Supercomputing and AI: How the New Jersey Supercomputer Initiative is unleashing innovation

Reading Time: 8 minutes

Supercomputing and AI fuel modern research

The scale of modern research is accelerating.

Universities are generating and analyzing massive amounts of data—far more than traditional computing environments were designed to handle. These data insights are opening doors to new discoveries on how to cure disease and understand climate patterns. But the data quickly confronts physical constraints that limit how far innovation can go.

Research increasingly requires massive amounts of compute to power to fuel its data-driven outputs. This is why supercomputing is becoming so foundational for modern research and innovation. Supercomputing brings powerful computers and specialized networks to solve complex, data-heavy problems at speeds traditional computing can’t match.

At the same time, many organizations lack the infrastructure firepower—compute, networking speed, energy and cooling, and physical space— to drive this innovation. Leaders are recognizing these constraints.

“In research, computing power acts like a currency,” noted Gov. Mikie Sherrill at the announcement of the New Jersey Supercomputer Initiative (NJSCI) on Oct. 20, 2026. “There is a ceiling on the types of questions that researchers can ‘afford’ to take on with their machines.” Gov. Sherrill announced the partnership, which aims to address these infrastructure constraints and fuel research and innovation.

What is the New Jersey Supercomputer Initiative?

States like New Jersey—home to myriad higher-education institutions—recognize they have an opportunity to strike up partnerships that bring stakeholders and their resources together to give modern research momentum.

The New Jersey Supercomputer Initiative is a $125-plus million public‑private partnership that joins universities, state institutions, and researchers. It provides shared computing resources for AI, healthcare, environmental, and other high‑demand research. Housed in SHI’s existing New Jersey facilities, NJSCI will double the combined supercomputing capacity of New Jersey’s research universities. The project spans several years and takes a phased approach to building a supercomputing facility with security, redundancy, and cutting-edge infrastructure to support modern research, collaboration and “build new skills and prepare for emerging roles,” noted Chris Malachowsky, NVIDIA co-founder, New Jersey native, and benefactor to the NJSCI in the amount of $75 million.

Data suggests that universities have warmed to these kinds of projects, which aggregate supercomputing and AI resources because to drive innovation at speed and scale. High-performance computing capacity at leading U.S. research universities is growing about 18 percent annually as institutions respond to rising demand from AI and data-intensive science.

‘A big problem and an even bigger opportunity’

For researchers like Tiana Noll -Walker, a Ph.D. student at Rowan University (one of the partner institutions of the NJSCI), supercomputing offers the right tools to explore human disease and therapeutics to treat those diseases.

Noll-Walker is using genomics to understand disease and therapeutic drugs that could address those diseases. Access to supercomputing could change the types of questions that her team can ask and answer. But as she noted during the announcement event, her ability to ask those questions is defined by her access to GPUs, networking, power, and physical space.

“If a human genome were printed in 12-point type, it would stretch from Houston to Boston,” Noll-Walker explained. While that is the length of a single genome, genomics may compare thousands of genomes, correlate them with other biological data and then train AI models to detect relationships buried within them. “Right now, in the biology space, we have a big problem and an even bigger opportunity,” Noll -Walker said at the announcement of NJSCI.

“We have massive amounts of biological data,” Noll-Walker noted. “AI gives us the opportunity to find patterns and connections within that biological data that humans wouldn’t be able to do otherwise. These connections help us understand genes, proteins, and cellular interaction. …. We are able to make better medicines, design safer drugs, and find cures for disease [as a result].”

In the area of climate pattern research, Dr. Michel Boufadel at the New Jersey Institute of Technology (NJIT), discussed the benefits of supercomputing to predict flooding. NJIT can help communities prevent or mitigate flooding with solutions like building a reservoir or water diversion and predict where flooding may have severe impact.

“We are dealing with different towns with different infrastructure, different landscape, and different flood risk. We have to be very careful in our analysis.” Boufadel said. Using drones that generate “billions of data points,” NJIT can map the area. IT pairs the drone data with rainfall data collected every 15 minutes from the National Weather Service, which also generates massive amounts of data. It can then predict waterfall in regions with potential flooding.

“Supercomputing is really helpful for us in New Jersey,” Boufadel said. “It allows us to move from big data to well-informed action.”

What technologies is the NJSCI using?

To deliver the vision—a flexible and scalable foundation for high performance—SHI played a central role as coordinator and integrator of a complex set of technology partners, including NVIDIA, Dell Technologies, DataDirect Networks, Inc. (DDN), Cisco, Palo Alto Networks, ClearML, Secuvy, and Vertiv.

