GTC 2024, NVIDIA Officially Announces Four Quantum Breakthroughs
On March 18, NVIDIA® announced an innovative cloud service. The service is designed to help researchers and developers advance the exploration of quantum computing in key scientific areas such as chemistry, biology and materials science.
Tim Costa, head of NVIDIA's High Performance Computing and Quantum Computing Division, said the company's Quantum Cloud initiative will initially include a data center. This center will incorporate artificial intelligence chips and systems that work together to simulate a quantum computer. Unlike other cloud services, NVIDIA does not currently have its own quantum computer; however, the company plans to provide access to third-party quantum computers in the future.
NVIDIA CUDA-Q Integrates and Programs Quantum Processing Units (QPUs), GPUs and CPUs in One System
Official website link:
https://developer.nvidia.com/cuda-q
The NVIDIA Quantum Cloud is built on the company's open source CUDA-Q™ quantum computing platform, which has been adopted by three-quarters of companies deploying quantum processing units (QPUs). As a microservice, it enables users for the first time to build and test new quantum algorithms and applications in the cloud, including the availability of powerful simulators and hybrid quantum-classical programming tools.
NVIDIA® CUDA-Q allows hybrid code to be executed directly on all types of quantum processors, both analog and physical. Researchers can utilize the cuQuantum accelerated simulation back-end, as well as partner-provided QPUs, or connect their own simulators or quantum processors.
NVIDIA CUDA-Q significantly improves the execution speed of quantum algorithms compared to other quantum frameworks. Quantum algorithms can be up to 2,500 times faster compared to CPUs, and the number of quantum bits can be scaled by using multiple GPUs.
Typical QML workflow in CUDA-Q using a multi-threaded CPU with multiple NVIDIA A100 Tensor Core GPUs
Quantum Cloud has powerful features and a wealth of third-party software integrations to accelerate scientific exploration, including:
- A generative quantum feature solver, developed in collaboration with the University of Toronto, which utilizes Large Language Models (LLMs) to enable quantum computers to determine the ground state energies of molecules faster.
- Classiq's integration with CUDA-Q enables quantum researchers to generate large, complex quantum programs and perform in-depth analysis and execution of quantum circuits.
- QC Ware's Promethium is capable of solving complex quantum chemistry problems such as molecular simulations.
At GTC 2024, Tim Costa noted that NVIDIA's quantum simulator will be integrated into products from Alphabet's Google Cloud, Microsoft Azure and Oracle Cloud Infrastructure.
It will also be used by quantum companies such as IonQ, IQM Quantum Computers, OQC, Orca Computers, qBraid, Promethium and Quantinuum.
In fact, the artificial intelligence chipmaker has only recently entered the already fairly crowded cloud computing space.
Microsoft launched Azure Quantum in 2019, the same year Amazon Web Services launched Braket, while IBM-backed startup Strangeworks Inc. has had dozens of quantum computers on its cloud platform since it was founded in 2017.
The world's most powerful countries and companies have committed billions of dollars to quantum computing research, even though there aren't many practical use cases for quantum computers.
Global enthusiasm for this stems from the promise of computing speeds hundreds of millions of times faster than today's supercomputers, as well as concerns about the military and commercial consequences that laggards might face.
For much of the 2010s, NVIDIA was largely on the sidelines in this space. But with its graphics processing units, initially used to accelerate video games, showing the ability to perform certain types of calculations quickly, NVIDIA is now a key player in the field of quantum research.
Quantum computer simulations use similar computational methods, which makes NVIDIA's AI chips, such as the H100 Tensor Core GPU, an important part of quantum research. Costa mentioned that GPUs are now powerful enough to run classical algorithms that were previously too large for computers, such as those used in drug and materials discovery and other areas of science.
NVIDIA said its Quantum Cloud is supporting the entire quantum ecosystem and some of the most important research going on today. From quantum computing startups to some of the world's largest corporations, academic labs and supercomputing centers to Fortune 500 companies, they are committed to helping their partners develop and leverage quantum technologies.
The supercomputer, equipped with more than 2,000 Nvidia H100 Tensor Core GPUs interconnected via Nvidia Quantum-2 InfiniBand - a globally unique, fully offloadable in-network computing platform - is expected to be operational early next year.
In addition to the Nvidia H100s with Nvidia Quantum-2 InfiniBand interconnect, the ABCI-Q supercomputer is specifically designed to integrate with future quantum hardware.
The supercomputer follows the University of Tokyo's Artificial Intelligence Bridging Cloud Infrastructure (ABCI) and ABCI 2.0 systems. The 2.0 supercomputer achieved a peak Linpack performance of 54.34 petaflops in the latest Top500 ranking.
Tim Costa said, "Researchers need high-performance simulations to solve tough problems in quantum computing. cUDA-Q and Nvidia H100 enable pioneers like ABCI to make critical advances that accelerate the development of quantum-integrated supercomputing."
The company also revealed that the Gefion supercomputer at the Novo Nordisk Foundation in Denmark, which is dedicated to quantum computing research, and at the Supercomputing Research Center in Posey, Australia, will also use Nvidia hardware. Installation is being handled by Eviden, part of the Atos Group, and the system is expected to be ready for pilot projects by the end of this year.
At GTC24, NVIDIA also announced that it is partnering with more than 20 universities to develop quantum computing-related programs to accelerate the development of quantum computing talent.
The collaborative program, called NVIDIA CUDA-Q Academic, is designed to teach learners the fundamentals of quantum computing and programming skills. The program will include workshops, lectures, hands-on exercises, Jupyter notebooks and CUDA-Q training. In addition, NVIDIA further supports this educational initiative by sponsoring events such as hackathons, such as this year's QHack in February.
