For the first time in history! Nvidia's H100 chip has been successfully operated in space.

Almost simultaneously, SpaceX launched the largest IPO in history.

The ancient Chinese saying "Man's plans are no match for heaven's designs" has been given a new technological interpretation in the new era.

While the world is still competing for electricity and land for computing power on the ground, Musk's space empire has quietly opened the prologue to the "space computing era."

On Wednesday local time, CNN reported that the world's richest man, Elon Musk, is planning to create the largest initial public offering in history for SpaceX, raising $30 billion. If the IPO is successful, its valuation will reach $1.5 trillion, far exceeding OpenAI's current valuation of $500 billion.

Almost at the same time, CNBC reported that startup Starcloud launched a satellite equipped with Nvidia's H100 GPU chip last month, training an artificial intelligence model in space for the first time.

"Greetings to Earthlings, a charming collection of blue and green." This is the greeting to Earth from the AI model trained in space.

It is reported that Starcloud's satellite is in operation and can query and respond to Google's open-source large model Gemma, marking the first time that a large model has been run using a high-performance GPU in space.

Whether it is the trillion-dollar valuation IPO fundraising or the technical verification achieved by the first successful operation of the H100 in space, their core points consistently point to "computing power and energy" — the key engine of the AI era.

The World's Largest IPO: SpaceX Becomes the "Musk" of the Capital Market

"Based on the Starship's carrying capacity, 300GW to 500GW of solar-powered AI satellites can be launched into orbit every year." In mid-November last month, Musk posted this comment.

This means that Musk has long targeted space for his artificial intelligence energy supply strategy.

According to current information, Musk is actively promoting SpaceX's IPO, which is expected to raise more than $30 billion with an overall valuation of $1.5 trillion (equivalent to approximately 10.62 trillion yuan). The final plan is to list on the capital market in the mid-to-late 2026.

If SpaceX goes public as scheduled according to current market news, it will undoubtedly become a "talk of the town" like its founder Elon Musk, attracting strong market attention.

On the one hand, as an IPO company, SpaceX's $30 billion fundraising amount has broken the previous record of $29 billion set by oil giant Saudi Aramco. It not only refreshes the global IPO scale record but also shows the market's extreme recognition of the future value of commercial aerospace.

Moreover, in terms of SpaceX's target valuation, the company's valuation of $1.5 trillion after the IPO will not only be equal to Tesla's total market value but also far exceed OpenAI's valuation of $500 billion.

On the other hand, according to CNN, if the raised funds belong to the company, given that Musk holds nearly half of the company's equity, the fundraising income will also significantly increase his net worth.

Meanwhile, Bloomberg Billionaires Index data shows that Musk's current net worth is $461 billion, most of which comes from Tesla's stocks and options. If SpaceX's valuation reaches $1.5 trillion, Musk's net worth may double.

From the industrial perspective, SpaceX's IPO has more important strategic value.

Public information shows that SpaceX plans to use part of the funds raised from the IPO to develop space-based data centers, including purchasing chips needed to operate these facilities. By equipping high-performance chips to build orbital AI computing infrastructure, the in-depth integration of commercial aerospace and artificial intelligence will enter a new stage of development.

Currently, SpaceX's Starlink business has deployed more than 8,000 satellites, covering more than 10 million users, and its annual revenue accounts for 80% of the total revenue.

Sales are expected to be about $15 billion in 2025 and further increase to between $22 billion and $24 billion in 2026. Most importantly, relying on this core business, SpaceX has achieved positive free cash flow, moving from the "money-burning expansion" stage to the "self-sustaining" stage.

For the First Time in History! The "Space Computing Era" Begins

Starcloud launched the H100 GPU into space, marking the start of competition in the space computing power track.

It is understood that Starcloud, the company that executed this satellite launch, was founded in 2024 and is a space data center startup supported by Nvidia. Data shows that its core vision is to deploy data centers in low Earth orbit to address the energy and resource bottlenecks faced by the development of AI computing power on the ground.

