The "BYD" moment in commercial aerospace

Author: Zuoye

Since the law of universal gravitation, no invention has decisively shaped the destiny of civilization like the rocket, allowing humans to gaze at Armstrong and Buzz on the Moon, briefly becoming a multi-planet species, only to be abandoned before a new era begins.

The dilemma stems from the fading passion of the Cold War, and post-Cold War humanity lacks the courage to move toward the future.

Elon Musk and other right-wing Silicon Valley figures’ call for a “Tech Republic” is a retracing of decades of U.S. industrial policy guiding national engineering; bureaucrats’ reimagining after ideological failures, rebuilding the red imagery of national and commercial entities from new energy, artificial intelligence to commercial space.

The outcome of new energy is already decided, AI is in full combat, and commercial space eagerly seeks to become the new high ground.

Deconstructing this practice, “BYD” has driven the establishment of the supply chain, leading to extreme division of labor, resulting in localized overcapacity, then “Xiaomi” enters to drive a second growth curve, ultimately creating the DeepSeek miracle—non-consensus, counter-cyclical pure technological exploration.

When the world was young, humanity was full of desire for new frontiers; now, the ships of time have rounded the last cape of youth, and it’s time for the rocket capacity race.

The tail flame of rockets will burn away all ignorance.

Half a life of floating and reuse tides

Once spring ends, beauty ages; flowers fall, and people perish, both unaware.

The rocket industry belongs to all humanity. This is not human-centeredism at work, but the intertwining of scientific principles and engineering practice has always been so.

Newton’s mathematical principles of the universe, Tsiolkovsky’s chemical rocket equation based on them, Nazi engineer von Braun’s V2 illuminating the UK sky, American von Kármán’s division of V2 engineering, and Soviet Korolev witnessing the miracle of the V2’s physical reality.

Von Kármán’s Chinese students Qian Xuesen and Guo Yonghuai made important contributions to “supersonic Mach number,” laying the theoretical foundation for hypersonic and suborbital vehicles. After Qian returned to China as director of the Institute of Mechanics at CAS and head of the Fifth Academy of the Ministry of National Defense, he built the backbone of China’s space research and engineering.

Meanwhile, the U.S. absorbed von Braun as the core of its response to Soviet space efforts, Spunik became the morning star of Earth, Yuri Gagarin was a hero for all humanity—second only to the fish’s return to land in evolutionary history.

The roar of Saturn V was backed by NASA’s 4.5% of U.S. government expenditure. In 1962, Qian Xuesen wrote “Introduction to Interstellar Navigation,” envisioning engineering routes toward Alpha Centauri. But reusable rockets are merely the high mountains of an era with poor imagination; the Moon is the natural and ideal interplanetary station, and Europa/Io/Europa could serve as interstellar stations.

Let’s assemble reusable rockets, using 1960s technology routes. Don’t think this lowers the difficulty; it’s actually a high-end game. After the Moon landing, von Braun planned to use 1,000 Saturn Vs to reach Mars, powered by nuclear propulsion for reusable spacecraft.

Men are inherently proud of their strength; ascending to the sky and descending to the earth are always difficult.

Thrust forward, resistance backward; lift upward, gravity downward.

When thrust > resistance, movement forward; when lift > gravity (weight), flight upward. Human history, as we know it, is nothing but differences in work modes, but fundamentally, it’s all about the practice of mechanics.

Don’t be afraid, we won’t elaborate on Newtonian mechanics or Kármán’s formulas. Just remember two points:

Pressure difference is the fundamental driving force for sails, aircraft, and rockets, exemplified by the interstellar fast transport in “The Three-Body Problem”—light sail pressurization.

Pressure difference comes from the interplay of working fluid, structure, and ratio. Without knowing the linear solutions of chaotic systems, humans can only simulate through “alchemy”—trial and error.

Alchemy is truly manual parameter tuning. From wind tunnel tests of aircraft to asteroid exploration with “Tianwen-2,” all require an iterative process of “collect data—model analysis—conduct experiments.” This is fundamentally different from Einstein’s gravitational wave prediction—LIGO’s detection—meaning all human spacecraft are empirical products.

