Cal Poly students and faculty now have the computing power to build artificial intelligence models of their own.
The university’s new AI Factory, located underneath Building 14, began operating in June of this year and was formally unveiled Oct. 22. Funded through the Noyce School of Applied Computing, the $3 million facility intends to expand research opportunities and give students hands-on experience developing artificial intelligence.
“The AI Factory exemplifies Cal Poly’s learn by doing philosophy, enabling our students and faculty to move beyond using AI to shaping it,” Provost Al Liddicoat said in a university statement.
Unlike ChatGPT and other AI chatbots, the factory isn’t an artificial intelligence model itself. Rather, it’s the computing infrastructure that allows researchers to build, train, test, and operate different types of AI models, including large language models.
The facility consists of four NVIDIA DGX B200 systems equipped with graphics processing units, or GPUs—specialized computer chips capable of performing the large numbers of calculations required to train advanced AI models.
Computer science assistant professor Alex Bisberg, who helped prepare the system for use, said artificial intelligence has existed for decades, but recent advances have dramatically increased the computing power needed to develop it.
“When I was a student, AI was something that could run on hardware small enough—hardware meaning, literally just a computer, like a laptop—it could run on hardware that basically everyone had at that point,” Bisberg told New Times.
He explained that GPUs, originally developed to process computer graphics, are particularly useful for AI because they can perform mathematical calculations simultaneously, allowing models to analyze large amounts of information.
“The rise of model complexity has led [to] increased demand for hardware sophistication and complexity as well,” he said.
For students and researchers, accessing that equipment can be expensive.
“Students literally could not afford to buy the hardware to run these models,” Bisberg said.
Much of today’s advanced AI infrastructure is controlled by private technology companies, leaving researchers dependent on commercial platforms to access the computing power needed for their work.
Bisberg said operating its own equipment gives Cal Poly greater control over the models researchers develop, a concept known as data sovereignty.
“The key thing here is that these closed-door companies, we can’t alter or change the models that they give us,” he said. “We just have to kind of take what we can get.”
The factory’s opening comes as the expansion of artificial intelligence infrastructure draws scrutiny over its environmental footprint, particularly the electricity and water required to operate large commercial data centers.
Cal Poly’s system is considerably smaller than those facilities, occupying approximately 50 to 100 square feet in an existing campus building, according to Bisberg.
“These servers are sitting in the basement of … a building on Cal Poly where … we’ve had the space available for … 50, 60 years,” he said.
According to Cal Poly spokesperson Keegan Koberl, the university’s solar farm helps supply electricity to the facility, although specific figures for its energy consumption were not provided.
The factory’s equipment and software are maintained by university employees, allowing researchers to store data and operate models without relying on third-party cloud services, Koberl said.
“Much of generative AI is administered by massive … private companies,” Bisberg explained. “The problem is all this research is happening behind closed doors and these companies are releasing studies without peer review. So, for us as scientists and universities, our job is to shepherd this this technology for the public and ensure that it’s being deployed in the best way it possibly can.”
“I think it’s maybe not a stretch to say that these companies might not always necessarily have the public’s best interest at heart,” Bisberg added.
NVIDIA and technology partner Mark III Systems provided software licenses, technical support, and access to experts, Koberl said.
“As an endowed school, Noyce exists to drive innovation like this, and this system will let us train our own large language models from the ground up and design applications around them,” Chris Lupo, founding director of the Noyce School, said in a university announcement.
The factory began operating in June with a small group of faculty researchers. Current projects involve video analysis, bioinformatics, marine science, business analytics, and public health.
Computer science professor Jonathan Ventura is using machine learning to analyze drone imagery of eelgrass in Morro Bay. According to Cal Poly, his system has produced results comparable to manual analysis while processing images approximately 10 times faster.
Computer science professor Paul Anderson is using AI for research involving spinal cord injuries and genetic risk. Keeping the computing infrastructure on campus allows researchers to work with sensitive medical information without transferring it to outside AI services, according to the university.
“Before the AI Factory, we couldn’t work on a problem like this,” Anderson said in a university statement. “Now we have the computer to train frontier models.”
Students in advanced AI courses began using the factory this fall, although broader access remains limited as the university develops training materials and policies governing its use.
Koberl said the system will eventually be available to students across all majors through their Cal Poly accounts.
“I think we do have to just be careful on … who we let use the factory [and] when, just because it is a limited resource,” Bisberg said.
Tobias Halpern, a sophomore business student involved with the campus organization Tech for Liberation, said he has mixed feelings about the university’s investment.
“I don’t like the way AI is used in our school system, and I don’t like the way that it’s pushed,” Halpern told New Times.
He expressed concern about the environmental consequences of AI infrastructure, particularly large data centers constructed in communities that may not directly benefit from the technology.
Despite his reservations, Halpern said he considers the factory a step toward more responsible use of the technology.
“You’re realistically not going to get the [computer science] department to stop using AI,” he said. “They would fall behind if they did. You have to figure out how to manage that use and make that sustainable and responsible. And I think this is one way to do it.”
Bisberg said universities need to teach students how AI functions, rather than simply how to use commercial tools.
“We do have a responsibility to teach the tools. A holistic education about how they work is really important,” he said. ∆
Reach Staff Writer Chloë Hodge at chodge@newtimesslo.com.
This article appears in Nov 5-12, 2026.