In early November, Starcloud launched its first prototype satellite, Starcloud-1, via a SpaceX rocket. The satellite weighs about 60kg and is about the size of a small refrigerator. As mentioned earlier, its highlight is that the satellite is equipped with Nvidia's H100 GPU, which is said to have 100 times the computing power of previous space chips.

This time, the satellite trained the NanoGPT model using the complete works of Shakespeare in orbit and responded to Google's open-source large model Gemma, sending greetings to Earth from space. This move verified the feasibility of conducting high-performance AI computing in the space environment.

Philip Johnston, CEO of Starcloud, stated that all new data centers built in the next decade may be fully transferred to space.

Starcloud will also launch commercial services, performing inference through satellite images for applications such as maritime search and rescue and forest fire location. Subsequently, Starcloud will deploy Blackwell architecture chips, with the long-term goal of building an orbital data center with a power of 5 gigawatts and a span of about 4 kilometers.

As a startup, Starcloud is gaining momentum, but it is not the only player. In the competition for space computing power, U.S. tech giants have never given up this "lucrative opportunity."

On November 4th, Google announced "Project Suncatcher," planning to launch two satellites equipped with its self-developed TPU in 2027 to verify the feasibility of distributed machine learning tasks in the space environment.

On December 10th, media reported that Sam Altman, CEO of OpenAI, is quietly planning a deal that may disrupt the space industry pattern. He plans to acquire or take a controlling stake in a commercial rocket company through cooperative review (rumored to be Stoke Space) and may compete directly with Elon Musk's SpaceX.

In addition, U.S. commercial aerospace startup Lonestar Data Holdings is also committed to deploying the first commercial data center on the lunar surface.

Dion Harris, Senior Director of Nvidia AI Infrastructure, commented: "Starting from a small space data center, we have taken a big step towards a future where orbital computing utilizes the infinite energy of the sun."

Is Space Computing Power the Optimal Solution?

Taking GPT-3 as an example, its training requires thousands of GPU chips and takes several months; OpenAI invests as much as billions of dollars annually in training large models. Behind the high training costs is the sharp increase in computing power costs.

Nvidia H100

According to the seventh annual AI Index report released by Stanford HAI Institute, it reveals the current situation of the artificial intelligence industry. The report points out that over time, the training costs of models have increased sharply, especially the greater the computational training demand, the higher the training cost of the model. (As shown in the figure below)

In addition, according to the International Energy Agency's "Electricity 2024" report, global data center electricity consumption may exceed 1000 TWh by 2026 (about 114 GW of continuous load), of which the proportion of AI data centers is rising rapidly. Goldman Sachs also predicted in a report in April last year that by 2030, AI-related electricity demand will increase by 160%, and total data center electricity consumption may account for 8% of global electricity, exceeding the current proportion of 1% to 2%.

Space data centers can utilize all-weather high-intensity solar energy without being affected by day-night cycles, weather, or atmospheric losses. In space, there are no adverse factors such as atmospheric refraction and cloud cover, so the solar energy conversion rate has significant advantages compared to the ground. The extremely cold environment of about -270°C in deep space only requires the deployment of thermal conductive materials to complete heat dissipation, without consuming precious water resources.

Therefore, realizing the advantages of space computing power has become a breakthrough to solve the constraints of computing power and energy in the AI era.

At the 2025 Shanghai Bund Summit in September, Wang Jian, founder of Alibaba Cloud, put forward 9 assertions: "Space has always been our greatest resource. Today, AI should absolutely not only be used on mobile phones and computers; in fact, artificial intelligence should not be absent from space."

Wang Jian believes that with the emergence of artificial intelligence, there will be a fourth type of satellite, which can be temporarily called a "computing satellite." Only by sending AI and computing power into space can humans truly step out of the Earth. On the way to Mars, humans cannot do without the company of computing and AI, which is the most exciting thing in the next ten or even twenty years.

