Why does the primary market need large-scale sectors that cannot be disproven for a decade?
In November 2025, the debate about the AI bubble reached a new climax. The fluctuating stock price of NVIDIA, the absence of killer applications, and the crazy expansion of data centers - every signal is reminding investors: After selling so many shovels, can we really strike gold?
However, this very anxiety actually proves that AI is far from being a bubble. When Cisco's price - earnings ratio soared to 200 times in 2000, no one used the price - earnings ratio to evaluate the Internet; when WeWork's valuation reached $47 billion, no one questioned when the shared office space would turn a profit. In a real bubble, investors don't care about commercialization at all, they are just selling the space of imagination.
To find such bubbles, you don't need to look far. While AI is causing global anxiety, two sectors are quietly attracting unprecedented capital: controlled nuclear fusion and quantum computing. Their common feature is that they push the moment of falsification ten years later, and you can't even find a clear metric.
"When society needs fusion technology, fusion can be achieved." In the 1950s, Soviet nuclear physicist Lev Artsimovich said this after developing the tokamak device. More than 70 years have passed, and the way this prediction is being realized is unexpected.
In the 1970s, the industry predicted that "there would be several commercial nuclear fusion reactors by 2000". In 2000, the timetable was pushed to 2030. The first plasma discharge time of the International Thermonuclear Experimental Reactor (ITER) was postponed from 2020 to 2025, and then to 2034, and the cost soared from 5 billion euros to over 20 billion euros. There is a sarcastic saying in the industry: "Controlled nuclear fusion is always 50 years away."
What exactly is a bubble?
From my observation, whether in the primary or secondary market, while everyone pretends to be sober and warns about the bubble, everyone actually needs the bubble. There are many bull stocks in the secondary market. Whether it's opening a new business in a hot area or having a restructuring of assets, the space of imagination will instantly expand.
Generally speaking, it goes like this. Rumor stage: The stock price starts to rise, and the space of imagination is infinite; Confirmation stage: The stock price continues to rise, and people say "this time it's real"; Plan announcement: The stock price rises first and then falls, and the details start to be scrutinized; Shareholders' meeting: The stock price stagnates, and funds start to diverge; Formal implementation: The stock price opens high and closes low, "The good news is realized, and the funds are fleeing."
It's better if you haven't commercialized yet, and it's better if it's not time to count the costs. When people really start to discuss whether it's worth it, it's probably not worth it. The economic calculation is very cruel.
The logic in the primary market is even purer. The fund's lifespan is 10 - 15 years. LPs are looking for "strategic layout" rather than quarterly returns, which means GPs need to find sectors that can tell a ten - year story and won't be falsified within three years. For consumer brands, the repurchase rate is evaluated in 18 months. For enterprise services, the ARR growth is evaluated in 3 years. For biomedicine, there are clear timetables for Phase I, II, and III of clinical trials. These are all too fast. Before the next phase of the fund is raised, the projects of the previous phase have already been judged as successful or failed.
In 2021, the financing in the nuclear fusion industry suddenly exploded. After the middle of that year, $2.8 billion was raised, almost equal to the total of all previous financings. By mid - 2023, the cumulative financing reached $6.2 billion, with private capital accounting for as high as 96%. Interestingly, this wave of capital frenzy occurred before a major technological breakthrough - the nuclear fusion ignition at the US NIF was only achieved in December 2022, while large - scale financing started in 2021.
This is not capital following technology, but capital actively looking for a suitable container.
What does "suitable" mean? There are three criteria: a big story, a long cycle, and slow verification.
The story has to be big enough to accommodate a fund scale in the tens of billions. "Optimizing supply chain efficiency" can't hold 5 billion RMB, but "solving the ultimate energy problem of humanity" can. The same goes for quantum computing. "Improving an algorithm by 10%" won't excite anyone, but "subverting the existing computing paradigm" can make LPs' eyes light up.
The cycle has to be long enough for the fund to reasonably exist for 10 - 15 years without being questioned. Internet projects show results in 5 years, and new consumer brands are judged in 3 years. But for nuclear fusion, you can say "we are optimizing plasma confinement parameters", and for quantum computing, you can say "we are breaking through the bottleneck of decoherence time" - these terms themselves filter the audience. Those who understand won't question, and those who don't understand dare not question.
The verification has to be slow, slow enough to have "milestones" to show every year, but commercialization is always "one step away".