This is also the significance of SpaceX’s revival of reusable rockets: empirical products require continuous experimentation for improvement. But don’t forget Kármán’s chemical rocket formula: in a sense, it depicts the prospects of human interplanetary travel, at the cost of locking humanity’s path to the stars.

Before flying dreams, first define what’s achievable.

Image caption: Classification of orbits and spacecraft

Image source: @zuoyeweb3

Fleeting as mayflies in the universe, a speck in the vast ocean.

According to common orbit classifications, there are suborbital (below 100 km), LEO (160 km to 2000 km), MEO (2000 km to 35786 km), and geostationary orbit (GEO) at 35786 km.

GEO, as the name suggests, is synchronized with Earth’s rotation, appearing stationary from Earth, suitable for navigation satellites. For example, BeiDou has three satellites in GEO. MEO is relatively higher, covering larger areas, with BeiDou’s main constellation in this orbit.

In fact, the four major global navigation satellite systems—GPS (USA), BeiDou (China), GLONASS (Russia), and Galileo (Europe)—are all in MEO and GEO.

Below 2000 km, LEO satellites’ communication coverage is further limited. Countries’ constellations (Iridium, Starlink, OneWeb, StarNet, Qianfan) are competing for this resource. It’s estimated that LEO’s total capacity is about 60,000 satellites. Starlink already has about 10,000 in orbit, with plans for 42,000, leaving little time for China.

Higher orbits require fewer satellites for global coverage. Theoretically, three GEO satellites suffice, but for communication, GEO’s latency exceeds 500 ms, MEO’s over 27 ms, and LEO’s over 2 ms.

On January 2, SpaceX chose to lower Starlink’s altitude from 4400 to 480 km, not just for orbital safety but also to reduce latency.

However, high-orbit resources beyond LEO, especially Musk’s Mars exploration and settlement plans, will remain commercial fantasies for the next decade. Lacking commercial demand like Starlink, even contracts with the International Space Station are insufficient to cover Falcon 9’s costs, let alone Starship.

Without venturing into the vast universe, it’s hard to see our insignificance. Newton and Tsiolkovsky’s theories have already taken us the first step toward the stars, but unfortunately, it’s only the first step.

Since we are destined to be confined within the solar system, our engineers face two common problems:

How to increase propulsion speed—either increase specific impulse (thrust per unit propellant) or load more propellant;

How to reduce propulsion costs—optimize manufacturing under chemical rocket structures (reusable), or develop non-chemical rockets.

Gravity comes from the mass of objects; to accelerate, one can only enhance their own energy. This is the essence of Newton’s first and second cosmic velocities. Sadly, most commercial space efforts in the next 100 years won’t reach the third cosmic velocity; we will probably forever orbit the Sun.

In fact, both halves of these problems are impractical. Non-chemical rocket theories are feasible, but nuclear fission rockets risk orbital contamination that can’t be fully avoided. Fusion rockets still face the twin hurdles of practicality and miniaturization—The 50-year law still applies.

As for RTGs (radioisotope thermoelectric generators), electric propulsion, light sails, or even antimatter drives, they all face the issues of insufficient thrust or engineering challenges. If nuclear fusion becomes practical, remaining problems could be solved; conversely, if fusion remains elusive, better to dream of Orion’s nuclear pulse propulsion.

Restricting ourselves to chemical rockets, further excluding other propellants, the unexpanded Kármán formula shows that rocket propellant and thrust grow logarithmically. This means fuel mass must increase exponentially to achieve linear velocity gains. Usually, propellant accounts for 85%-95% of total rocket mass; beyond that, escaping Earth becomes impossible.

Thus, Musk’s envisioned system is “stainless steel, series of arrow bodies + LOX-methane (liquid hydrogen) + parallel engines + full reusability,” not just simple recoverability. The distinction is crucial.

Only by fully implementing each link can we achieve truly fully reusable rockets.

Qian Xuesen and von Braun both envisioned reusable rockets—or rather, they thought more broadly. In 1949, Qian Xuesen envisioned a spaceplane concept with vertical takeoff and glide landing at JPL. In 1962, he considered liquid fluorine propulsion and first-stage recovery. In 1969, von Braun envisioned a nuclear-powered shuttle + Saturn V reuse network. Nixon ultimately approved the space shuttle plan based on this blueprint, and China followed the Shenzhou route.

