Chapman Debuts 1st ‘Community Supercomputer’

Chapman University unveiled the world’s first “community-operated” supercomputer on Sept. 23.

Nvidia Corp., the $1 trillion-valued maker of graphics processing units for artificial intelligence (Nasdaq: NVDA), is backing the initiative.

“There are hundreds, maybe thousands, of supercomputers in the world, but there’s not one single supercomputer governed by the community,” Keith Strier, Nvidia’s vice president of Worldwide AI Initiatives, told the Business Journal.

“We are trying to learn how to make this technology not just accessible, but really add value to a community governed by the community.”

The supercomputer is one of the leading efforts of the CEO Leadership Alliance of Orange County (CLAOC), a group of high-level executives, to kick-start artificial intelligence and higher paying jobs in Orange County.

Strier calls it a community supercomputer because while it’s located in Chapman’s data center, other entities in Orange County like local governments and community colleges will have access to it.

The first community project with OC Public Works is already underway, using the computer to make digital renderings to help with building and material plans.

Trace3 Support

The supercomputer, capable of performing 8 quadrillion calculations a second, spans 64 GPUs, or graphics processing units, taking up 4-by-6 square feet. The supercomputer isn’t as powerful as ChatGPT, which uses about 10,000 GPUs, Strier said.

The computer will be at Chapman’s Beckman Hall data center due to its size and power/cooling requirements, according to Chapman President Daniele Struppa. All computation can be done by connecting remotely from other parts of campus, according to Struppa, a mathematician by training.

The computer is among the fastest in Orange County.

The fastest supercomputer in California is at Nvidia’s campus in Silicon Valley, according to Strier, who works out of OC.

Besides Nvidia, other backers of the project included Trace3 Inc., the Irvine-based tech consultancy, which installed the system with the help of companies like Dell Inc.

“This is a complete investment from our side to the community, which we’re big fans of,” said Josh Lindstrom, managing director of data intelligence at Trace3.

“Five years ago when I came to Trace3, we had to go knock on the doors and ask if they do algorithms or model building and that type of thing. Now we’re actually seeing these folks come to the forefront. So, when Chapman came to me, I said, ‘Yeah, absolutely.’”

The computer also possesses cutting-edge software from upcoming names in the AI space: Run:ai and Deci AI.

The Israeli-based companies automate resource management using generative AI.

“We have not just a state-of-the-art supercomputer, but a software stack that would rival most Fortune 500 companies or any research one university,” Erik Linstead, senior associate dean at Chapman’s Fowler School of Engineering, said at the launch event.

Chapman’s engineering school, which oversees the supercomputer, plans to give students hands-on experience for their engineering and computer science courses. It will be made available to all students regardless of age, particularly lower-income people with fewer resources.

“We need to make sure students have access to AI to compete in the classroom,” Linstead said.

Tech Talent

Strier said the leading university for AI nowadays is the University of Florida, which has set up a team of interdisciplinary experts to teach all its schools to use its supercomputer, called HiPerGator, which received a $25 million partnership from Nvidia.

Strier noted that AI is far more pervasive than realized.

“Today, the vast majority of average citizens don’t think we need to engage with AI all the time, but every time you search on Google or hop into your Tesla, you’re engaged with AI,” Strier said. “AI is pervasive.”

The experts at Chapman downplayed concerns that AI will take away jobs.

“I always laugh at the fear that [AI] is going to take jobs,” Lindstrom said. “If you actually embrace it like what we’re seeing in real time in all the industries, it is actually going to help you.”

Declining Costs

In 2008, the IBM-backed Roadrunner was the first supercomputer to deliver 1 petaflop of processing power, or 1 quadrillion calculations per second.

It cost approximately $120 million to build, according to Strier.

He said an 8-petaflop computer—what’s being installed at Chapman—in 2008 would have cost about $1 billion to build.

Prices have come down as technology has improved. The cost of the project in Orange, which are donated from the companies involved, altogether ran around $1 million.

AI’s ‘New Reality’

A “new reality” is emerging worldwide about artificial intelligence, according to Keith Strier, who is vice president of Worldwide AI Initiatives for Nvidia Corp., the world’s biggest maker of graphics processing units for artificial intelligence.

About 15% of the world’s nations have a supercomputer within their borders, said Strier, who lives in South Orange County.

“You have only 31 countries in the world that can compute a large-scale model; the other 170 countries are fully dependent on those 31 for AI—that is the new reality,” Strier said.

The problem is that supercomputers require a tremendous amount of energy at big data centers.

“One of the main bottlenecks in developing countries, even a sophisticated country like Singapore, is running out of data center space,” Strier said.

“They have nowhere to power it. It has profound implications for their economy and national security.”

Strier last month spoke at the 78th UN General Assembly in New York about the “digital divide.”

“Up until a year ago, no one cared about access to AI,” Strier said, “Now it’s become such a critical problem. Can we protect our country’s infrastructure if we can’t compute large scale models?”