Qnetic white paper defines "AI-grade energy storage" as new category
Qnetic, a New York-based flywheel energy storage developer, has published a white paper arguing that neither battery energy storage systems (BESS) nor long-duration energy storage (LDES) are adequate for the power demands of AI data centres. The company is positioning a third category, which it calls "AI-grade energy storage", as a necessary addition to the infrastructure stack.
The white paper, released on 18 June 2026, is accompanied by an active crowdfunding round on the DealMaker platform, meaning the publication serves both an editorial and a promotional purpose. Readers should weigh that context when assessing the category definitions it advances.
The case against conventional storage
According to Qnetic, the core problem is that AI training and inference clusters produce load profiles that differ fundamentally from traditional grid demand. At gigawatt-scale facilities, the company says power ramp rates can exceed 1,000 MW per second, a figure that conventional battery chemistries were not designed to handle continuously. Lithium-ion cells are subject to capacity fade over repeated deep cycles, which Qnetic argues makes them poorly suited to infrastructure that may cycle continuously for decades.
The white paper identifies six requirements it says define AI-grade storage: millisecond response times, unlimited daily cycling without degradation, high peak-power output, multi-hour energy capacity, decades-long operational life, and intrinsic safety without thermal runaway risk.
Chief executive Michael Pratt framed the shift as categorical rather than incremental. "AI changes the problem," Pratt said. "Data centers now require storage resources that can respond in milliseconds, cycle continuously for decades, operate safely alongside critical digital infrastructure, and support increasingly dynamic power loads."
Qnetic's own flywheel systems store energy in high-speed composite rotors suspended on magnetic bearings inside vacuum enclosures. The company says the technology delivers multi-hour storage with no capacity fade and a rated 30-year operating life, characteristics it argues satisfy all six of its own criteria.
Market and regulatory context
The broader data centre power challenge is well documented. Global data centre electricity demand is widely projected to grow sharply through the end of this decade, driven primarily by AI workload expansion, and grid operators in high-concentration states including Virginia, Texas and California are already revising their capacity planning assumptions.
The energy storage market is correspondingly busy. Established players in grid-scale BESS include Tesla Energy, Fluence (a joint venture of Siemens Energy and AES), and a number of Korean battery manufacturers supplying turnkey systems. Long-duration storage developers, including Form Energy (iron-air) and Energy Vault (gravity-based), are targeting multi-day storage for renewable integration rather than the sub-second response requirements Qnetic emphasises. Flywheel systems are not new: Beacon Power operated grid-frequency-regulation flywheels in the US over a decade ago, and Amber Kinetics has continued development. What Qnetic is attempting is to reframe flywheels as an AI-infrastructure component rather than a grid ancillary service.
From a regulatory standpoint, the US Federal Energy Regulatory Commission's Order 841 (energy storage participation in organised markets) and ongoing NERC reliability standards both bear on how behind-the-meter storage at large data centres is classified and compensated. The EU's revised Renewable Energy Directive and the emerging AI Act's indirect power-consumption provisions create parallel pressures for European operators. Neither framework specifically mandates a particular storage chemistry, leaving the door open for flywheel deployments to compete on technical and commercial merit.
Editorial assessment
Qnetic's framing of a new category is a marketing construct as much as a technical one, and the white paper is not peer-reviewed. That said, the underlying engineering argument, that millisecond-response, high-cycle-rate storage deserves a distinct specification class from bulk energy shifting, is a legitimate point that grid planners and data centre operators are beginning to discuss seriously. Whether flywheels at scale can compete on cost-per-megawatt-hour with next-generation battery chemistries remains an open question the white paper does not address with independent benchmark data. Investors and infrastructure buyers should treat the category definition as a vendor position pending third-party validation.