The proposed AI factory architecture combines NVIDIA DGX Rubin NVL8 accelerated computing platforms, DDN’s high-performance storage, NVIDIA InfiniBand and Ethernet networking, Dell infrastructure and cluster management capabilities,  as well cybersecurity and data governance solutions, and enterprise-grade AI orchestration and automation tools. In addition, the AI factory includes ClearML’s enterprise machine learning operations (MLOps) and AI orchestration platform, Secuvy’s AI-driven discovery and classification software, and Vertiv’s advanced liquid cooling,

Supporting these technologies is a modern, scalable high-performance compute infrastructure featuring the latest rack systems, liquid cooling technologies, coolant distribution units (CDUs), and power management solutions designed to maximize performance, efficiency, sustainability, and operational resilience for the most demanding AI and high-performance computing workloads.

The NJSCI is built based on NVIDIA’s reference architecture and key best practices, such as the following:

• Establishing an engine for repeatable success
• Building a foundation for high performance
• A platform approach, vetted by a center of excellence (COE)
• Unifying expertise (that is, engineering, architecture, and deployment knowledge) from top NVIDIA Partner Network Solution Providers)

Today, in phase one, SHI is providing approximately 500 kW of power to the initiative—power that was already available at the facility. But the project will require additional in the future. In phase two, SHI will work with partners to “provide 2.9 MW of additional power to the building that will be dedicated to the New Jersey Supercomputer Initiative cluster, said Jon Stumpf, principal architect for AI and data center at SHI. This will also involve building a new server room at the New Jersey facility to provide greater resilience than that of the current lab on its own.

“We are going to build a server room dedicated for this purpose, on a slab (as opposed to traditional raised-floor server rooms), with liquid cooling, with much higher redundancy than we currently enjoy in the lab,” Stumpf said.

Phase two of the project will also add capabilities so that stakeholders can work on the projects with government entities that require more stringent data security protocols. This opens doors for researchers to participate in grants or contracts from which they might now be excluded.

In phase three, currently scheduled to begin in 2028, the SHI facility can drive even further toward the full vision of an AI factory. While a traditional data center typically handles diverse workloads and is built for general-purpose computing, AI factories are optimized to create value from AI. They orchestrate the entire AI lifecycle—from data ingestion to training, fine-tuning and, most critically, high-volume inference. As the project evolves, the infrastructure will be built out to support this AI factory end state.

Public-private partnerships as difference makers

Individual universities cannot marshal the resources to solve problems that demand megawatts of power, thousands of GPUs, petabytes of memory and storage, networks capable of moving data at hundreds of gigabits per second and systems requiring liquid cooling. This is where collaboration and partnership can make the difference.

Combining government investment and research capacity with private-sector technology, and expertise can accelerate major initiatives.

“Technology is changing how we discover new medications, solve problems, and build businesses,” Malachowsky said. “Making that happen works best with a public-private partnership. Universities bring teachers, researchers, and students. Industry and philanthropy bring technology, funding, and experience. Government helps bring institutions and stakeholders together. … Working together, we can give people the tools to do more.”

These partnerships do more than pool funding—though aggregated resources are beneficial: Public-private partnerships enable participants to share expensive computing platforms, large data sets, laboratory resources, specialized talent and development risk. That can make research possible at a scale no single university, agency or company could easily support on its own, while creating shared findings and tools that benefit the broader scientific community.

How SHI is bringing the AI factory to reality

While the state is a “broker” of sorts that bring parties together in the NJSCI, SHI serves as the strategic architect and integrator for the initiative SHI has designed the architecture that houses NJSCI and established the technology requirements to bring the initiative to reality.

SHI has also worked to bring stakeholders together to realize the vision of the project—at a feasible pace. As Stumpf notes, this is no small feat. Stumpf noted that the planning process for the project accelerated in August 2026, when several New Jersey-based universities and institutions came together to discuss their requirements, agree to a phased approach, and develop consensus on what was achievable in the first phase, which will be operational by June 2027, Stumpf estimated. SHI also helped map out milestones for subsequent phases, such as key security provisions.

“What we were able to accomplish on August 11, [2026], was creating that funnel of getting everyone moving in lockstep,” Stumpf recalled. “We’re not trying to make it perfect on day one. We are trying to get it off the ground and it will evolve,” he said.

In August 2026, SHI created a workshop to establish stakeholders’ aspirational vision and then translate that vision into workable phases. “We put everything on the table and grouped those [requirements] into what can be reasonably accomplished in phase one, deferring what is left to phases two and three,” recalled Stumpf. “We wanted phase one to deliver meaningful benefits, while demonstrating the successful operation and governance of this critical resource.”

Stumpf noted this approach was particularly important for the project given that it was “the first time the state is trying to centralize computing resources at this scale.”

Stumpf noted that the goal is ultimately to balance futuristic vision with a healthy dose of pragmatism. “You can always shoot for the moon. But in terms of the execution, you have to take into account the capabilities of the organization.”

Stay tuned as NJSCI continues to build its AI factory capabilities.

NEXT STEPS:

Join SHI at SC26, booth 2000

If you need help, speak to an SHI expert.