AI drives new opportunities and risks in space

Researchers at Stanford first brought machine learning to robots aboard the International Space Station in 2025, helping them plan movements 50% to 60% faster and opening a new chapter for artificial intelligence (AI)-supported robots in space. This is just one example of how AI has moved to the center of the space sector, transforming how we design, operate, and govern activity in orbit.

Hitting a record $613 billion in value in 2024 (78% of which was in the commercial sector), McKinsey estimates the space economy could grow to $1.8 trillion by 2035. As the space sector rapidly advances—driven by technological advances and market dynamics—its growth is increasingly reliant on petabyte-scale data streams, mega-constellations numbering in the thousands, and high-tempo operations that exceed human decisionmaking capacity. This means AI will be a major player and constructor of what the future of the space industry means from a business, innovation, and governance perspective.

The accelerating trends shaping AI in space

The number of satellites in orbit is approximately 15,000 but is projected to reach 100,000 by 2030. If all current filings for planned low Earth orbit (LEO) satellites at the Federal Communications Commission (FCC) launch, the number would reach half a million satellites by the end of the 2030s. This staggering growth is driven by commercial expansion in mega-constellations in LEO, with Starlink currently accounting for about two-thirds of all active satellites—more than all countries combined.

The growth in satellites also brings growth in data. NASA’s Earth Observation System Data and Information System archive has already hit 100 petabytes (PB) of data and is expected to reach 320 PB by 2030. Meanwhile, stakeholders access 5 PB of data per month from the satellites and sensors from the National Oceanic and Atmospheric Administration (NOAA)’s Open Data Dissemination (NODD) program. Europe’s Copernicus program also has more than 78 PB in online data. 

The reality of this acceleration in satellite launches and explosion of available data means that the true value for future companies, governments, scientists, and innovators will be the “time-to-insight” from this data. Edge AI directly on satellites and cloud-to-edge AI on the ground will be the backbone for handling these increased volumes. Edge computing enables real-time processing directly aboard spacecraft rather than routing all data to Earth for analysis. This is critical when communication delays can range from minutes to hours in deep space. Its distributed intelligence allows satellites to filter and prioritize data before transmission, which reduces bandwidth requirements and enables autonomous decisionmaking.

Opportunities

Space industry’s value-add

Raycho Raychev, founder of EnduroSat, and James Mason, Chief Space Officer at Planet, recently argued that as intelligence and AI models become more commoditized, unique data and service provision will likely become the scarcest and most valuable resources. They iterated how space companies should be at the forefront of reaping the benefits from this shift given their sheer amount of accessible data and ability to make leaps in big data analytics. 

Integrating AI to unlock new use cases

The Stanford researchers’ advance is a successful example of integrating AI into existing technology to improve performance and unlock new uses. The same is happening in navigation and planetary landing; technologies like AI-enabled computer vision and terrain-relative navigation enable missions, like NASA’s 2020 Mars mission, to touchdown on sites previously too hazardous, while also helping to identify discoveries like geological features or signs of water or life. The Ingenuity Mars Helicopter autonomously flew over Martian terrain on 72 flights, with testing underway to allow for landing on even more difficult conditions. AI’s space capabilities are continuing to expand to include both Earth-based innovations and planetary/extra-terrestrial discoveries.

Challenges

The space sector has its own challenges, whether business or governance related, but the greater integration and shaping power of AI for the industry opens up new and amplifies existing complex challenges.

Scaling and vertical integration

Raychev and Mason discussed how vertical integration for space companies can be extremely prohibitive for reaching scale. Because space companies currently do most things themselves—from building the actual infrastructure and developing the data, analytics, and AI capabilities—their resources are stretched incredibly thin, making it difficult for any of them to scale. If the industry could have some players developing spacecrafts and infrastructure, others could focus their resources and talents on intelligence applications.

Market concentration

AI integration in space also creates a challenge for governments, as it can further concentrate the market into a few very capital-intensive and powerful companies. The concentration of a few companies in a few countries creates a challenge to prevent exclusion and promote coordination in the two complex domains of AI and space. If Earth observation data and data powering AI algorithms are concentrated in specific regions, countries can be left behind in accurately applying models to their own conditions. If then a single entity controls both the physical infrastructure and the AI systems that process space-based data, nations could find themselves dependent on foreign commercial platforms for everything from agricultural monitoring to national security intelligence.

Governance and cybersecurity

Both AI and space industries already face complex cybersecurity and governance challenges that are only amplified and further complicated as they continue to integrate. With commercial satellites supporting military and defense intelligence, new avenues of cyberattacks are becoming more common, such as GPS jamming in Europe, attacks against space agencies in Japan and Poland, and ransomware attacks across 25 different space-sector organizations in 2024 alone.

The convergence of AI and space creates a double “dual-use technology” problem. Both technologies are inherently dual-use individually (can be applied for civilians and militaries), but their combination creates entirely new categories of risk that traditional governance frameworks are not capable of handling. AI-driven space-based decisions take microseconds, which means governance structures that assume human decisionmakers are in the loop do not apply.

Strategies

With satellite launches expected to rapidly accelerate, there is a pressing imperative to ensure that safety is at the forefront of the public and private sectors’ concerns. 

Recent United Nations Office for Outer Space Affairs (UNOOSA) recommendations call for “human‑in‑the‑loop for low‑latency operations, and human‑on‑the‑loop with robust safeguards for deep‑space missions where real‑time intervention is impossible,” which supports governance frameworks that pre-authorize AI decisions within defined parameters, similar to how nuclear power plants have automated safety systems that do not wait for human approval. UNOOSA recommends that academia develop new technical standards like explainable AI for space-grade hardware; the private sector incorporate decision logs and risk-based safety assessments; and UN governments develop an “international code of practice for AI in space.”

The public sector must play an active role in integrating these use cases while also adapting to the new reality where commercial satellites and private companies are working with the public sector to provide military and defense intelligence. For example, the U.S. Department of Defense recognized the need to have integrated defense strategies given the role of commercial space systems in military cyberattacks and created a “Commercial Space Integration Strategy.”

To address fairness and inclusivity related to the market concentration challenge, UNOOSA recommends academia to spearhead international collaborations that “co-design AI-for-[Earth observation] projects”; the private sector to participate in knowledge-transfer partnerships; and governments to ensure the publication of data from publicly funded Earth-observation (EO) missions and expand UNOOSA’s Access to Space for All initiative to include targeted scholarships on AI and EO.

In order to strengthen coordination, a global cybersecurity protocol for space will be a key area to develop and deploy in the near future, with some experts calling for global information sharing in real time and coordinated responses to incidents. Strategically integrating AI innovations can serve as part of the solution, detecting potential cybersecurity threats by identifying anomalies in data and preventing physical threats by anticipating intrusion points and overcoming communication delays.

Conclusion

These challenges and opportunities are amplified as AI is further entrenched as a key shaper of the future space industry. Coordinated yet agile governance will be critical for the success of commercial companies and their deployments—from agricultural precision tools to land-use analytics for urban planning—as well as for governments striving to ensure their citizens are included in technological advances and can benefit from unprecedented opportunities to improve quality of life on Earth and beyond.