anywhere from $8.5 billion to $18 billion, depending on who you ask.
Many argue that the trend is being driven by AI, which requires large amounts of data and computer power to train and run. National headlines continue to raise concerns over a potential AI investment bubble that, if it were to pop, could drag data center development down with it.
But Northwestern University computer science professor Kristian Hammond said he’s more worried about the move towards centralization for data centers and AI.
Hammond is a cofounder of Narrative Science, an artificial intelligence company that created an AI model that would take data and generate a readable narrative for the less data-inclined. It was acquired by Salesforce in 2021.
When an AI responds to a question or prompt, it requires data and computation power, two things data centers provide. Current versions of AI models are “data hungry,” Hammond said, requiring massive amounts of information to learn.
Current trends see the creation of larger and larger AI models, he said. Essentially, one model is being trained “to be able to do everything.” But a possible path forward could be to go the opposite direction, shifting to smaller models that each require less data but have more expertise in a specific field.
“You wake up in the morning, and you have a headache and a fever, you want to put together a family trust and your taxes are due. It’s not like you call one person,” Hammond said. “You get hold of your accountant, your lawyer and your doctor.”
Currently, major companies in the AI field are building large data centers and pushing the idea of large-scale models, “as opposed to thinking in terms of innovation in the direction of smaller, lighter-weight, local models,” he said.
But all this data being held or sent to centralized locations presents a few issues, he said. First, there are potential security risks, both for individual companies and the country at large. For businesses, they could be concerned about proprietary information being leaked, and might prefer moving processing inside their “firewall,” a term for a network’s security system.
T5 Data Centers CEO Pete Marin said in a statement that while some workloads may “eventually” move behind corporate firewalls, “we haven’t seen any widespread strategy or investment trend supporting a shift away from large-scale infrastructure.”
“AI at scale remains the dominant and most efficient model for training and deploying advanced systems,” Marin said.
More broadly, Hammond also said such data centers are enticing targets.
“If I have a map of all the high-powered data centers that exist in the United States, that’s great in terms of understanding how to get information, and fantastic in terms of looking for targets for things I can take out and hurt our information infrastructure,” he said.
When it comes to concerns about an AI bubble, while Hammond said there is “genuine concern” about not “seeing enough in the way of companies returning on their investment,” and a focus on consumer rather than enterprise innovations, he views it as a case of “ just having not found the right use cases.”
“But in terms of real utilization — oh my God, people are using it all the time,” he said. “We just aren’t using it in a way that is clean, precise, enterprise solutions.”
He worries the emphasis on centralization could muffle innovation on more local, smaller models that could “get us to real enterprise business solutions, as opposed to the really big, really flashy, really powerful.”
Right now, Hammond said, consumer-facing products are helpful, but there hasn’t been “a real focused effort” to bring the technology into businesses at scale and integrate it into workflows.
“Everyone in your organization using ChatGPT is not business at scale,” he said. “Having a system that is integrated with all your other systems and is invisible to people, that’s the scale we are looking for.”
Hammond wonders if the push for these large-scale data centers was making assumptions about a future that might not come about. While the modern world requires compute power, that doesn’t necessarily mean centralization.
“It’s an odd thing to decide we need these data centers when it’s not clear how businesses are going to use this technology,” Hammond said. “Everyone is excited about this technology. It’s unbelievably powerful, but we haven’t figured out how to use it yet.”
AI is not the driver, DCC says
The Data Center Coalition, for its part, argues the data center boom isn’t entirely tied to AI, but rather a general need for digital infrastructure in the modern world.
The coalition pointed to a study by Chicago consulting firm Bain & Company, which showed that in 2024 seven gigawatts of global data center capacity were AI-related, as opposed to 61 gigawatts called “general compute.”
Those numbers are predicted to shift significantly in the coming years, however, with Bain & Company expecting AI to take up nearly half of global data center capacity by 2030, even as capacity rises.
Outside of AI, there is still a “steep demand curve,” Diorio said, and cloud services and the proliferation of items connecting to the internet are still the “significant growth drivers” for the industry. The vast array of other technologies and services, from apps, online learning, telehealth, even autonomous vehicles, fall under the digital infrastructure provided by data centers, Diorio argued.
“It’s important to remember that this is the backbone of the modern internet,” he said.
Diorio also said the industry is working with “important stakeholders” to avoid over- or underbuilding, and “always evaluating demand, evaluating needs.”
https://www.chicagotribune.com/2025/11/13/grayslake-data-center-2/