In 2022, the NIF achieved successful ignition, with the output energy exceeding the input for the first time, and the Q value exceeding 1. This is a real milestone, and Science magazine called it "a historic moment in nuclear fusion research". Sounds like a breakthrough? Looking closer, the total energy consumed in this experiment was 200 times the output energy.
It's like spending $200 on lottery tickets, winning $1, and then announcing "I'm profitable". There is indeed technological progress, but it's still a long way from being able to open a lottery - selling business and make money.
For commercial power generation, the output needs to be at least 10 - 20 times the input, and a series of problems such as continuous operation, material wear and tear, and fuel recycling need to be solved. Optimistically estimated, it will take 15 years; conservatively, 30 years, and some even say 50 years - this time span itself is an art of ambiguity.
In November 2025, three experimental devices reported breakthroughs simultaneously: the US Zap Energy set a plasma pressure record of 1.6 GPa, the UK's Tokamak Energy achieved a fusion - level magnetic field of 11.8 Tesla, and the efficiency of the heavy - ion beam probe of Japan's LHD device increased by 2 - 3 times. Each piece represents real technological progress and can be written into the financing PPT.
China's BEST device is expected to generate electricity in 2027. It sounds like there are only two years left. But "generating electricity" itself is an elastic concept. Is it experimental power generation or continuous and stable power generation? Is it to verify physical principles or engineering feasibility? Even if it is successfully connected to the grid, it can be said that "this is an engineering prototype, and there is still a 10 - 15 - year distance from an economically viable commercial power plant".
Besides power generation, another important application scenario for controlled nuclear fusion is interstellar travel. A fusion - powered spaceship using the recoil of fusion products can fly from the Earth to Mars in only 90 days, saving more than half of the current time. Moreover, the proportion of the effective payload to the total weight of the rocket can jump from 1% to 50%, making it an excellent assistant for humanity to explore the universe. This can really be compared with Elon Musk's ideas.
The narrative structure of quantum computing is even more ingenious. It doesn't need to prove that "the sun is using it", but directly appeals to the fundamental revolution in mathematics - classical computers process 0 or 1, while quantum computers use superposition states to process 0 and 1 simultaneously, resulting in an exponential increase in computing power. This logical chain is so elegant that it's hard to question - would you say that quantum mechanics is wrong?
In 2019, Google announced "quantum supremacy" - its 53 - qubit Sycamore processor completed a calculation in 200 seconds that would take a classical supercomputer 10,000 years. IBM immediately refuted: "A classical supercomputer actually only needs 2.5 days."
This debate itself says a lot: When the definition of "success" can be reinterpreted, failure will never come.
But this doesn't prevent people from telling stories.
Every increase of 100 qubits is a "major breakthrough". Every 1% reduction in the error rate means "getting closer to practical application".
As for what "practical application" means, it could be cracking RSA encryption (still 10 years away), simulating drug molecules (still 15 years away), or building a general - purpose quantum computer (still 20 years away).
The ambiguity of the standard is an asset in itself - it allows every investor to find a timeline they believe in. Early - stage investors can exit at the C or D round, mid - stage investors can exit at the pre - IPO round, and late - stage investors can wait for the "first application". As long as the finish line is always ahead, everyone can get off at the stop they believe in, without waiting until the final accounting.
Actually, according to the views of many investors interviewed by the author, at least the commercial spaceflight and low - altitude economy are playing this game. But they are probably at the level of the metaverse and can be easily punctured. What can really be compared with nuclear fusion and quantum computing is probably synthetic biology and AGI.
Why does the primary market need a "reservoir"?
When people hear the word "reservoir", their first reaction is often a scam, thinking that capital is "cutting leeks". But this is a fundamental misunderstanding of the operating logic of the primary market.
The essence of the primary market is a market for investment and fundraising, not a market for exiting.
What do LPs want when they give money to GPs? It's not next year's dividends or the year - after - next's IRR, but a self - consistent story: how this money will appreciate in the next 10 years. The keyword here is "10 years". If the projects you invest in show results in 2 - 3 years, why would LPs lock their money in for 10 years? They could directly invest in the secondary market or buy convertible bonds, which have much better liquidity.
The time mismatch in the primary market naturally requires long - cycle projects to fill the gap.
Think about how a fund with a scale of 10 billion operates. The first 3 years are the investment period, and the money needs to be invested. But what to invest in? If you invest all in consumer brands, you'll know whether they can succeed in 18 months. Then what will you do in the 4th - 10th years? Just wait for the exit? LPs will ask: What have you been doing in these 7 years? Why don't you return the money to me? If you invest in enterprise services, you evaluate ARR in 3 years and profitability in 5 years. It's the same problem - why should they pay management fees in the middle years?