In 1981, Columbia’s first flight made it the first reusable spacecraft in human history. In 1993, McDonnell Douglas’s DC-X achieved vertical landing. In 1995, Apollo program director George Muller joined Kistler Aerospace to design the K-1 commercial reusable launcher.

Finally, in 2015, SpaceX’s Falcon 9 successfully landed on land, becoming the world’s first reusable orbital rocket. But note:

Not fully reusable: only first stage “reusable.” SpaceX’s truly fully reusable rocket is “Starship.”

Not stainless steel: still aluminum alloy body. SpaceX’s real stainless steel rocket is “Starship.”

Not methane: still LOX-kerosene. SpaceX’s real LOX-methane rocket is “Starship.”

Compared to methane (natural water) rockets, LOX-liquid hydrogen has higher specific impulse, but hydrogen storage is more difficult. Kerosene is easier to store but prone to coking; single-use discard, or full cleaning for reuse.

In SpaceX’s practice, versatility is pushed to the limit. Engines are only Merlin and Raptor; adding or reducing parallel units per mission suffices.

In fact, the Soviet N-1 rocket, parallel engine route, was developed during the same period as Saturn V, but due to inferior engineering, Musk ultimately took the crown of the parallel engine king.

Versatility can be further simplified: the first stage engine accounts for over 50% of total rocket cost. Achieving full reusability is very difficult; the most effective approach is to make the first stage recoverable and increase specific impulse, with thrust boosted by engine clustering.

Overall, the “reusable rocket” you see now, besides Musk’s Starship, is mostly “semi-reusable,” a more accurate term.

Image caption: Main parameters of commercial space engines

Image source: @zuoyeweb3

Most core first-stage engines of reusable rockets have a sea-level specific impulse of about 300 s, which is the passing line. The debate over methane versus kerosene/hydrogen is mainly about different engineering optimization paths. For example, Blue Arrow Aerospace is building a methane launch site in Jiuquan, similar to Musk’s visual approach at Tesla.

Besides, the most advanced is Blue Arrow’s Zhuque-3, using a stainless steel first stage + methane propulsion, with the second stage still aluminum alloy. Compared to SpaceX’s Falcon 9, with aluminum + kerosene, Zhuque-3 already shows a latecomer advantage.

Thus, fully reusable stainless steel chemical liquid hydrogen rockets can be simplified to first-stage reusable methane/kerosene rockets. Those capable of this can be considered to have entered the reusable rocket club.

But that’s not the whole story. To reach the stars, we must also win in the chaos of reality, initiating a complex game between Musk and national engineering, as well as the happiness and worries of Eastern peers.

Industry policy toward Silicon Valley

There is also a galaxy in the human world; a smile and a toast on the journey.

Since the founding of the nation, the U.S. has long implemented industrial policies and market access controls. It’s only since Reagan in the 1980s that free-market policies became the exception, leading to the stereotypical image of Silicon Valley tech elites and Wall Street financiers.

This is not the full truth. At least for the internet and commercial space, following the “state investment—laboratory development—civilianization” triad, the space field has been fully controlled by NASA from the start.

While American companies participate in lunar and other national projects, it’s clearly a buyer’s market, with all property rights and order allocations dictated by NASA.

Private companies initially participated in U.S. space industry, but it’s not accurate to say that the private commercial space industry started then. The industry was in the B2G (business-to-government) stage, very different from Starlink’s B2C communication services.

In moderation, from B2G to B2B, B2B2C, B2C, and future C2C, the active guidance of the U.S. government is inseparable, even a living fossil of American industrial policy.

Image caption: Subsidies for Musk’s companies

Image source: @washingtonpost

Even for Musk himself, his multiple industries grew gradually through subsidies, not relying on VCs or market demand. Tesla and SpaceX are precisely the recipients of subsidies.

In other words, Musk’s funds are converted into capacity growth, while Silicon Valley right-wing peers like Palantir and Anduril lack manufacturing capacity. Old industrial giants like Boeing and Lockheed are indeed decayed beyond salvation.

