Source: Tencent Technology

A capital and technology race centered on nuclear fusion is redefining humanity's imaginative boundaries and narrative approaches regarding the energy revolution.

The most direct signal comes from the capital market: nuclear fusion concept stocks are continuously gaining momentum globally. In the United States, venture capital is pouring into nuclear fusion startups; in China, concept stocks in the industrial chain have been strengthening and rising consecutively since the start of 2026; even relevant concept stocks in Japan and Europe have begun to show unusual movements. Real money is flooding into this track at an unprecedented rate.

This high-stakes race features a 'star-studded' lineup. Its participants form a rather fantastical mix: scientists, entrepreneurs, tech billionaires, and even former U.S. President Trump.

December 2025,$Trump Media & Technology (DJT.US)$In December 2025, a high-profile announcement was made about partnering with U.S. nuclear fusion companies; in Silicon Valley, tech giants like Sam Altman, Bill Gates, and Jeff Bezos have placed bets on different technological approaches: tokamaks, inertial confinement, magnetic mirror technology... Everyone is chasing the same dream: creating an 'artificial sun' on Earth.

Amidst this collective fervor, Elon Musk, one of the most radical 'gamblers' in the Silicon Valley tech scene, has chosen a distinctly different path.

"Solar energy is the only answer to humanity's energy freedom." In early January 2026, Musk reiterated his ambition for 'space-based solar power' during a public interview: launching AI-equipped solar satellites into space, where there is no night, no clouds, and sunlight can be absorbed 24 hours a day.

On one side, the world is pooling resources in hopes of igniting an 'artificial star' deep underground; on the other, efforts are reaching out to space to harness readily available sunlight. The contest over future energy pathways has thus evolved into a dramatic showdown between Elon Musk and the rest of the world.

Regardless of whose bet ultimately pays off, it will profoundly reshape the energy landscape of human history.

01. Nuclear Fusion Driven by AI

Nuclear fusion has lost twice before, first to physics and then to engineering challenges. In the 1970s, plasma proved difficult to contain, and net energy output remained elusive; later, in the 1990s, although breakthroughs were achieved with tokamaks, progress was hindered by material longevity, fuel cycles, and economic constraints, eventually leading to its decline due to low energy prices and waning policy support.

To date, nuclear fusion has once again entered a period of opportunity. The current opportunities are largely driven by demand. For Trump and Silicon Valley tech leaders, nuclear fusion is not merely a forward-looking strategic move but an important step based on intense international technological competition and geopolitical dynamics.

The most direct reason for this strategic focus is that the United States is facing an electricity shortage, which will worsen in the future, as power scarcity is becoming a new battleground in technological competition.

Over the past decade, growth in U.S. electricity consumption was primarily driven by residential use and traditional industrial sectors, with relatively moderate growth rates. However, starting from 2023, with the concentrated surge in emerging fields such as AI model training, hyperscale data centers, cloud computing infrastructure, semiconductor manufacturing, and encrypted computing, a clear turning point in electricity demand emerged. For the first time, electricity became a real bottleneck to tech expansion, and it became popular in Silicon Valley to say: 'Electricity is scarcer than chips.'

Taking the current large-scale model training as an example, while different institutions provide varying energy consumption figures, a consensus has been reached within the industry: training one of the most advanced large models consumes electricity at the 'terawatt-hour' level, equivalent to the annual electricity usage of several thousand American households.

One of the core reasons behind this surge in energy consumption is the extremely high power consumption of high-performance accelerator chips themselves. For instance, commonly used$NVIDIA (NVDA.US)$ H100 graphics cards have a single-card power consumption of nearly several hundred watts. A large-scale model training cluster often requires tens of thousands of graphics cards stacked together, and the annual electricity consumption of the entire cluster can be equivalent to that of a medium-sized city.

This trend has begun to exert tangible pressure on the U.S. power system. According to data from the U.S. Energy Information Administration and other organizations, AI data centers currently account for approximately 3% of the total electricity consumption in the United States. Industry experts widely predict that by around 2030, this figure may approach 8%. In some regions, the increase in electricity demand brought about by AI data centers has already significantly outpaced the growth in traditional industrial and residential electricity consumption.

Faced with such massive electricity demands, the U.S. government also feels the pressure. In the 'AI Infrastructure White Paper' published by the U.S. government in 2025, energy constraints were explicitly listed for the first time as one of the core bottlenecks restricting the further expansion of artificial intelligence. The document pointed out that AI competition is no longer limited to advanced chips and model architectures but has evolved into a comprehensive contest involving computational infrastructure and energy supply capabilities.