However, while full of infinite advantages, there are also scattered risks on the other side of the coin. Analysts at Morgan Stanley pointed out that harsh radiation in space, difficulty in on-orbit maintenance, debris hazards, and regulatory issues related to data governance and space traffic may become challenges for the development of space computing power centers.

Faced with energy problems such as power shortages, exploring space computing power has become a must-fight battle for U.S. tech companies. And in this global competition for space computing power, China is not absent.

According to Xinhua News Agency on November 29th, the China National Space Administration has recently established the Commercial Aerospace Department, and related businesses are being gradually carried out. China's commercial aerospace industry now has a dedicated regulatory agency.

On November 27th, a promotion meeting on the construction of space data centers was held in Beijing. It is reported that Beijing plans to build and operate a centralized large-scale data center system with a power exceeding gigawatts (GW) in the 700-800 km dawn-dusk orbit to realize the deployment of large-scale AI computing power in space. Domestic space computing power has joined the global competition.

As recently as the 10th of this month, Zhongke Aerospace's Lijian-1 Y11 carrier rocket successfully launched 9 satellites, accurately sending satellites such as the UAE 813 satellite and the Jixing Gaofen 07B01 satellite into their predetermined orbits.

Up to this mission, the Lijian-1 carrier rocket has served 32 satellite customers, with launch service business covering Europe, North America, South Asia, the Middle East, North Africa and other regions. China's commercial aerospace is moving from "serving the domestic market" to "serving the global market."

As global tech giants have laid out space computing power one after another, perhaps a new era of computing power is coming. From the ground to space, from "space sensing and ground computing" to "space sensing and space computing," the boundaries of computing power are being redefined.

When Nvidia's H100 chip operates in space, when China builds a distributed computing power network in low Earth orbit, and when global data centers begin to migrate to space — humans are standing at the threshold of a new era.

Above the sky, the computing power revolution has quietly begun, and this revolution may reshape the future of humans and AI.

For the first time in history! Nvidia's H100 chip has been successfully operated in space.

Almost simultaneously, SpaceX launched the largest IPO in history.

The ancient Chinese saying "Man's plans are no match for heaven's designs" has been given a new technological interpretation in the new era.

While the world is still competing for electricity and land for computing power on the ground, Musk's space empire has quietly opened the prologue to the "space computing era."

On Wednesday local time, CNN reported that the world's richest man, Elon Musk, is planning to create the largest initial public offering in history for SpaceX, raising $30 billion. If the IPO is successful, its valuation will reach $1.5 trillion, far exceeding OpenAI's current valuation of $500 billion.

Almost at the same time, CNBC reported that startup Starcloud launched a satellite equipped with Nvidia's H100 GPU chip last month, training an artificial intelligence model in space for the first time.

"Greetings to Earthlings, a charming collection of blue and green." This is the greeting to Earth from the AI model trained in space.

It is reported that Starcloud's satellite is in operation and can query and respond to Google's open-source large model Gemma, marking the first time that a large model has been run using a high-performance GPU in space.

Whether it is the trillion-dollar valuation IPO fundraising or the technical verification achieved by the first successful operation of the H100 in space, their core points consistently point to "computing power and energy" — the key engine of the AI era.

The World's Largest IPO: SpaceX Becomes the "Musk" of the Capital Market

"Based on the Starship's carrying capacity, 300GW to 500GW of solar-powered AI satellites can be launched into orbit every year." In mid-November last month, Musk posted this comment.

This means that Musk has long targeted space for his artificial intelligence energy supply strategy.

According to current information, Musk is actively promoting SpaceX's IPO, which is expected to raise more than $30 billion with an overall valuation of $1.5 trillion (equivalent to approximately 10.62 trillion yuan). The final plan is to list on the capital market in the mid-to-late 2026.

If SpaceX goes public as scheduled according to current market news, it will undoubtedly become a "talk of the town" like its founder Elon Musk, attracting strong market attention.