SpaceX is a product of American industrial policy and capital, ruthlessly replacing the “old space” of Boeing and Lockheed, and leading in the race against Blue Origin and Rocket Lab.

Meanwhile, we see that SpaceX’s real commercial scene is emerging—like Tesla’s entry into China, both as a “catfish and shark,” Musk strives to avoid binding with NASA, engaging with the U.S. military, aiming to build space Tesla purely through market mechanisms.

But the sensitivity of space and the complex U.S. political-business relations mean the U.S. government remains Musk’s largest single customer, whether through investment or restrictions. AT&T can’t avoid being split up, and Starlink can’t avoid being used.

Image caption: SpaceX’s long march

Image source: @zuoyeweb3

The forced arrival of B2B era.

In 1984, Reagan signed the Commercial Space Launch Act, responding to European and Chinese state-owned rockets seizing the civilian market, especially China’s Long March series, which at that time began to dominate about 10% of the market with “low-cost” launches.

The subsequent story is a lesson from the large-scale trial-and-error of America’s industrial and internet elites. For example, Microsoft co-founder Paul Allen sponsored Burt Rutan’s development of SpaceShipOne, which won the Ansari X-Prize in 2004—awarded to spacecraft that could cross the Kármán line twice in a week.

In fact, after the Space Shuttle’s second accident in 2003, the Bush administration signed the 2004 Commercial Space Launch Amendments Act, explicitly requiring NASA and other agencies to procure private launch services.

Looking back, Bezos’s Blue Origin and Musk’s SpaceX mostly emerged around 2000. Their appearance is a natural continuation of history.

U.S.-China industrial competition has always been a contest of national capacity in the commercial field. Whether in space or AI, it’s irrelevant. Great power competition has no retreat; the USSR will follow the Star Wars plan, and the U.S. will seize orbital resources.

Interaction between state and commercial entities has gradually transitioned commercial space into B2B2C.

In 1999, CIA established In-Q-Tel (IQT), a venture capital firm, following Silicon Valley trends, guiding commercial innovation to align with national interests. Its main member, Michael G. Griffin, not only accompanied Musk to buy missiles in Russia but also promoted the Commercial Orbital Transportation Services (COTS) program during his tenure as NASA administrator (2005–2009).

In 2023, after 21 years, SpaceX finally turned a profit with Starlink subscriptions. But 2008 was a life-and-death year; Peter Thiel’s Founders Fund invested $20 million, helping Musk sustain until the fourth successful test launch, ultimately securing NASA contracts.

A side note: IQT also invested $2 million in Peter Thiel’s Palantir in 2005, remaining its sole client for a long time, and helped evolve Palantir’s anti-fraud models from PayPal into intelligence surveillance analysis systems.

To date, Musk has secured over $10 billion in NASA orders, with the overall development cost of Starlink borne by U.S. venture capital and government.

Musk completed the final B2C commercial loop—the Starlink plan.

An interesting phenomenon: what’s called commercial space is actually satellite subscription industry, but its market penetration is far less than the stars and oceans. People fantasize about space travel around rocket flames, not about satellites orbiting Earth.

In fact, the lower the cost and the larger the capacity of commercial rockets, the smaller their proportion in the entire commercial space sector. This is why I deliberately skipped Musk’s prediction of $100/kg Starship costs—it’s not disbelief, but the potential for even lower costs.

But when reducing to 60,000 low-orbit satellites can’t meet capacity needs, rocket capacity will suddenly plunge into a brutal price war, and within five years, capacity shortages will turn into oversupply.

Take SpaceX as an example: its Starlink revenue exceeds $12 billion, while launch services are only about $3 billion. Yes, commercial space capacity has never been the main part of the space economy; the $20 billion in launch services accounts for only about 3–4%, with most revenue from satellite navigation, remote sensing, and communications.

SpaceX’s plan is to move toward the private market. In satellite navigation, remote sensing, and communications, navigation and remote sensing are long dominated by government, military, or B2G/B2B/B2B2C models. For example, AutoNavi (Gaode) involves BeiDou, ground stations, chip manufacturing, and subscription services—huge market share but complex profit chains.

