Fusion Gains and Liquid-Air Storage Reshape Clean Energy Grids

The phrase "limitless clean energy" has been the carrot dangling in front of the nuclear fusion community for roughly seventy years, and for most of that time it has been more marketing than milestone. That is finally changing — but not in the way the breathless headlines suggest. The real story is two stories braided together: a genuine acceleration in fusion physics, and a quieter, arguably more consequential build-out of grid-scale storage that will determine whether any clean energy source, fusion included, can actually keep the lights on. The fusion experiments are winning the attention; the storage projects are winning the deployment race.

The Fusion Milestones Are Real, but Modest

The foundational moment remains the National Ignition Facility's December 2022 ignition shot at Lawrence Livermore National Laboratory, where 192 lasers converged on a hydrogen fuel pellet and, for the first time in a controlled setting, a fusion reaction gave back more energy than the laser delivered to it1719. Politicians called it a history-book moment, and the science press duly framed it as the dawn of "near-limitless" clean power17.

What the coverage agrees on, though, is that the margin was thin and the framing generous. The NIF's original result produced 3.15 megajoules from 2.05 megajoules of laser input — scientific breakeven, but nowhere near the energy ratios a power plant needs16. CBC's more sober assessment made the point bluntly: getting out slightly more energy than was put in isn't good enough for a commercial reactor, and the ratio needs to climb dramatically before fusion generates electricity anyone can buy20.

The encouraging news is that NIF has not stood still. By April 2025 it had pushed output to 8.6 megajoules — more than four times the 2.08 megajoules the ignition laser supplied — while standardizing a process that once seemed like a once-a-decade fluke1620. Reporting on the facility's trajectory through 2024 and 2025 describes repeated ignition shots and a shift in research focus toward making the reactions reliable and replicable rather than merely possible13. That is the correct metric to watch: repeatability is the difference between a physics demonstration and an engineering program.

Beyond Livermore, the fusion landscape has diversified in ways the 2022 coverage barely anticipated. Commonwealth Fusion Systems' compact SPARC tokamak, built on high-field magnets spun out of MIT, is chasing net energy gain in a far smaller footprint than traditional designs, with first plasma now expected in late 2026 after schedule slips13. TAE Technologies published peer-reviewed results in Nature Communications showing a streamlined method for forming and optimizing plasma in its Norm machine, which the company says cuts reactor complexity and cost and speeds its route to net energy15. In the UK, First Light Fusion claimed a first-of-its-kind demonstration of a plausible path to "high gain" inertial fusion — its projectile-based approach sidestepping lasers entirely — which its coverage frames as a stepping stone toward cheap, sustainable power1241.

The money follows the physics. Private investment in fusion reached $6.2 billion by the end of 2023 according to the Fusion Industry Association11, and later industry tracking puts cumulative funding at $14.24 billion by 202526. Google, meanwhile, has reportedly committed €411 million to a reactor design critics once considered impossible to build24. Even the defense sector has begun courting fusion startups as partners27 — a signal that serious money now believes the timeline, whatever the technical risk.

Where the Reporting Diverges

The divergence among sources is less about facts than about interpretation, and it maps cleanly onto enthusiasm versus institutional patience. Optimistic coverage points to commercial viability projections clustered around 2026 milestones and late-2030s deployment11. Institutional voices — the World Economic Forum's explainer among them — note that after nearly seventy years of research, key technological hurdles remain, even as they acknowledge NIF's steady gains and the Wendelstein 7-X stellarator's record eight-minute plasma sustainment in Germany16. The CBC piece is the most useful corrective in the set, emphasizing that a net-gain demonstration and a power plant are separated by decades of engineering, materials science, and tritium supply chains20.

There is also a quiet geography story. China has broken ground on the Helong-2 fusion platform, targeting first hydrogen-boron electricity by 203022, and is building enormous magnet infrastructure that its boosters tout as the key to unlimited energy21. Whether that timeline is credible or not, the competitive framing matters: fusion is now a geopolitical race, not just a scientific one, and races attract funding.

My own reading of the combined coverage: the physics genuinely turned a corner in 2022 and has kept moving, but "limitless" remains the wrong word. Fusion's fuel — deuterium extractable from seawater — is effectively inexhaustible, which is where the near-limitless framing earns its keep18. But delivered electricity in the 2030s, from any of these designs, would already be a triumph against historical baselines. Anyone promising limitless power on a 2026 grid is selling something.

The Unglamorous Breakthrough That Matters More

Here is the part the "limitless energy" framing obscures: the grid's problem right now is not generation. It is time. Solar and wind are already the cheapest new generation across most markets; what they lack is the ability to store afternoon sun for a windless February evening. That is a storage problem, and storage is where the actual gigawatts are being deployed.

The United States added a record 57.6 GWh of battery energy storage in 2025, a 30% jump from 2024 — enough, by one estimate, to power roughly 5.1 million homes65. Globally, 112 gigawatts of batteries were installed in 2025, ten times the amount added just four years earlier8. Through the first nine months of 2025, grid-scale battery systems totaling 49.4 GW / 136.5 GWh came online, a 36% year-over-year increase in gigawatt-hours9.

But nearly all of that is lithium-ion running two to four hours6 — enough to shave evening peaks, not enough to carry a grid through a multi-day wind lull. Long-duration energy storage crossed 15 GWh of global installations in 2025, a 49% year-on-year increase according to Wood Mackenzie, yet the sector is squeezed by declining venture capital and by lithium-ion manufacturers themselves extending into four-to-eight-hour products backed by supply chains exceeding 1,000 GWh of global capacity2.

The most interesting development here is the commercial maturation of liquid air energy storage. MIT researchers have modeled LAES as potentially the lowest-cost solution for reliable long-duration storage on a future carbon-free grid, noting that pumped hydro is geographically exhausted in the US and lithium-ion becomes prohibitively expensive beyond four hours7. Two large-scale projects — one in China's Gobi Desert near Golmud, and Highview Power's facility at Carrington near Manchester — are moving the technology from pilot to commercial operations, using surplus renewables to liquefy air at minus 196 degrees Celsius and expand it more than 750-fold to drive turbines when demand peaks10. Alongside them, Finland's Polar Night Energy is operating the world's first commercial sand battery, heating silica to 500–600°C and holding that heat for weeks with under 2% losses1.

Even AI is pulling the market. Surging data center demand has developers actively exploring long-duration storage as a way to guarantee clean power around the clock, though consultants warn most new LDES deployments may be delayed past 2030 thanks to supply chain rules and absent market mechanisms6.

Why the Two Stories Need Each Other

The COP29 Energy Storage and Grids Pledge commits nations to 1,500 GW of global storage capacity by 2030 — roughly six times 2022's installed base — and current deployment trajectories fall well short3. That shortfall is the context in which every fusion breakthrough should be read.

A fusion plant, if one arrives, would be baseload-like: it would produce power continuously, reducing — but not eliminating — the need for storage in a grid still dominated by intermittent renewables. In the nearer term, storage is what allows the renewables we already have to displace coal and gas. The two technologies are complements in a sequencing argument: storage is the bridge for the 2020s and 2030s; fusion, if the timelines hold, is the destination for mid-century.

The honest synthesis is this: fusion is progressing faster than skeptics expected and slower than its promoters claim, while grid-scale storage is progressing faster than almost anyone noticed and still not fast enough. When "limitless clean energy" finally means something, it will be because both of those sentences stopped being true — a grid with cheap multi-day storage, eventually anchored by fusion baseload. Until then, the smart money watches the unglamorous number: gigawatt-hours of long-duration storage actually delivered, year over year.