Nuclear fusion: The science behind the energy technology, explained

This article has been updated.

The buzz around nuclear fusion energy is continuing to build, moving science closer to reaching the 'holy grail' of generating a virtually inexhaustible supply of emission-free energy.

Per kilogram of fuel, fusion could produce nearly four million times more energy than coal or oil, the International Atomic Energy Agency says. It's easy to see why this could be a game-changer in the fight against climate change. Yet, even after nearly 70 years of research, some key technological hurdles remain.

After a series of breakthroughs in the energy output of existing reactors in recent years, China's Academy of Sciences announced that its nuclear fusion reactor, dubbed the "artificial sun", has overcome another major obstacle to scalable fusion energy. The Experimental Advanced Superconducting Tokamak (EAST) reactor maintained plasma stability – key to initiating and sustaining the fusion reaction – at extreme densities previously considered impossible.

China had entered the nuclear fusion race with an estimated annual investment of around $1.5 billion – almost double that allocated by the US government to research in 2024, according to reports in Nature and the Financial Times.

Alongside government investments, private funding has also mushroomed, growing to $10.6 billion between 2021 and 2025. The number of companies involved in fusion projects more than doubled from 23 to 53 in the same timeframe. The technology has attracted investors including tech giants Microsoft and Google and oil major Chevron, alongside billionaires Bill Gates and Jeff Bezos.

Our current nuclear power stations use nuclear fission – essentially splitting an atom’s nucleus. Nuclear fusion, the process that powers the Sun and stars, merges two atomic nuclei into a larger one.

Fusion occurs when two light atoms bond together, or fuse, to make a heavier one. The total mass of the new atom is less than that of the two that formed it; the 'missing' mass is given off as energy, as described by Albert Einstein's famous E=mc2 equation.

Both reactions release large amounts of energy, but nuclear fusion yields much more energy per kilogram of fuel – four times as much as fission. There is also no risk of meltdowns or out-of-control chain reactions and no high-level, long-lived radioactive waste.

There are several 'recipes' for cooking up nuclear fusion that rely on different atomic combinations.

The most promising combination for power on Earth today is the fusion of a deuterium atom with a tritium one. The process, which requires temperatures of approximately 72 million degrees Fahrenheit (39 million degrees Celsius), produces 17.6 million electron volts of energy.

Deuterium is a promising ingredient because it is an isotope of hydrogen. In turn, hydrogen is a key part of water. A gallon of seawater (3.8 litres) could produce as much energy as 300 gallons (1,136 litres) of petrol.

While nuclear fusion power offers the prospect of an almost inexhaustible energy source for future generations, it also presents many scientific and engineering challenges that have been holding back progress.

Massive gravitational forces inside the sun create the right conditions for nuclear fusion in the star’s core, but on Earth, they are much harder to achieve.

Fusion fuel must be heated to extreme temperatures of approximately 50 million degrees Celsius, maintained under intense pressure, and sufficiently dense and confined for long enough to allow the nuclei to fuse.

Fusion happens within a super-heated plasma, most commonly confined in a doughnut-shaped or toroidal chamber, or tokamak, using strong magnets. Keeping this plasma stable over a prolonged period of time remains a challenge.

Google-owned AI company DeepMind is working with fusion startup Commonwealth Fusion Systems (CFS) and the École Polytechnique Fédérale de Lausanne to deploy AI to control plasma. To date, the collaboration has shown that deep reinforcement learning can help stabilize plasma in the tokamak. These learnings will now be applied to CFS's work, aiming to optimize fusion power and accelerate the route to commercializing nuclear fusion.

Meanwhile, AI chip giant NVIDIA and energy and defence company General Atomics, supported by a range of academic partners, are developing a digital twin model to simulate plasma behaviour at the DIII-D National Fusion Facility in San Diego. The goal is to stress-test the reactor virtually without causing real-world damage.

This level of AI control is a critical enabler for the "density-free" milestone achieved by EAST, providing the millisecond-level precision needed to maintain stability at extreme densities.

China's success at pushing the boundaries of fusion technology is only the latest in a series of breakthroughs that have moved fusion closer to delivering for the energy transition.

