For some, NVIDIA has almost become a meme stock for everything AI-related. The bell-wether of the Artificial Intelligence innovation we see, such as the rise of the LLM with tools such as ChatGPT, is making it into our everyday vocabulary. However, some might not be aware of the impact that the chip giant is having on the development of quantum computers and quantum computing.
NVIDIA has been busy developing its tools for developers who want to simulate quantum computers. Let’s be clear: NVIDIA does not create a quantum computer but instead helps developers of quantum devices simulate their devices. Eventually, the hope is that quantum computers will become powerful and cheap enough that their hardware can be explored to run quantum circuits, which offer a quantum advantage over traditional or classical algorithms.
Simulating a Quantum Computer with a GPU
Simulating quantum computers is something that GPU chips can do rather well. NVIDIA is considered the foremost leader in this technology, which it has pioneered for decades. The current lineup of NVIDIA chips includes the H100. There is a waiting list for some of its most powerful chips. the AI revolution sometimes appears to have developed upon the back of a single hardware company: NVIDIA.
The GPU started life-accelerating games, but researchers and developers have found that the chips perform specific calculations brilliantly. Until there is widespread quantum computing, we can keep using the power of GPUs to compute quantum circuits on conventional or non-quantum hardware. The drawbacks are, of course, that the purported quantum advantage is not present, but it does allow researchers and developers to explore quantum without the hardware available right now.
When more powerful quantum hardware shows up, we can exploit that hardware using quantum effects such as superposition and entanglement to generate an advantage over classical algorithms. Think of it as riding a bike with stabilizers. It will never be like riding a bike on two wheels with all the benefits of speed and efficiency, but it prepares the rider for what is to come. Deploying GPUs enables developers to get to grips with developing quantum workflows that can exploit quantum computers. They can toggle a switch from running on simulated quantum hardware to real quantum hardware.
As real quantum hardware accelerates in qubit count. A qubit is simply a quantum bit. The number of qubits roughly translates as a proxy for the power of quantum computers. Companies such as IBM, Google, Rigetti, and IonQ are all busy increasing their qubit count. IBM is busy executing its Quantum roadmap, which has reached over 1,000 qubits for its machines.
The point of a quantum computer is that it will be able to perform calculations that are simply impossible on a classical computer. As the number of qubits advances, the power of the conventional computer that simulates that device must also increase, and that has led to the GPU as the foundation for quantum simulation. The GPU market leader, NVIDIA, often powers that foundation.
Of course, other chip makers have GPUs. AMD and Intel have GPUs, but so far, they have not managed to maintain NVIDIA’s dominance. In part, this is likely due to the tools that NVIDIA has produced, which make interacting with standard software tools a relative breeze. As the de facto standard for GPU workloads, its CUDA tools have seen common integration with some of the leading machine learning model frameworks, such as Torch and TensorFlow, further solidifying the support for the US-based GPU giant.
From GPU to QPU
The GPU will be the precursor to the QPU (Quantum Processing Unit). The GPU is at the forefront of the AI revolution, enabling services such as chatGPT. Without the GPU, these services would not exist and likely would never have been created in the first place because of the sheer size of the networks that GPUs can help train the LLMs (Large Language Models) that are behind the wave of exciting new tools that promise to upend a variety of industries, from graphic design to copywriting to customer services.
We see the synergies around computation, whether that is the rise of the GPU, which has further been accelerated by the early promise of LLM models such as chatGPT. The technological improvements won’t stay siloed, and the powerful breakthroughs we see in GPUs will likely bleed over to other domains. That means that more and more developers will not be exploring the quantum and quantum-classical workflows, and they already have the tooling on their desks or in the cloud.
