Quantum computing is a path to energy-efficient AI
Three ways annealing quantum computing could head off AI’s escalating computational crunch
Dr. Alan Baratz
The AI
boom is driving an explosive surge in computational demands and
reshaping the landscape of technology, infrastructure, and innovation.
One of the biggest barriers to widespread AI deployment today is access
to power. Some estimates suggest AI-driven data centers now consume more electricity
than entire nations. The World Economic Forum projects a doubling of
energy use by data centers from 2024 to 2027, driven by the
energy-intensive nature of AI workloads.
This
surge in electricity demand is transforming the utilities industry and
redefining how and where data centers are built—power is no longer a
given. In the U.S, electricity usage is growing for the first time in over a decade largely because of data center consumption. Meanwhile, big tech is even turning to nuclear power to fuel their long-term AI strategy, while data center builders
are searching for land parcels in areas with excess power or resorting
to building their own power infrastructure, often relying on natural gas
generators.
ENTER QUANTUM COMPUTING
Quantum
computers could be the key to reducing AI’s rising energy consumption,
offering a more efficient, scalable solution. Unlike traditional
computers that evaluate one possibility at a time, quantum computers are
designed to explore complex problem landscapes more efficiently, making
them well-suited for tackling certain challenges that can be difficult,
time-consuming, or costly for classical systems. This enables them to
potentially provide solutions faster, at higher quality, and with
greater efficiency. While AI excels at uncovering patterns and
predictions, quantum computing identifies the most efficient solutions,
making these two powerful technologies complementary. Quantum computers
address problems that AI and classical methods struggle with, such as
factoring large numbers and solving hard optimization challenges like
vehicle routing and supply chain structuring.
Here are three ways quantum computing could help mitigate the expected disruptive impact of AI’s rising computational demands:
Optimize data center placement and utility grid management
Quantum
computing could be used to identify optimal data center locations based
on power availability or assist utility companies in streamlining grid
planning and management to support both consumer and data center needs. GE Vernova,
a global energy company, is using quantum computers today to identify
weaknesses in the power grid and optimize responses for potential
attacks on the grid. E.ON, a European multinational electric utility company, is now using annealing quantum computing to explore energy grid stability.
Unlock opportunities for greater energy efficiency
Early
research shows the potential for quantum computing to reduce the amount
of computational power needed to run AI workflows. A breakthrough published in Science
demonstrated that our D-Wave quantum computer solved a magnetic
materials simulation problem in minutes using just 12 kilowatts of
power. This task would have taken one of the world’s most powerful
exascale supercomputers, a massively parallel GPU system, nearly one
million years to solve, consuming more electricity than the world uses
annually. Applying these quantum computing techniques to blockchain
hashing and proof of work could also result in substantial enhancements
to security and efficiency, potentially reducing electricity costs by
up to a factor of 1,000. Quantum computers are very energy efficient and
may soon perform complex computations like those needed for blockchain
or AI at a fraction of the power required today.
Some of the
world’s largest supercomputing facilities are now actively exploring how
GPUs and quantum processing units could work together to improve
problem solving and reduce energy consumption. In February, Forschungszentrum Jülich,
a leading supercomputing center in Germany, purchased an annealing
quantum computer to integrate with the Jülich UNified Infrastructure for
Quantum computing (JUNIQ). This integration is expected to enable JUNIQ
to connect to the JUPITER exascale computer, potentially enabling
breakthroughs in AI and quantum optimization. JUPITER is anticipated to
surpass one quintillion calculations per second. This will likely be the
world’s first pairing of an annealing quantum computer with an exascale
supercomputer, providing a unique opportunity to observe the
technology’s impact on AI computational challenges.
Boost model efficiency and performance with quantum AI architectures
Early
evidence suggests that annealing quantum computers can be integrated
into quantum-hybrid AI workflows, which could potentially enhance model
efficiency and performance. Japan Tobacco’s
(JT) pharmaceutical division recently conducted a project that involved
using a quantum-hybrid AI workflow to generate new molecules. Using
this hybrid approach, JT enhanced the quality of its AI drug development
processes, demonstrating that the quantum AI workflow generated more
valid molecules with better drug-like qualities compared to classical
methods alone.
TRIUMF, Canada’s particle accelerator center, recently published a paper in npj quantum information
demonstrating the first use of annealing quantum computing and deep
generative AI to create novel simulation models for the next big upgrade
of CERN’s particle accelerator, the Large Hadron Collider—the world’s
largest particle accelerator. Traditional simulations of particle
collisions are time-consuming and costly, often running on
supercomputers for weeks or months. By merging quantum computing with
advanced AI, the team was able to perform complex simulations more
quickly, accurately and efficiently.
HOW TO ADDRESS AI’S POWER DRAIN WITH QUANTUM INNOVATION
As
AI adoption continues to accelerate, its insatiable demand for
computational power is upending industries and straining global power
resources. We need a better solution for addressing AI’s power demands
than simply adding more GPU clusters or building nuclear power plants.
From optimizing energy grids and data center placement to reducing GPU
power consumption and enhancing AI model performance, annealing quantum
computing offers a promising path forward. Tools like PyTorch plug-ins
are even making it easy for developers to incorporate quantum into AI
workflows to explore how the technology could address computational
challenges. For business leaders navigating the energy-intensive AI era,
adopting annealing quantum computing could unlock transformative
efficiencies today and tomorrow.
Alan Baratz, PhD is CEO of D-Wave.
ABOUT THE AUTHOR
(9)