On the one hand, as an IPO company, SpaceX's $30 billion fundraising amount has broken the previous record of $29 billion set by oil giant Saudi Aramco. It not only refreshes the global IPO scale record but also shows the market's extreme recognition of the future value of commercial aerospace.

Moreover, in terms of SpaceX's target valuation, the company's valuation of $1.5 trillion after the IPO will not only be equal to Tesla's total market value but also far exceed OpenAI's valuation of $500 billion.

On the other hand, according to CNN, if the raised funds belong to the company, given that Musk holds nearly half of the company's equity, the fundraising income will also significantly increase his net worth.

Meanwhile, Bloomberg Billionaires Index data shows that Musk's current net worth is $461 billion, most of which comes from Tesla's stocks and options. If SpaceX's valuation reaches $1.5 trillion, Musk's net worth may double.

From the industrial perspective, SpaceX's IPO has more important strategic value.

Public information shows that SpaceX plans to use part of the funds raised from the IPO to develop space-based data centers, including purchasing chips needed to operate these facilities. By equipping high-performance chips to build orbital AI computing infrastructure, the in-depth integration of commercial aerospace and artificial intelligence will enter a new stage of development.

Currently, SpaceX's Starlink business has deployed more than 8,000 satellites, covering more than 10 million users, and its annual revenue accounts for 80% of the total revenue.

Sales are expected to be about $15 billion in 2025 and further increase to between $22 billion and $24 billion in 2026. Most importantly, relying on this core business, SpaceX has achieved positive free cash flow, moving from the "money-burning expansion" stage to the "self-sustaining" stage.

For the First Time in History! The "Space Computing Era" Begins

Starcloud launched the H100 GPU into space, marking the start of competition in the space computing power track.

It is understood that Starcloud, the company that executed this satellite launch, was founded in 2024 and is a space data center startup supported by Nvidia. Data shows that its core vision is to deploy data centers in low Earth orbit to address the energy and resource bottlenecks faced by the development of AI computing power on the ground.

In early November, Starcloud launched its first prototype satellite, Starcloud-1, via a SpaceX rocket. The satellite weighs about 60kg and is about the size of a small refrigerator. As mentioned earlier, its highlight is that the satellite is equipped with Nvidia's H100 GPU, which is said to have 100 times the computing power of previous space chips.

This time, the satellite trained the NanoGPT model using the complete works of Shakespeare in orbit and responded to Google's open-source large model Gemma, sending greetings to Earth from space. This move verified the feasibility of conducting high-performance AI computing in the space environment.

Philip Johnston, CEO of Starcloud, stated that all new data centers built in the next decade may be fully transferred to space.

Starcloud will also launch commercial services, performing inference through satellite images for applications such as maritime search and rescue and forest fire location. Subsequently, Starcloud will deploy Blackwell architecture chips, with the long-term goal of building an orbital data center with a power of 5 gigawatts and a span of about 4 kilometers.

As a startup, Starcloud is gaining momentum, but it is not the only player. In the competition for space computing power, U.S. tech giants have never given up this "lucrative opportunity."

On November 4th, Google announced "Project Suncatcher," planning to launch two satellites equipped with its self-developed TPU in 2027 to verify the feasibility of distributed machine learning tasks in the space environment.

On December 10th, media reported that Sam Altman, CEO of OpenAI, is quietly planning a deal that may disrupt the space industry pattern. He plans to acquire or take a controlling stake in a commercial rocket company through cooperative review (rumored to be Stoke Space) and may compete directly with Elon Musk's SpaceX.

In addition, U.S. commercial aerospace startup Lonestar Data Holdings is also committed to deploying the first commercial data center on the lunar surface.

Dion Harris, Senior Director of Nvidia AI Infrastructure, commented: "Starting from a small space data center, we have taken a big step towards a future where orbital computing utilizes the infinite energy of the sun."

Is Space Computing Power the Optimal Solution?