Only the communications market, already validated by systems like Iridium, is ready for large-scale expansion, perfectly matching the demand for reusable rockets. Referencing the distribution of 4G/5G base stations, China holds 40%–60% of the market share. Starlink should be included in 6G discussions—an American overtaking maneuver.

Unlike China, after AT&T’s breakup, major telecom operators fell into low-quality internal competition, unable to meet stable communication needs in fringe areas. Starlink bypasses existing infrastructure and channels, essentially a B2C victory.

Currently, Starlink has about 850 active users, generating $12 billion annually. Musk also takes the most profitable satellite subscription revenue in commercial space. Falcon 9 launches are conducted every 2–3 days, continuously replenishing and networking, supporting a daily operation of 7,500 active satellites.

Meanwhile, competitors like Bezos’s Blue Origin, OneWeb, Google, and Microsoft have different visions for space, but their commercial loops are less complete than SpaceX’s. Especially after OneWeb shifted to Europe, falling into traditional profit-sharing models. SpaceX’s remaining rivals are only across the ocean.

Fragmented attacks on Musk

Once, the court was full of songs and dances.

Musk’s series of explosions to success began with finance.

SpaceX’s valuation at $1.5 trillion, dreams of Mars, reality of Starlink, promotion of Falcon 9. SpaceX, beyond delivery capacity, skillfully navigates between financial markets and real industry, driving civilian commercial space toward low Earth orbit constellations.

Eastern peers’ good news is that SpaceX has explored constellation models; national teams’ StarNet and Shanghai’s Qianfan constellation have huge real demand.

Bad news is they only have two years to sprint. LEO orbital resources follow a first-come, first-served model. China’s orbital resource application in 2020 will expire by 2027, so in 2025, StarNet might even deploy Long March 5 to secure a position.

By late 2025, Long March 12A and Zhuque-3 aim at “satellite internet technology test satellites,” with remarkably consistent results: first-stage recovery failed, second-stage achieved orbit. Both national and private teams face the reality of 2026.

Musk’s low-slow-small business: low orbit, small satellites, slow colonization.

Image caption: Musk’s affiliated companies

Image source: @theinformation

Musk is a highly capable project manager, with unique approaches in new energy, AI, commercial space, even solar and brain-computer interfaces, with interconnected commercial needs.

China’s model is driven by national demand, guiding private companies to emulate Musk’s attributes, achieving a balance of public and private, and preventing the rise of super-wealthy conglomerates, avoiding excessive dependence on the national economy.

BYD emulates Tesla; DeepSeek emulates Grok; LandSpace emulates SpaceX. Interestingly, LandSpace has its own low-orbit constellation plan.

Taking low Earth orbit as an example, the national team controls the overall demand for StarNet, while private aerospace companies provide capacity support for financing, capacity explosion, and IPOs. They can’t equate domestic commercial space companies with rocket companies, but in capacity shortage times, they command the highest premiums.

Just as commercial space isn’t the same as low-orbit constellations, within ten years, they won’t reach Mars or the Moon.

For satellite manufacturing, services, remote sensing, and even computing, we’ll leave detailed discussion for future articles. Currently, capacity is the bottleneck of all space economies.

For current commercial space (rocket) enterprises, the path of imitation and benchmarking SpaceX is very clear:

First, develop mature “small thrust” LOX-kerosene engines Merlin

Attach engines, achieve VTVL (vertical takeoff and landing) controllable testing, called “Grasshopper” by Musk

Achieve orbital launch capability, exemplified by Falcon 1, mainly for orbit validation

Based on these, develop the first fully reusable rocket Falcon 9, which is SpaceX’s main launcher

Repeat the above steps, develop larger LOX-methane engines Raptor, and a larger rocket—Starship, achieving full reusability

Of course, since the focus is on rocket capacity, omit the steps of SpaceX’s crewed Dragon; within ten years, orbital crewed flights won’t be the main commercial focus. Refer to Wang Chun’s $200 million, ten times more expensive than suborbital tourism.