There must be some assets that can reasonably occupy the entire fund cycle.
This is not because GPs are lazy or greedy. It's a structural requirement. If a 10 - year fund liquidates all projects in the 6th year, what should it do in the remaining 4 years? Raise the next phase of the fund? Then the LPs of the previous phase will feel that their money has been idle.
Nuclear fusion and quantum computing solve this problem. They allow GPs to tell LPs: We are laying out for the technological revolution 15 - 20 years from now. Every penny we invest now is accumulating strategic value for the future. Is this "strategic value" real? Maybe yes, maybe no. But the key is that this statement can support the 10 - 15 - year lifespan of the fund, and there are "technological progress" to report to LPs every year.
Even better, these sectors have a natural "escape clause". If a nuclear fusion company fails, it can be attributed to "a problem with the technology route selection, but the industry direction is correct". If the number of qubits increases but the practicality doesn't improve, it can be explained as "we are solving fundamental physical problems, which takes time". Failures are individualized, and successes are collective. This is the risk structure that the primary market dreams of.
This is why mainstream VCs suddenly flocked to nuclear fusion after 2021.
It's not just because of the NIF ignition - that happened at the end of 2022; nor is it because the technology suddenly became mature - commercialization is still far away. It's because traditional hard - tech sectors such as new energy and semiconductors have entered the "accounting" stage. The prices of photovoltaic modules, the costs of lithium - ion batteries, and the yields of chip manufacturing processes are all too clear. It's so clear that it's difficult to tell a story when raising funds for a fund. LPs will ask: Is this sector a red - ocean market? Where will your excess returns come from?
The visible technological progress is to prove to LPs that "we are not burning money, but doing R & D".
The high - temperature superconducting magnet technology breakthrough in the SPARC project of Commonwealth Fusion are real engineering achievements that can be written into the annual report and used for roadshows. The distant commercial prospects are to avoid early valuation disputes. More importantly, these projects can be reasonably financed in multiple rounds.
Traditional project financing goes like this: angel round, A round, B round, C round. At each round, performance growth is evaluated, and a stagnant valuation is a signal. But nuclear fusion is different. It can be financed 10 times, and each time a different story can be told: the first few rounds are about technological breakthroughs, the middle rounds are about engineering prototypes, and the later rounds are about commercialization preparations. The valuation can increase at each round because "it's getting closer to the goal", but it's still far from real profitability at each round.
This financing rhythm is perfect for the primary market. Early - stage investors can exit at the C or D round without waiting for an IPO. Mid - stage investors can exit at the pre - IPO round without waiting for commercialization. In this whole chain, everyone can find a successor because "the story is still going on".
This is the real value of the "reservoir" for the primary market: it provides a time buffer.
Moreover, in this process, technology is indeed being advanced. The $6.2 billion in private capital allows more than 20 companies to explore different routes in parallel, which government funds can't achieve. SpaceX used private capital to reduce the launch cost by an order of magnitude. Nuclear fusion may be able to repeat this story. As long as there are enough attempts, someone will surely find the viable path.
Conclusion
In the 1990s, Mou Qizhong proposed the "blasting through the Himalayas" plan - using nuclear bombs to blast a 50 - kilometer - wide passage in the mountains to guide the warm and humid air from the Indian Ocean to transform the climate in northwestern China. Is this plan absurd? Of course. But it has all the characteristics of a "reservoir".
Mou Qizhong also proposed a "phased implementation" plan: In the first phase, blast a small opening to test the effect; in the second phase, expand the passage; in the third phase, continuously optimize it. Each phase can be financed, and each phase is "making progress", but each phase is still "a little bit" away from the final success.
The only problem is that he chose the wrong era. In the 1990s, the capital market in China was not mature enough to understand the value of "strategic losses". If it were today, this project could be packaged as the "Himalayan Climate Project", supported by a few papers in Nature and endorsed by a few academicians. With the policy narratives of "carbon neutrality" and "western development", it might really be able to raise funds.
When Mou Qizhong was imprisoned, people laughed at his arrogance. But 30 years later, when we see tens of billions of funds flowing into nuclear fusion, perhaps the question we should ask is not whether it's a scam, but why the capital market always needs a "Himalayas".