New York University (NYU) has announced the deployment of Torch, a supercomputer to power academic discovery in New York State.

New York University (NYU) has deployed Torch, a supercomputer that is the highest-performing academic computing system in New York State and ranks within the top tier of U.S. higher education institutions. Named after NYU's logo, the system expands computing capacity for the university's research community. Torch delivers an official 10.79 petaflops on the LINPACK benchmark, a fivefold increase over NYU's previous cluster, Greene. By bringing this capacity on-premises, NYU has created a local computing environment that gives faculty, researchers, and students broader access to high-end hardware.

**Accelerating Research Through Advanced Architecture**

Torch is designed for data-intensive workloads and multidimensional simulations and uses a parallelized, hybrid architecture. NYU worked with HPE and NVIDIA to build the infrastructure and scale it to the university's research operations.

Key specifications include:

- **AI engines:** Core workloads run on more than 500 NVIDIA H200 Tensor Core GPUs, which use HBM3e memory. The GPUs allow NYU researchers to train large language models (LLMs) and run deep learning tasks.

- **System integration:** HPE engineered, integrated, and deployed the HPE Cray XD670 cluster, drawing on its experience with academic and enterprise supercomputing to maximize operational throughput.

By deploying this hardware on campus, NYU research teams can shorten the time needed to turn theoretical models into scientific results.

**Focus on Sustainable Innovation**

Torch balances computational speed with environmental considerations. It can process more than 10 quadrillion calculations per second and is ranked No. 40 on the Green500 list of the world's most energy-efficient supercomputers.

The system uses liquid cooling that removes heat at the component level, which reduces the data center's reliance on traditional air conditioning. This energy profile supports NYU's institutional goal of achieving carbon neutrality by 2040.

NYU leadership said Torch balances academic computing power with energy efficiency. University officials said it will support research in artificial intelligence, climate modeling, and biomedical sciences, and will contribute to international collaborations such as the Global AI Frontier Lab.

**Putting the Fivefold Performance Gain in Context**

Greene, Torch's predecessor, provided approximately 4 quadrillion calculations per second. It was a significant asset for the university and initially ranked No. 50 on the Green500 list. However, the rapid growth of machine learning, quantum mechanics simulations, and neural network training required an upgrade.

Greene used an array of 332 NVIDIA GPUs. That configuration was well suited to general high-performance computing but faced operational limits when running modern large language models with hundreds of billions of parameters.

Torch changes that baseline. Its NVIDIA H200 Tensor Core GPUs handle matrix multiplications at speeds previously unavailable on campus. A high-throughput host processing architecture prevents data delivery delays, keeping the GPUs' execution pipelines continuously supplied with data. Overall, Torch can complete jobs that previously took weeks on the Greene cluster in a fraction of the time.

**Eliminating Bottlenecks in the Research Pipeline**

Time to results is a critical metric in academic computing. Universities without on-premises infrastructure have had to apply for compute time at national laboratories or rely on public cloud providers. Cloud alternatives can bring recurring operating costs, data egress fees, and privacy constraints that complicate institutional research grants.

With on-premises infrastructure, NYU departments can test models locally, debug network configurations immediately, and run continuous cycles without waiting in queues. This benefits fields including:

- **Artificial intelligence and embodied intelligence:** Training transformers and computer vision systems for real-world robotics applications.

- **Climate modeling:** Simulating localized weather patterns, ocean heat distribution, and long-term carbon variations at a granular scale.

- **Biomedical sciences:** Accelerating protein folding computations, gene sequencing analysis, and molecular dynamics modeling for Alzheimer's disease research.

**Engineering for Sustainable Infrastructure**

Energy consumption is a central concern for data centers that house top-tier AI hardware. NVIDIA H200 GPUs require significant power and produce substantial heat. Without changes in environmental engineering, a cluster of this size would need large, power-intensive traditional HVAC systems.

Torch addresses this through its environmental design. Its liquid cooling system distributes coolant directly to the processors and memory components. This removes heat more efficiently than air-based systems and lowers the facility's overall Power Usage Effectiveness (PUE) ratio.

This efficiency helped Torch earn its No. 40 ranking on the global Green500 index. The deployment supports NYU's effort to reduce its environmental footprint and shows that academic research capacity can scale while staying aligned with a carbon-neutral goal.

**Long-Term Institutional Impact**

Torch is intended to establish NYU as a center for computational science. In higher education, access to computing resources helps universities attract faculty, postdoctoral researchers, and graduate students.

The university also plans to integrate Torch into its academic curricula. Graduate and undergraduate students in data science, physics, engineering, and the digital humanities will receive direct training on high-performance computing clusters. This gives them hands-on experience with hardware configurations like those used in global research laboratories and advanced technology companies.

As part of its international strategy, Torch will serve as the foundational computing engine for the Global AI Frontier initiative. The initiative connects NYU's campuses and global partners to support multi-institutional research on global challenges. With Torch operational, NYU says it has the technical infrastructure needed to support data science developments over the coming decade.