As mentioned earlier, SpaceX and Blue Origin started around 2000, in sync with the internet privatization process. But unlike the internet, which quickly shifted to B2C or C2C after infrastructure completion, commercial rockets and satellites have long lacked physical layer independence.

This contrasts with the “disappearance of the encrypted physical layer”: commercial space’s persistence shows signs of absorbing the internet and AI. Space computing and satellite internet are booming, and after Ethereum’s shift to PoS, it can’t even become the internet’s economic layer—at most, SaaS for finance.

Under the narrative of an independent physical layer, China’s commercial space industry policy lagged behind the U.S. by about 30 years, starting roughly around 2014/15, reaching its first funding peak in 2018, with familiar companies like LandSpace and Tianbing mostly founded then.

After the official establishment of the “Commercial Space Office” by the China National Space Administration in 2025, coupled with rumors of a $1.5 trillion IPO for SpaceX and the 2027 low-orbit constellation orders, domestic commercial space (rockets) officially enters the elimination race.

Image caption: Progress of China’s main commercial space (rocket) projects

Image source: @zuoyeweb3

According to incomplete statistics, by 2026, at least over ten types of reusable rockets will be ready for launch. Besides Long March 12A representing the national team’s reusable route, only CASC’s China Rocket has a strong “national team” flavor, incubated by the Institute of Mechanics of CAS, a peculiar hybrid enterprise. As previously mentioned, the Institute of Mechanics is also where Qian Xuesen returned to work.

After detailed node segmentation of SpaceX, LandSpace is the closest private enterprise, even surpassing the national team. In a sense, LandSpace’s approach is a “two-in-one” of SpaceX’s three steps: directly adopting LOX-methane engines + stainless steel bodies + first-stage recovery, most similar to Falcon 9, even slightly ahead in engine technology.

Tianbing Technology’s Kang Yonglai participated in the development of the DF-17 hypersonic missile and Long March 11 rocket. Its Tianlong-3 rocket has experienced delays at the “test stand,” but its technical strength is top-tier. If all goes well, it’s close to Falcon 9’s level.

It’s worth noting that Dongfang Space, as a “solid first, then liquid” route representative, mainly focuses on maritime launches. The “Yinli-1” is currently the world’s largest solid launch vehicle, and “Yinli-2” is heading toward LOX-kerosene + recovery route. Solid rockets first occupy part of the market, then support liquid-fueled rockets.

Previously, I mentioned that domestic aerospace companies don’t need to consider crewed spacecraft or full reusability of Starship. Whoever builds the most successful domestic Falcon 9 first will gain market share comparable to Tesla Model 3, with at least 1,000 satellite launch orders annually.

In market tolerance, this is a rare scenario where private companies can compete fairly with the national team. Under the national will embodied by StarNet and Qianfan, no one dares to risk losing to the U.S. Meanwhile, using private rockets or national team rockets isn’t an issue.

In the blue ocean market, the national team and private enterprises have already deeply integrated. In the most critical engine products, a “state-led private” and “private-led state” dual approach has emerged:

State-led private:

Engines: YF-102v Rocket: Li Jian-2 and Zhihang-1

Engines: YF-102 Rocket: Yinli-2 (rumored)

Engines: YF-209 (LOX-methane) Rocket: Yueqian-1

Private-led state:

Engines: Jiuzhou Yunqian Longyun engine (LOX-methane) Rocket: CZ-12A

Star-class and machine-class rockets, the overall capacity gap prevents private and state-owned satellites from filling the network. This isn’t just anxiety in the communication satellite market; companies like Geely’s SpaceTime and Jilin Changguang remote sensing satellites are waiting anxiously for private rocket capacity to ramp up.

When capacity ramps up, capital enthusiasm will surge. Many companies, including LandSpace, are lining up for IPOs. Compared to the internet IPO wave in the U.S., aerospace companies can only choose between A-shares and Hong Kong stocks.

Compared to SpaceX’s $1.5 trillion valuation, with the combined imagination of rockets, capacity, Starlink, and Starship, domestic rocket companies’ valuations are at most 100 billion RMB. But we can imagine that after the 2026 elimination race, the valuation of commercial space will concentrate on a few enterprises.

