Dissecting Curiosity about Space Data Center
“As electricity, water, and land became scarce on Earth, large tech launched sconcludeing computer warehoutilizes up into space, where sunlight shines around the clock and cools itself like an ice sheet.”
The battlefield of artificial innotifyigence (AI) technology competition between major hegemonic countries and technology companies is shifting into space. This is becautilize AI’s explosive growth has reached a level where global (ground) infrastructure is difficult to handle.
Representatively, the Space-Based Data Center (SBDC) is rapidly emerging as an alternative to ground AI infrastructure facing power shortages and cooling limitations. The idea is to build a data center in space to solve energy supply, cooling, and site problems at once.
Morgan Stanley diagnosed in a report published in December last year that “space data centers are becoming a new grand narrative of AI infrastructure.” McKinsey, a global consulting firm, predicted that the global space economy ecosystem will expand to 1.8 trillion dollars (about 2,500 trillion won) by 2035.
The concept of a space data center was considered a distant future only two years ago, but now it has become a real business where all large tech such as Google, Amazon, SpaceX, and Nvidia have jumped in and trillions of won of funds are actually shifting.
This
What’s a space data center
If I question a question on ChatGPT, where will the answer be created.
It doesn’t come from our smartphone.
For example, somewhere in Texas or Nevada, the United States, tens of thousands of computers (servers) are running 24 hours a day in a huge warehoutilize the size of dozens of soccer fields to create answers. Such a huge warehoutilize is a ‘data center’.
The ‘space data center’ literally plans to launch this huge computer warehoutilize into outer space.
The concept is to float hundreds to tens of thousands of sanotifyites in low-Earth orbit (LEO) much higher than an airplane and around 400 kilometers around the International Space Station (ISS), and put a high-performance computer into the sanotifyites to perform AI operations.
In November last year, U.S. start-up Starcloud put the 60-kg sanotifyite “Starcloud-1” carrying Nvidia H100 graphics processing unit (GPU) into actual orbit. It succeeded in turning Google’s AI models ‘Gemma’ and ‘Gemini’ directly in space. In other words, AI has started believeing outside the Earth.
How far is technology commercialization
StarCloud launched its first sanotifyite, ‘StarCloud-1’ in November last year. The sanotifyite, which weighs 60 kilograms, is equipped with Nvidia H100 GPU, the highest performance AI accelerator for data centers. It was the first time that a top performance GPU worked in outer space.
Based on the technology demonstration, StarCloud attracted an additional $170 million in large-scale investment in March this year, 17 months after its foundation, and quickly rose to the rank of a ‘unicorn company’ with an enterprise value of $1.1 billion. StarCloud is planning a 5 gigawatt (GW) orbital mega cluster with a solar panel area of 4㎢.
Canadian space communications company Kepler Communications successfully put 10 sanotifyites called AETHER into orbit in January this year.
These sanotifyites are equipped with 40 Nvidia ‘Jetson Orrin’ edge processors. The sanotifyites are interconnected by optical laser communication (OISL) and operate like one large distributed GPU computing cluster that dynamically performs operations in outer space.
The strategy is to immediately infer and analyze fires or military relocatements on track without waiting for the “downlink delay time” to sconclude vast amounts of data down to the ground.
Elon Musk’s SpaceX applied to the U.S. Federal Communications Commission (FCC) on January 30 for permission to deploy up to 1 million data center sanotifyites at an altitude of 500 to 2,000 kilometers.
Why do you have to go all the way to space
AI is an electric hippo. Each time we question a question to an AI chatbot, it has an average electricity of 0.24 watt-hours (Wh). For example, if the era of full-fledged AI agents arrives and ChatGPT estimates that it processes 2.5 billion questions a day, this alone will sweep 0.4% of the total U.S. electricity.
The International Energy Agency (IEA) predicts that global data center power consumption will exceed 1000 TWh by 2026. This is large enough to be compared to the demand for electricity at the national level. Furthermore, data centers around the world are expected to utilize 120 nuclear power plants by 2030.
According to McKinsey, $5.2 trillion in AI-related data center infrastructure is required by 2030, and additional AI data center capacity of 156 GW is required.
The ground grid cannot meet all of these demands. In Virginia, the United States, you have to wait seven years for a new data center to be connected to the electrical grid. According to Kushman & Wakefield, a real estate service company, more than 25 GW of data centers are currently being built in the United States alone.
On the other hand, the sun shines for free 24 hours in space. There are no clouds, no nights, no weather.
High-performance AI computers perform calculations and emit enormous heat. Currently, ground data centers are pouring about 40% of their total power consumption into cooling this heat alone.
In an uninhabited universe, there is no opposition to humans.
According to Data Center Watch, a $18 billion data center project has been blocked in the U.S. over the past two years due to regional opposition. 142 activist groups in 24 U.S. states are opposed to the construction of data centers. On the other hand, there are no residents, no civil complaints, and no borders in space.
The reality of the space data center is
The advantages of space data centers include △ solar-based virtually unlimited power △ radiation cooling applying space vacuum △ freedom from land, water, and civil complaints △ global low-latency edge computing △ carbon neutrality.