In addition to maintaining plasma stability, a major challenge for researchers is increasing fusion reactor output beyond the input energy required to initiate the reaction.

Since 2022, the National Ignition Facility (NIF) at the Lawrence Livermore National Laboratory in California has been pushing net energy gain from fusion from strength to strength. Initially achieving 3.15 megajoules (MJ) of fusion energy from 2.05 MJ of input energy delivered by lasers, the NIF team has since standardized the ignition process, achieving greater output. In April 2025, it delivered 8.6 MJ, more than four times the 2.08 MJ provided by the laser for ignition.

Similar progress has been made in Germany with the Wendelstein 7-X stellarator, which uses a slightly different technology from the more common tokamaks. In 2023, it achieved an energy conversion of 1.3 gigajoules, a record for this technology. Plasma was maintained for eight minutes. In May 2025, energy turnover was raised to 1.8 gigajoules.

Much hope rests on the completion of a new fusion reactor in France, ITER, a collaboration of 34 nations that will create the world's largest tokamak. The goal is to generate 500 megawatts (MW) of fusion power as the next step towards demonstrating net energy gain. However, following construction delays, research operations are now set to start in 2034, with the facility aiming for full-scale fusion reactions by 2039.

Alongside research facilities, private companies are ramping up efforts to make fusion a commercial reality, accelerating the energy transition.

Commonwealth Fusion Systems has raised nearly $3 billion to date for its experimental Sparc reactor, including investments from Nvidia and Google. The ultimate goal is to build a fusion power plant capable of generating 400 MW to power around 280,000 average US homes.

Pacific Fusion raised $900 million in Series A funding last year, while Helion – a startup backed by OpenAI's Sam Altman and SoftBank – has also received additional funding and begun construction of its plant slated to supply power to Microsoft's data centres from 2028. Germany's Marvel received a €113 million Series B investment, bringing total funding to €385m from private sources such as Siemens Energy Ventures and public investment from the European Innovation Council (EIC) Fund. It aims to commission a fusion power plant by the mid-2030s.

At Davos 2026, Francesco Sciortino, Co-Founder & CEO of Proxima Fusion, another competitor in the fusion race, said that the goal of all these companies is to develop "reactor concepts that are not just experiments but are trying to actually target energy production".

Public and private efforts are building on the earlier work of pioneering researchers, including those at the JET fusion lab. The Joint European Torus site in the UK, to give it its full name, was a collaboration of European nuclear scientists.

The JET lab, once the world’s largest and most advanced fusion reactor, began its pioneering experiments in 1983. Thanks to its success in advancing fusion technology, it operated for 40 years before entering decommissioning in October 2023. In its final experiment later that year, it achieved a record 69 MJ on just 0.2 milligrams of fuel.

Its successor will be ITER in France. Though the UK is not involved in ITER, the government committed £2.5 billion to fusion programmes last year, adding to previous investments. These include the experimental STEP (Spherical Tokamak for Energy Production) in Nottinghamshire.

At this year's Davos meeting, Proxima Fusion's Sciortino said: "Fusion is not just one technology. It is about understanding several enablers." These include computation, materials such as superconductors and powerful lasers, but also integrated designs and supply chains, he added.

Many of these are now beginning to converge, bringing fusion closer to a point at which it is no longer experimental but can deliver grid power. "But the value creation through intellectual property in fusion means we're creating maybe one of the most important industries of the new century."

"It's a very exciting ecosystem, and this ecosystem really relies on public-private partnerships to a significant extent. We need governments not to step back but engage even more. And working together is also part of how we can accelerate."

Kim Budil, Laboratory Director at Lawrence Livermore National Laboratory, where the NIF is based, added: "Historically, we've always said fusion energy is 30 years away from whatever day you ask and will always be that. And I think that's not true anymore."

"But I have to manage expectations. Fusion is hard. It's taken us a very long time to get to this point, and while there are many favourable things coming together, there's a lot of work to be done to realise this opportunity," she cautioned.

Like Sciortino, she stressed the importance of private-public partnerships.

"This is a moment for a different kind of operating model, where the private sector and the public sector need to work together in new ways to realize the opportunity."