Additionally, there are more practical considerations stemming from geopolitics and industrial structure.

Modern energy systems heavily rely on oil and natural gas, both of which exhibit significant geographic unevenness in distribution. Their prices are subject to international relations, policies of oil-producing countries, and financial market fluctuations, leading to periodic and sharp volatility. For industrial systems, transportation systems, and data center infrastructures, this implies dual uncertainties in cost and security.

Against this backdrop, nuclear fusion has begun to regain the attention of engineers and investors.

Nuclear fusion boasts extensive fuel sources, high energy density, zero greenhouse gas emissions, low radiological burden, minimal land use, and strong sustainability. It is regarded as the most promising clean, efficient, and long-term energy solution for the future.

For AI, the most critical issues are unpredictable challenges such as power outages or electricity rationing. After all, data centers cannot wait for the sun to shine or pause training due to unstable wind conditions. The capability of nuclear fusion can be understood as its role as the "foundation" in energy supply, requiring no consideration of weather or day-night cycles, nor the grid’s constant adjustments between peak and off-peak hours.

Nuclear fusion holds a key hidden advantage: it does not rely on geopolitical factors. Unlike resources such as oil that are geographically constrained, its primary fuel is derived from seawater. Given the current international geopolitical conflicts, this implies fewer potential conflict risks.

In terms of safety, nuclear fusion does not exhibit the chain reaction characteristics of nuclear fission, thus avoiding catastrophic meltdowns. Once conditions deviate from requirements, the fusion reaction will automatically cease. While this does not mean zero risk, it indicates a lower psychological threshold for society, fewer regulatory nightmares, and a relatively lighter narrative burden for governments.

Most crucially, it leaves room for the future. The issue with AI's energy consumption is not about "cost," but rather the absence of an upper limit. Humanity can build more solar panels, increase energy storage capacity, and construct additional natural gas power plants, but these efforts compete with real-world constraints like resources, land, climate, and emissions.

Therefore, in a world where AI raises the baseline energy demand, nuclear fusion reemerges as a focal point for capital investment.

Section 02: Representative Approaches to Nuclear Fusion in Silicon Valley

To understand the current state of nuclear fusion technology, we must not view it merely as a singular "scientific challenge."

Over the past 50 years, the scientific community has demonstrated that nuclear fusion is physically achievable (plasma can reach the required temperature, and fusion reactions can occur). However, there remain several gaps before it can be used for power generation: whether it can operate continuously, whether costs can be reasonable, and whether it can integrate into the grid (providing stable output rather than laboratory pulses).

After all, fusion in laboratories often occurs in pulses lasting only a few seconds, tens of seconds, or even shorter, whereas power plants require an industrial system that operates continuously throughout the year, is maintainable, regulatable, and capable of grid integration. Precisely for this reason, the focus of today's discussions on nuclear fusion has shifted from 'whether it can light up' to 'whether it can be industrialized.' This point is highly significant.

When nuclear fusion transitions from scientific experimentation to an engineering system, a new reality emerges: there is no universally acknowledged 'correct path.' Instead, multiple distinct technical routes have formed globally, each based on different assumptions, engineering philosophies, and even timelines.

Take Silicon Valley as an example. In a high-uncertainty field like nuclear fusion, Silicon Valley capital exhibits a diversified investment structure: not betting on a single champion, not seeking short-term wins or losses, but hedging risks through multi-route allocation.

The first approach is pulsed magnetic confinement. Taking Helion Energy, invested in by Altman, as an example, it directly converts the energy generated by fusion into electricity via magnetic fields, bypassing intermediate steps.

To date, Helion has built six generations of working prototypes and is the first privately held nuclear fusion company to achieve a plasma temperature of 100 million degrees Celsius. Its seventh-generation prototype, Polaris, is currently under construction.$Microsoft (MSFT.US)$It has even signed a fusion power purchase agreement in advance, with a clear rationale behind it: AI's electricity consumption is becoming increasingly substantial, so securing a power source takes precedence.

The second approach is the hydrogen-boron route. Unlike Altman’s engineering pathway, Peter Thiel has placed his bet on a route more focused on fundamental physics and long-term stability. The logic of this route involves using FRC (Field-Reversed Configuration) plasma structures and neutral beam heating to make a nearly neutron-free high-temperature fusion reaction controllable, stable, and suitable for engineering.

