Can quantum take the load off AI's power problem? - InformationWeek

Machine Learning & AIIT InfrastructureEnergyData CentersIndustry TrendsCan quantum take the load off AI's power problem?Can quantum chips help ease AI's growing power problem? Here's what's plausible, what's hype and what a CIO should watch.Pam Baker,Contributing WriterAugust 7, 20267 Min ReadGetty ImagesPicture this: Giant AI data centers keep multiplying, pushing already strained power grids into a full-blown energy crisis. The strain on the grid forces Big Tech to slow the pace of new AI data center construction. But some quantum computing researchers are saying there is another option. They argue that the hybrid quantum-classical approach offers a narrow but real escape valve by offloading the most energy-brutal computations to processors that barely sip power.Is that a plausible answer to the power problem, or is it just hype?Rethinking approaches to solving the IT energy problemThe energy issue isn't a future problem. It exists now — and not solely because of new mega-sized AI data centers, but because of record heat waves and other factors. A recent Reuters report cited a regional U.S. grid operator warning it could order rolling blackouts for 14 states "because the power system was on the brink of an electricity shortfall due to record demand."Related:Quantum computers edge toward industrializationAI is expected to intensify those pressures.

The International Energy Agency (IEA) projects that global data-center electricity consumption could nearly double by 2030 to around 945 terawatt-hours (TWh), slightly more than the entire electricity consumption of Japan today. Much of that growth is expected to come from AI-optimized data centers, whose electricity demand is projected to more than quadruple by 2030, according to that same IEA report. In the U.S., that growth is already raising the alarm about whether the electric grid can keep up. The U.S. Department of Energy estimates that data centers could consume up to 12% of all U.S. electricity by 2028.The AI bottleneck is shifting from chips to power supplies."For CIOs, this means higher compute costs, longer deployment timelines and AI infrastructure increasingly being built where electricity is available, rather than where customers are located," said Albert Liu, Ph.D., founder and CEO of Kneron, a provider of edge AI products.Nuclear power, natural gas, renewables and small modular reactors may all help meet rising energy demands eventually. But if AI's computing requirements continue to grow, expanding electricity supplies alone may not solve the problem. Finding more energy sources now could lead to a costly cycle: bigger models drive bigger data centers, which consume more power and demand still more infrastructure, Liu warned."I think the industry has spent too much time asking, 'How do we produce more electricity?' and not enough time asking, 'Why are we using so much compute in the first place?'" Liu said.Related:Quantum computing faces security, skills shortage problemThat raises a different question: Instead of producing more electricity, could reducing the computing needed to power AI become a significant part of the solution to the looming power crunch?Shrinking the compute to shrink energy demandThe idea that quantum can take the energy load off AI rests on two concepts: shrinking the time it takes to complete complex tasks and reducing the amount of power needed to do so."It boils down to the fact that quantum computers can manage incredibly complex variables compared to classical systems, while requiring significantly less power," explained Pranav Gokhale, CTO and co-founder of Infleqtion, a global provider of neutral-atom quantum technology and related products.Unfortunately, swapping out computing modes isn't that simple."The capabilities of quantum computers are well understood. The hype is treating "energy-intensive computation" like one big category that quantum can just take over. It can't. The overstatement is saying this energy advantage is ready today for any compute-heavy job," said Marta Estarellas, CEO at Qilimanjaro Quantum Tech, a builder of full-stack, analog quantum computers.But even the reduced-energy claim from faster computing comes with an important caveat: Quantum computing requires more than just efficient processors. Factors such as refrigeration to cool quantum systems to absolute zero can dramatically change the overall energy equation.Related:Architecting a quantum node economyModeling of quantum data centers by the National Renewable Energy Lab found that cooling energy consumes far more energy than computation itself."In other words, the processor itself barely sips power, but the system around it does not. As such, [the costs] even out in a way," said Arif Gasilov, a partner at the consultancy Gasilov Group.How much energy can quantum really save?Does that mean quantum and hybrid computing does or doesn't have a place in energy and sustainability planning?"Trying to figure out how much energy savings we'll see in the future can be tricky," said Konstantinos Karagiannis, senior director of quantum computing services at Protiviti, a global consulting firm. "We don't know how the energy demands of the classical stack and the [quantum] refrigeration infrastructure will scale as qubit counts increase in the future."But there are other techniques that use laser control systems at room temperature rather than extreme cooling, according to Scott Buchholz, quantum computing leader at Deloitte.The laser power requirements in those systems dictate the total energy usage of the quantum computers. "In either case, energy usage is largely capped and comparable to or less than a modern high-performance rack of computers," Buchholz said.What