Quantum Computing Can Learn Sustainability Lessons from AI

One of the unintended consequences of the AI revolution has been a massive surge in energy demand, with the electricity usage of the world’s data centers forecast to double by 2026.

Quantum computing is waiting in the wings as the next transformational emerging technology. It is still early enough in its development to ensure sustainability is built in as it progresses towards commercialization.   

Quantum computers are expected to solve problems more efficiently than classical supercomputers, leading to energy savings and lower carbon emissions when they are integrated into data centers.  

Quantum computing startup Riverlane is on a mission to make quantum computers more useful sooner, notably through its Deltaflow quantum error correction stack, which it is building with energy efficiency as a core pillar.

Riverlane was recently accepted into the Quantum Energy Initiative (QEI), a global community of quantum technology companies and research organizations committed to better understanding the physical resource cost of quantum technologies.

Marco Ghibaudi is vice president of engineering at Riverlane. He worked on one of Europe’s first smart city testbeds in Pisa, received his doctorate degree at CERN and worked on one of the first commercial wireless virtual reality headsets.

Related:AI, Sustainability Converge in the Manufacturing Sector

In this Q&A, Ghibaudi elaborates on quantum computing’s potential to combat climate change by creating energy savings and the role of quantum error correction.

How does joining the Quantum Energy Initiative align with Riverlane’s goals?

Marco Ghibaudi: The QEI is a cross-industry group that aims to prevent the anticipated scaling-up problem in terms of the energy consumption of quantum computers. We know that this links quite strongly with error correction. We are expecting that as quantum computers scale in the number of qubits, the power consumption per qubit will reduce but the overall level of consumption might be prohibitive if not addressed properly.

One year ago, I was visiting IBM in the U.S. and we joked that you can always have a nuclear plant next to a computing facility to power everything. From a research perspective, power is no problem but coming into a data center they have to demonstrate energy efficiency every day. New facilities have to be designed considering energy efficiencies from day one.

What is the opportunity for quantum computing to ensure that energy efficiency is built in from the ground up?

There are two different answers but they are linked together.

The first one is if you look at quantum computing as a standalone problem in designing efficient systems, we have seen some of the missed opportunities that AI has not taken. Recent papers showed that by reducing the amount of data required for training you can get a 20% to 40% power reduction.

Related:Quantum Energy Initiative Adds New Member; Joins IBM, Microsoft

While it's not too late to change this it will require more effort and more financial investments than five years ago. Quantum computing needs to learn from AI in terms of engineering what would need to happen to make quantum computing as efficient as you could get.

But at the same time, there is an element in which AI could benefit and it's supposed to benefit from quantum computing in terms of power. Training for AI simulations has been the most expensive part as the system needs to be fed with a huge amount of data. While this will probably plateau in language or image recognition, applications that are like pharmaceutical, chemistry materials driven could use quantum computers to ask for some specific simulations, get the results back quantum simulation, get the result back and use those for the training of the models.

What it means is that instead of using that part that would be extremely demanding in terms of energy footprint, we are using instead optimized quantum computers that have an exponential speed-up in performing those types of computation and get the value in the system and the models are everyone then can just run interference on these models and get in see a lower footprint power wise overall.

How does Riverlane’s error correction stack support sustainability?

The quantum error correction stack is a required component in quantum computers from this era of research machines as they develop towards commercial ones. Even now on your laptop, there is something that corrects errors on your disk that is running continuously.

Quantum errors are more difficult to correct and they happen more often. You want something that keeps your computation alive and correct for as long as possible. This is an extremely challenging problem and if we were to approach it just from a “let's make it work, let’s go to the market as fast as possible, let's deliver the maximum level of performance you could imagine and cut costs,” perspective we would be missing the opportunity to design the system in a power-efficient way. You would pay the price later, as AI is paying the price now.

This is not something that we want. We have made one of the multiple components of the quantum error correction stack in a way that enabled us to build the capabilities internally to deliver low-power solutions. There is an engineering process and the opportunity to deliver a compelling solution but, at the same time, be ready to switch when we want to reduce the power footprint of our systems.