The biggest advantage of this route is that it produces almost no neutrons that damage materials, meaning the reactor will have a long lifespan and low maintenance costs, making it suitable for long-term operation. However, the downside is that it is more difficult to achieve and therefore slower. To date, TAE, backed by Thiel, has operated five generations of devices, but these remain at the experimental prototype stage and have yet to generate actual electricity. The company plans to build a demonstration reactor by around 2030 and move toward commercialization.

Currently, the hydrogen-boron route, which Peter Thiel has invested in long-term, is primarily being advanced by TAE Technologies. It is supported not only by long-term investors $Alphabet-C (GOOG.US)$ ? $Goldman Sachs (GS.US)$ and$Chevron (CVX.US)$but has also entered into an all-stock merger deal with Trump Media & Technology Group, achieving a valuation exceeding $6 billion. Plans are underway to initiate site selection for constructing a commercial fusion power plant by 2026.

The third approach is the tokamak route. Commonwealth Fusion Systems (CFS), supported by Bill Gates, belongs to this category. Tokamak has been a well-researched path in the scientific community for decades, with abundant data, established theories, and regulatory foundations. However, previous devices were extremely large and costly. Therefore, CFS uses high-temperature superconducting magnets to reduce the size of the device. Currently, they are constructing the SPARC prototype, aiming to connect to the power grid in the 2030s.

Additionally, in January 2026, CFS announced a collaboration with NVIDIA and Siemens Energy to develop a digital twin system for the SPARC prototype reactor. This indicates that the tokamak route is moving beyond pure physical experiments and beginning to integrate industrial simulation, system integration, and operation and maintenance toolchains.

The fourth approach is the magnetized target route. For instance, General Fusion, backed by Jeff Bezos, focuses on reducing manufacturing complexity and costs, making it more akin to 'an industrial device capable of mass production.'

Other approaches, such as the laser inertial confinement method led by national laboratories, are also being explored, but these appear to be further from commercialization at present.

A Silicon Valley industry analyst stated: 'This multi-route parallel investment strategy essentially reflects Silicon Valley's rational response to extreme uncertainty. Nuclear fusion projects typically span over two decades, with an extremely high probability of technical failure, which almost entirely contradicts the traditional venture capital logic of a 7-to-10-year exit period. However, as the ultimate energy source and an ideal option, in Silicon Valley’s strategic view, whoever masters controllable nuclear fusion technology first will likely gain the upper hand in the post-fossil fuel era of technological competition. Thus, capital does not demand short-term results but instead seeks to secure long-term positioning.'

Section 03: Elon Musk, the 'Opponent,' Advocates Space-based Photovoltaics.

In the prolonged debate over nuclear fusion, Elon Musk has arguably been one of its most vocal critics. He has repeatedly mocked terrestrial nuclear fusion as a 'super dumb waste of resources,' reasoning in his characteristic style that humanity already has access to nuclear fusion through the Sun, which has been a stable fusion reactor operating for billions of years.

In his worldview, the issue is not whether nuclear fusion is 'advanced enough' but whether it can achieve a complete engineering loop and a controllable cost curve. This, he argues, is the core issue often deliberately avoided in most narratives about nuclear fusion.

Elon Musk has repeatedly emphasized his 'space-based photovoltaic' roadmap in various public forums: deploying large-scale 'Solar AI Satellites' to take advantage of nearly continuous sunlight in orbit, thus improving solar energy utilization. Musk stated plans to deploy approximately 100 gigawatts (GW) of Solar AI Satellites annually, equivalent to one-quarter of the entire U.S. power system.

The advantages of space-based photovoltaics stem from the unique environmental conditions. According to relevant science communication materials, outside the atmosphere, solar intensity can increase five to tenfold, and there are no day-night cycles or weather disruptions, enabling continuous power generation without the need for energy storage to maintain stable output.

In terms of orbital altitude, the differences in space-based photovoltaics are quite evident: Low Earth Orbit (LEO) satellites are exposed to sunlight approximately two-thirds of the time; Medium Earth Orbit (MEO) experiences even less shadow obstruction; and Geostationary Orbit (GEO) is almost continuously illuminated throughout the year, with only brief periods of shadowing near the spring and autumn equinoxes. In other words, the higher the orbit, the more continuous and stable the illumination, resulting in significantly longer power generation times compared to terrestrial photovoltaic systems.