Here’s a possible pathway: the commercial space wave propagates into China, similar to Tesla’s entry in 2019, China’s new energy industry exploding in 2021, and the aerospace sector may have 3–5 years to respond, leading to a Chinese version of SpaceX.

AI and deep space

Image caption: Heading toward Alpha Centauri

Image source: Unverifiable

The following text has little relation to commercial space, but after studying the history of semi-planetary spaceflight, I feel that human space capability today is actually inferior to the peak of the Cold War in the 1960s. At that time, Qian Xuesen was already designing routes toward Alpha Centauri.

In November 2025, the University of the Chinese Academy of Sciences established the Interstellar Navigation Institute, proactively cultivating interstellar travel talent within the solar system. After all, Mars is a planet, but human travel to the Oort cloud will remain in the realm of scientific research this century.

Chemical rockets give humans no future, but non-working fluid rockets might last forever—500 years. In this in-between era, we are destined to be the background of technological breakthroughs—the era of scientific stagnation.

At least, we still have artificial intelligence. If we can’t solve the chaotic linear solutions, AI might accelerate progress through simulation. Combustion, water flow, air—AI can optimize parameters efficiently. Especially for deep space, communication delays prevent humans from timely responses; AI could be a good helper.

We’ve experienced the “political space” of the Cold War era; now, in the post-Cold War era, it’s “commercial space.” But if we want to surpass LEO’s slow orbit and truly establish a long-term presence on Mars, space AI might be more real than fusion-powered space.

Here’s a brief summary of AI’s potential in space:

Design (turbulence, combustion, atmosphere, orbit calculations, micro-meteoroid and asteroid belt defense)

Manufacturing (thermal insulation, flares, high-energy radiation)

Guidance (mid-course correction, orientation adjustment)

Communication (filtering, compression, decoding)

Ecology (human monitoring, ecological cycles)

With OpenAI entering the hardware and rocket field, StarCloud is sending Nvidia H100 into space for computing platforms. The space race might extend from low Earth orbit to mid-high orbit computing power. Will it be Silicon Valley’s advantage or domestic private capacity overtaking? We’ll see signs within two years.

But regardless, turning human gaze back to space is a good thing. As Qian Xuesen said, “In science, anything that can be proven theoretically can always be realized in engineering.”

From political needs to commercial demands, the expected value-added space is gradually increasing: short-term near-Earth internet (energy, data centers, space stations, space factories), medium-term Moon, long-term Mars.

Beyond Musk’s first principles, we need to prioritize space in business, consider what scientific and engineering technologies are needed, and steadily advance within human commercial activities until reaching the goal—becoming an interstellar species.

Moreover, during the review, I found two long-held illusions:

Chinese only imitate; the U.S. has no industrial policy

Americans only do software; China has no commercial ecosystem

But from the development of commercial space, the U.S. SpaceX is entirely a product of industrial policy, while China’s constellation plans fully welcome private enterprise participation. Both sides highly agree on one point: having independent hardware networks is the fundamental cornerstone of self-reliance.

Additionally, to dispel stereotypes, a few wild ideas from Chinese engineers: Lingkong Tianxing is working on civilian hypersonics; weaponization is not the main goal, but aiming at suborbital tourism and ultra-high-speed point-to-point travel, like Concorde’s suborbital version.

There’s also Ceres-2, planning to try “electromagnetic launch” rockets. Since the first cosmic velocity is well understood, why not accelerate to break out of the atmosphere? After all, Qian Xuesen once said, “From an acceleration perspective, a single push is better than two stages.”

But note: chemical multistage rockets can achieve interplanetary travel within the solar system, called interstellar travel, with a fundamental difference from star-to-star travel. Even nuclear fusion rockets require millennia.

Whether a thousand or ten thousand years, they will inevitably be eroded and weathered by the long river of time, reaching an endpoint unseen by the world.

It’s like seeing a black stone fall, humans bowing to each other on the East African plains, witnessing migrations across icy peaks and ocean straits, the bursts and rebirths of powerful dynasties, finally settling on the tranquil blue dot and faint glow at the edge of the solar system.

May humanity, when dwelling on the satellites of Alpha Centauri, not forget to set out on a deeper, lonelier journey into the distant horizons.