does this all mean for CIOs?Throughout computing history, efficiency has come from specialized architectures such as GPUs for parallel computing and NPUs for AI inference."CIOs should think about quantum the same way enterprises thought about GPUs 15 years ago. At first, GPUs looked like a niche accelerator. Today, they are foundational. Quantum may eventually follow a similar path, but only for certain classes of problems," Liu said.That doesn't mean CIOs should expect quantum to replace classical computing. Instead, they should look for where quantum computing begins to complement it.But there's more for CIOs to worry about than quantum itself. AI data centers are heading for a stall in the face of worsening energy shortages, while AI workloads continue pushing against the limits of traditional computing."Even today's industry-standard optimization tools, which have saved billions of dollars, are beginning to reach their computational limits," Gokhale said.Energy planning becomes an IT issue"To be clear, quantum computers won't solve every energy problem confronting AI data centers today, but they undoubtedly offer a favorable path ahead for important computational tasks," Gokhale said.Computing in general is expected to cost much more in the foreseeable future due to a clash between demand and supply.Gasilov said the gap between rising demand and limited grid capacity has "a direct financial consequence for IT budgets" because it exposes organizations to swings in global energy prices. To address this risk, he said CIOs can "incorporate energy procurement risk into continuity planning the same way they model supply chain disruption." Not doing so can have fiscal consequences, he added.Beyond quantum, easing the power crunchMeanwhile, keep an eye on current efforts to reduce grid strain and to increase energy supply. There are a variety of approaches currently underway to address the energy needs of AI data centers, according to Pradeep Tagare, head of investments at National Grid Partners, the venture and innovation arm of National Grid, one of the world's largest investor-owned energy companies and operating in the U.S. and U.K.As examples, Tagare pointed to several technologies and companies National Grid Partners is investing in or following, including:Behind-the-meter generation, including gas, geothermal, nuclear and other sources.Real-time monitoring and analytics to unlock capacity on existing grids.Flexible AI data center designs that can adjust power demand.Next-generation electrical conductors to speed grid upgrades."We expect a mix of these over the next few years to solve the current issue," Tagare said.Not everyone agrees there's a power shortageThere's one other perspective to consider. Some say the energy demands of data centers may not be quite the problem it appears to be.Josh Wong, CEO of ThinkLabs AI, argued that the bigger problem is how utilities allocate and plan grid capacity.Hyperscalers are using a catch-all approach to data centers by "over-filing" with multiple utilities, he said, because today's "planning process is too slow and opaque" to determine where projects can be built most quickly."Filing permits in multiple places without the expectation of building is the problem plaguing our system," Wong said. He rejects the idea that the industry simply lacks enough power to support the AI boom.He added that CIOs can cut or control energy costs by using smaller task-specific AI models where appropriate."Training comparatively small, customized models to optimize the grid can take just 5 GPUs and 15 minutes, at a cost of as little as $5," Wong said, adding that this approach could be a game changer for companies looking to reduce AI's energy footprint.How is your organization preparing for rising AI compute costs and energy constraints? Share your strategy with us at [email protected].Read more about:Quantum ComputingAI managementIT Leadership and StrategyAbout the AuthorPam BakerContributing WriterA prolific writer and analyst, Pam Baker's published work appears in many leading publications. She's the author of several books, the most recent of which are "Agentic AI For Dummies," the second edition of "ChatGPT For Dummies" and "Generative AI For Dummies." She is an instructor for LinkedIn Learning on topics covering AI, IT, and innovation. Baker is also a popular speaker at technology and science conferences, and a member of the National Press Club, Women's Media Group, and the Internet Press Guild.See more from Pam BakerWant more InformationWeek stories in your Google search results?Add Us NowMore InsightsWebinarsAgentic AI: The Urgent OpportunityThe New Analytics Stack Is Converging.

Who Will Own It?The Future of ITSM: No-Code Automation Meets AIAgentic Analytics: From Legacy Architecture to Agentic Action on EDB Postgres AIBeyond the Hype: 2026 AI in ITSM Market Study Key FindingsMore WebinarsIndustry Reports2026 Cloud Security ReportBeyond Legacy Virtualization: Wind River's Modern Alternative for Enterprise InfrastructureThe Forrester Wave™: Unified Vulnerability Management Solutions, Q3 2025IDC MarketScape: Worldwide Exposure Management 2025 Vendor AssessmentInformationWeek 2025 US Tech and IT Salary ReportAccess More ResearchEditor's ChoiceAI InnovationsThe real heat behind OpenAI's new Jalapeño chipThe real heat behind OpenAI's new Jalapeño chipCybersecurityPoor UX undermines security policies, says Texas A&M University System CIOPoor UX undermines security policies, says Texas A&M University System CIOCloud outages, cybersecurity, agentic AI & federal policyGet Award-Worthy InsightsWant more InformationWeek stories in your Google search results?Get a snapshot of the issues affecting CIOs, three times a week in your inbox.Subscribe to our newsletters today.SIGN UP