Flywheel Energy Storage Prototype Begins Low-Speed Grid Testing

A spinning rotor goes up against lithium-ion

New York-based startup Qnetic has started low-speed testing of Pulsar, a full-scale flywheel energy storage prototype designed to hold 200 kWh. With this phase, the work moves past assembly and component checks and into testing the system as a whole2. The company's first goal is to confirm that the control systems, safety systems and power electronics work together under live conditions. Only after that will it push the rotor to higher speeds4.

The test follows Qnetic's August entry into the Electric Power Research Institute's deRISKED program, an independent review of pre-commercial storage technologies. The Sacramento Municipal Utility District (SMUD) is sponsoring Qnetic's evaluation11. Those two developments are why the story has been framed as a US test of a flywheel meant to store grid energy for decades. What matters more is the question behind it: can a mechanical battery capture part of the stationary storage market that lithium-ion chemistry, borrowed largely from the EV industry, now dominates?

I'll give my reading up front. The engineering milestones are real but early, and the independent review counts for more than any figure Qnetic has released so far.

What Qnetic has actually shown

The concept is simple. Electricity spins a rotor up to speed, the energy is held as rotation, and the same motor runs as a generator to turn that motion back into electricity when it's needed2. Qnetic's version uses composite materials, magnetic bearings and high-efficiency power electronics2. The rotor spins inside a vacuum chamber to cut friction3.

Qnetic reports two headline numbers. Before low-speed testing began, Pulsar ran 100 back-to-back charge-discharge cycles with no faults and no loss of capacity. In a separate test, the motor produced 130 kW at 97.3% efficiency at speeds above 11,000 rpm4. Mechanical assembly finished in June4. Starting this fall, Qnetic is sharing test data with EPRI, and EPRI will eventually present its findings to the utilities in the deRISKED program4.

The coverage agrees on these facts. It splits on how much they mean. Several outlets reported the 100-cycle run and the motor efficiency as signs of reliability36. A closer technical analysis was more cautious. It pointed out that 97.3% describes only the motor, not losses elsewhere in the system. It also noted that the full 200 kWh discharge, round-trip efficiency, standby losses and safety at high speed have not been verified yet7. The same analysis said Pulsar is running a temporary steel rotor for now. Because stored energy rises with the square of rotational speed, low-speed runs say little about the machine's final capacity7.

The skeptics have the better reading here. A hundred cycles at low speed shows that the subsystems can work together. It does not show that the machine lasts for decades, and Qnetic's own staged approach admits as much. CEO Michael Pratt has described a process of raising operating conditions gradually and fixing problems at each step before full-speed testing4.

Why independent validation is the real story

The EPRI program may matter more than the hardware test. Utilities are cautious buyers, and deRISKED exists to judge pre-commercial storage technologies by one consistent standard using lab tests, field demonstrations and data analysis11. Qnetic will hand over its prototype records and a full set of pilot-system performance data11. EPRI's Taylor Kelly framed the program as a way for stakeholders to put emerging technologies in context with a common methodology11.

This matters because flywheel marketing relies on claims that can only be checked over many years. Qnetic says its systems will run for decades with little degradation and almost unlimited cycling4. Its own materials list a 30-year design life and 22,000 cycles without capacity fade, and they acknowledge that this target has so far been validated only at prototype scale, not measured in the field5. The company also projects a levelized storage cost of about $101/MWh in 2030, compared with about $164/MWh for lithium-ion8. That is a vendor projection, not an audited result.

Qnetic has another validation track as well. It has agreed to install and test prototypes at the federal lab campus formerly known as NREL. The tests there will match wind, solar and AI data center operating profiles and demonstrate energy time-shifting813.

The geography is more complicated than the "US tests" framing implies. Qnetic says it plans to build Pulsar in the US13. But its main validation site, described as the world's largest dedicated flywheel test cell, is a facility costing more than a million dollars at its Technology Center in Shanghai15. One outlet described the company as having headquarters in both the US and China13. The EPRI and SMUD involvement is clearly American. The physical testing is more international than the headlines suggest.

The EV battery connection

The pitch against lithium-ion draws directly on the weak points of EV-derived batteries. Qnetic presents its technology as storage free of the degradation, fire risk and critical-mineral dependence that come with conventional lithium-ion15. Its own comparison puts lithium-ion service life at 14–16 years and says those batteries usually cycle about once a day. It claims 30 years and unlimited multi-cycling for its flywheel8. The company also says the build uses no lithium or cobalt5.

Flywheels and vehicles have a long history together. More than a decade ago, the Department of Energy asked Oak Ridge National Laboratory to assess flywheels as high-power storage for hybrid vehicles. That report described a University of Texas flywheel that ran 112,000 charge-discharge cycles with no drop in performance14. It also noted a Flybrid system for Jaguar that supplied 60 kW for up to seven seconds14. Those were power applications lasting seconds. Electrochemistry won the vehicle market because cars need energy density, and the industrial scale that followed is what made lithium-ion the default for grid storage too.

For that reason, the claim that flywheels will "replace lithium-ion" deserves skepticism. A more realistic role is complementary, and even sympathetic coverage suggests flywheels and batteries may work best together, each covering the other's weaknesses19. Qnetic's own materials say its systems can complement or replace lithium-ion15.

A history of near misses

The toughest context comes from the technology's commercial record. One detailed critique argues that DOE funded flywheels for roughly 35 to 50 years, and that when the department later looked at what might matter for storage lasting 10 hours or more, flywheels were no longer on the short list16. The same analysis calculates that Beacon Power's Hazle plant cost about $53 million for 5 MWh, or roughly $10,600 per kWh of stored energy16. It also argues that Amber Kinetics' 32 kWh modules show a scale problem: about 3,125 units would be needed to match a 100 MWh battery project16.

On this point Pulsar is a meaningful change in approach. Beacon's Gen4 units were designed for 25 kWh each and grouped by the hundreds to build 20 MW regulation plants. Pulsar's 200 kWh design is eight times larger per unit, though that compares design values, not demonstrated capacity7. Qnetic says it plans to launch with a 500 kWh unit and follow with a 1,000 kWh unit that discharges over 4–12 hours at 85% round-trip efficiency13. If those units work as specified, they would address the module-size objection directly.

The longevity argument does have some field support. Amber Kinetics reports more than 1.6 million cumulative runtime hours and 2.01 GWh discharged across its fleet. Those hours are spread over many units, not decades on one machine5. A California Energy Commission project recorded more than 38,000 operating hours and over 880 full cycles across a fleet of more than 15 Amber units20.

The market context

Two trends are improving the case for flywheels. The first is demand from AI data centers. Qnetic says its response time is under 3 milliseconds, fast enough for the sharp load swings of AI computing5, and data centers are among its target markets4. The second is competition from China. Boding Energy Storage has built what it calls the world's first 100 MW-class standalone flywheel frequency-regulation plant in Rushan. Phase I connected to the grid in July 202510.

On the commercial side, Qnetic cites $110 million in signed but non-binding letters of intent covering more than 900 units5. It is aiming for pilots in 202715.

The verdict

Pulsar has entered testing, which is a long way from proving decades of grid service. The 100-cycle run and the motor efficiency figure are encouraging engineering checkpoints. The key numbers, including full-capacity discharge, round-trip efficiency, standby losses and high-speed safety, are still unmeasured7. The development worth watching is Qnetic's decision to put its data through EPRI's independent process11. Flywheel companies have historically struggled to close exactly that credibility gap. If EPRI's findings support the claims at full scale, multi-hour mechanical storage gets a real hearing from utilities. If they don't, Pulsar joins a long line of well-engineered machines that never reached the grid at scale.