Taking GPT-3 as an example, its training requires thousands of GPU chips and takes several months; OpenAI invests as much as billions of dollars annually in training large models. Behind the high training costs is the sharp increase in computing power costs.

Nvidia H100

According to the seventh annual AI Index report released by Stanford HAI Institute, it reveals the current situation of the artificial intelligence industry. The report points out that over time, the training costs of models have increased sharply, especially the greater the computational training demand, the higher the training cost of the model. (As shown in the figure below)

In addition, according to the International Energy Agency's "Electricity 2024" report, global data center electricity consumption may exceed 1000 TWh by 2026 (about 114 GW of continuous load), of which the proportion of AI data centers is rising rapidly. Goldman Sachs also predicted in a report in April last year that by 2030, AI-related electricity demand will increase by 160%, and total data center electricity consumption may account for 8% of global electricity, exceeding the current proportion of 1% to 2%.

Space data centers can utilize all-weather high-intensity solar energy without being affected by day-night cycles, weather, or atmospheric losses. In space, there are no adverse factors such as atmospheric refraction and cloud cover, so the solar energy conversion rate has significant advantages compared to the ground. The extremely cold environment of about -270°C in deep space only requires the deployment of thermal conductive materials to complete heat dissipation, without consuming precious water resources.

Therefore, realizing the advantages of space computing power has become a breakthrough to solve the constraints of computing power and energy in the AI era.

At the 2025 Shanghai Bund Summit in September, Wang Jian, founder of Alibaba Cloud, put forward 9 assertions: "Space has always been our greatest resource. Today, AI should absolutely not only be used on mobile phones and computers; in fact, artificial intelligence should not be absent from space."

Wang Jian believes that with the emergence of artificial intelligence, there will be a fourth type of satellite, which can be temporarily called a "computing satellite." Only by sending AI and computing power into space can humans truly step out of the Earth. On the way to Mars, humans cannot do without the company of computing and AI, which is the most exciting thing in the next ten or even twenty years.

However, while full of infinite advantages, there are also scattered risks on the other side of the coin. Analysts at Morgan Stanley pointed out that harsh radiation in space, difficulty in on-orbit maintenance, debris hazards, and regulatory issues related to data governance and space traffic may become challenges for the development of space computing power centers.

Faced with energy problems such as power shortages, exploring space computing power has become a must-fight battle for U.S. tech companies. And in this global competition for space computing power, China is not absent.

According to Xinhua News Agency on November 29th, the China National Space Administration has recently established the Commercial Aerospace Department, and related businesses are being gradually carried out. China's commercial aerospace industry now has a dedicated regulatory agency.

On November 27th, a promotion meeting on the construction of space data centers was held in Beijing. It is reported that Beijing plans to build and operate a centralized large-scale data center system with a power exceeding gigawatts (GW) in the 700-800 km dawn-dusk orbit to realize the deployment of large-scale AI computing power in space. Domestic space computing power has joined the global competition.

As recently as the 10th of this month, Zhongke Aerospace's Lijian-1 Y11 carrier rocket successfully launched 9 satellites, accurately sending satellites such as the UAE 813 satellite and the Jixing Gaofen 07B01 satellite into their predetermined orbits.

Up to this mission, the Lijian-1 carrier rocket has served 32 satellite customers, with launch service business covering Europe, North America, South Asia, the Middle East, North Africa and other regions. China's commercial aerospace is moving from "serving the domestic market" to "serving the global market."

As global tech giants have laid out space computing power one after another, perhaps a new era of computing power is coming. From the ground to space, from "space sensing and ground computing" to "space sensing and space computing," the boundaries of computing power are being redefined.

When Nvidia's H100 chip operates in space, when China builds a distributed computing power network in low Earth orbit, and when global data centers begin to migrate to space — humans are standing at the threshold of a new era.

Above the sky, the computing power revolution has quietly begun, and this revolution may reshape the future of humans and AI.