Elon Musk has planned to launch solar-powered AI satellites into space, leveraging the advantage of 24-hour sunlight exposure in space to maximize the use of solar energy. It is expected that around 8,000 launches will be required to complete the deployment within a year. A solar panel array covering an area of about 100 square miles (approximately 259 square kilometers) would be sufficient to meet the entire electricity demand of the United States. Additionally, he plans to eventually relocate satellite production to the Moon, enabling local sourcing of materials and orbital deployment, thereby achieving larger-scale solar energy capture.

A key supporting factor behind this concept is the partial closed-loop conditions that Elon Musk has already established: SpaceX’s first-mover advantage allows it to dominate the development of space-based photovoltaics. SpaceX provides low-cost, reusable launch vehicles, making it no longer prohibitively expensive to send solar panels into orbit.

Of course, this does not mean that the photovoltaic route has reached its conclusion. Energy storage still faces three main practical bottlenecks: cost reduction, material innovation, and scaling up. However, within Elon Musk's framework, these issues are considered “iterative engineering challenges” rather than “unsolved physical problems.” This is why many Silicon Valley engineers tend to be more optimistic about Musk's “space-based photovoltaics” initiative: his approach already demonstrates a fully integrated closed-loop system.

04 The Significance of the High-Stakes Gamble: Even if They Fail, They Haven't Really Lost

Faced with numerous alternative approaches and controversies, as well as the rapid advancement of renewable energy, why are large amounts of capital in Silicon Valley still betting on the highly uncertain path of nuclear fusion?

In investing in nuclear fusion, Silicon Valley’s capital is clearly engaging in an “atypical” business venture. Once viewed from a long-term perspective, this choice itself challenges the fundamental assumptions of traditional venture capital: mainstream venture capital funds typically require an exit within 7–10 years, whereas nuclear fusion—from experimental validation to commercial power generation—is widely believed to require 20 years or more.

A technology investor noted, “One important reason Silicon Valley dares to invest this way lies in the fact that the R&D process for nuclear fusion inherently generates ‘byproducts.’ For instance, when fusion companies tackle complex problems, they engage with critical technologies such as ultra-high-temperature superconducting magnets, high-energy-density pulsed power supplies, precision plasma control, and advanced materials engineering. These technologies do not vanish if a particular fusion pathway fails; instead, they can be transferred to multiple high-end industries, including quantum computing, aerospace propulsion, precision manufacturing, and defense equipment.”

This logic of “technological spillover” has gained widespread acceptance in the industry. Michl Binderbauer, Chief Scientist at TAE Technologies, emphasized that the challenge of the hydrogen-boron approach does not lie in singular breakthroughs in physics but rather in long-term systems integration capabilities. In other words, whether or not a fusion pathway achieves power generation first is one matter, but the accumulated technological expertise can be transformed into actual productivity across multiple industries, ensuring that capital investment retains recovery potential even in a 'failure scenario.'

Moreover, the deeper rationale lies in achieving 'strategic control.' Within Silicon Valley’s cognitive framework, energy is not merely infrastructure—it is a foundational variable determining computational capacity, industrial boundaries, and the pace of technological expansion. Whoever can first master safe, scalable, and continually declining marginal cost clean energy will have the opportunity to secure a structural advantage in the next phase of technological competition.

Therefore, in such a highly uncertain technological environment, pursuing multiple approaches simultaneously has become the mainstream strategy in Silicon Valley. The simultaneous advancement of different methods such as pulsed magnetic confinement, hydrogen-boron fusion, tokamaks, magnetized target fusion, and laser inertial confinement is not due to investors being unsure of which to back, but rather an acknowledgment of the inherent unpredictability of technological evolution. Under this logic, failure is not considered an investment mistake, but rather a necessary cost of technological progress.

From this perspective, nuclear fusion is not a traditional energy business, but rather a strategic bet on the foundational capabilities of future energy, the scale of computing power, and the limits of technology. For Silicon Valley capital, the real risk is not investing too early, but choosing to sit out a competition that could reshape the landscape of energy and computing power.

Regardless of whether humanity ultimately succeeds in igniting "a sun on Earth," the contest surrounding energy pathways has already prompted a global rethinking of the future relationship between energy, computing power, and technological boundaries. The outcome of this race will profoundly influence the global technological landscape and the trajectory of civilization.

Editor/Rice

The translation is provided by third-party software.

The above content is for informational or educational purposes only and does not constitute any investment advice related to Futu. Although we strive to ensure the truthfulness, accuracy, and originality of all such content, we cannot guarantee it.