According to StarCloud, a 10-year operation of a 40MW data center on the ground costs about $140 million, while an orbital data center can be reduced to about $2 million through solar utilization.
But there are also a lot of problems. It is necessary to break through considerable obstacles to complete commercialization. The indusattempt explains, “The direction is right, but there is a long way to go.”
It is explained that it can be a realistic option only when the cost of launching is reduced by one-tenth, radiation-resistant semiconductors are developed, and space assembly robots are put into practice. The market sees this as the mid-2030s.
When will we all benefit from ‘space AI’.
Currently, there are only dozens of high-performance AI chips running in space. Compared to 4 million GPUs sold on the ground by Nvidia in 2025 alone, it is far short.
The prevailing view in the indusattempt is that the main game will launch in the 2030s. If SpaceX’s mega rocket “Starship” is commercialized in earnest and the launch cost falls below $200 per kg, it is expected that the space data center will be able to compete with the ground from then on.
Why is the universe connected to security
Competition in space data centers is now being reorganized into a key pillar of national security beyond competition among private companies.
Beyond the logic of simply streamlining AI operations, there is a calculation that considers it a “strategic asset” that can track △ hypersonic missiles in real time, break away from depconcludeence on △ enemy cloud infrastructure, and maintain command and control (C2) even in wartime situations where △ ground communication networks are paralyzed.
The United States is taking a path that combines missile defense and space computing. “Golden Dome for America,” announced by President Trump in 2025 as an executive order, is a multi-layered defense system that detects, tracks, and intercepts ballistic and hypersonic missiles from space. James O’Brien, head of sanotifyite communication and spectrum at the U.S. Space Command, attconcludeed a conference in March and questioned if an orbital data center is essential for the Golden Dome, stateing, “I can’t imagine it without it.”
China also has an ambitious strategy in place.
China’s “Three-body Yeonsan Seonggun,” launched from the Juquan Sanotifyite Launch Center on May 14 last year, is the world’s first space AI supercomputer network deployed in practice. It is named after Lutsin’s science fiction novel “Three Body”.
It was jointly developed by ADA Space (Guoxing Aerospace), based in Chengdu, China, and Zhejiang Lab. It is equipped with an AI model of 8 billion parameters per sanotifyite and performs 744 trillion operations (744 TOPS) per second. The first 12 units alone produce computing power of 5 PetaOPS.
China’s ultimate goal is to build 2,800 fortresses. Once completed, it will reach 1,000 Peta OPS and overwhelm the world’s strongest supercomputer on the ground.
Military implications are also noteworthy. General B. Chance Saltzman, commander of U.S. space operations, also defined China’s sanotifyite network as a “kill web” that poses a significant threat to U.S. space assets.
The three-body military can detect and track missile launches or fleet relocatements in real time without ground intervention. Even if the ground network is paralyzed by natural disasters, cyberattacks, and geopolitical collisions, it operates autonomously in orbit. China has created this a key pillar of the ‘new infrastructure’ national strategy and the ‘No. 1 AI world in 2030’ plan. The goal is to complete the entire military by 2035.
Europe seems to have put forward two pillars: ‘Sovereignty’ and ‘Net Zero’.
In the space data center sector, the ‘ASCEND Project’ led by Thales Alenia Space is considered a representative example. The EU Commission has funded feasibility studies, aiming to secure both the EU Green Deal carbon neutrality goal and digital sovereignty in 2050. The plan is to launch Europe’s indepconcludeent cloud on an ‘orbit outside the border’.
※Terminology explanation = Space Data Center
Space-Based Data Center (SBDC) refers to an infrastructure that performs data storage, processing, and AI operations by mounting high-performance semiconductors on sanotifyites placed in low-Earth orbit (LEO) or solar synchronous orbit (SSO). The mainstream method is to combine hundreds to tens of thousands of sanotifyites with laser optical communication and operate them as one huge cluster.
Summary of 5 lines
1. Space Data Center (SBDC) has emerged as the ultimate alternative to overcome the depletion of power grids, cooling water resources, and sites on the ground due to the explosion of AI operations. Space can generate 24-hour solar power without clouds or constraints, and can be cooled without radiation applying deep space at minus 270 degrees Celsius, which is the perfect physical condition for AI infrastructure. 3. Global large tech companies such as SpaceX, Google, and Nvidia have already succeeded in demonstrating sanotifyites equipped with high-performance GPUs by putting trillions of won into the competition for supremacy. 4. The astronomical cost of launching rockets, space radiation damaging semiconductors, and the physical challenges of on-orbit maintenance and assembly are still considered challenges to overcome. 5. It is expected that the real battle will take place in the mid-2030s when the unit price of launches will collapse due to the commercialization of next-generation rockets
※ The full text of The Tech Wave, including an interview with a space ecosystem strategy officer at Europe’s largest sanotifyite manufacturing company invested by French defense company Thales, will be available on the premium financial content platform “Maegyeong Plus,” which will open on May 7. In the future, we will meet more in-depth serials on the ‘Maekyung Plus’ site.
Copy the URL below and attach it to the address window, or take a QR code with your smartphone to see Maekyung Plus membership. ☞ https://seoulmoneyreveal.com/guide/02.php
Leave a Reply