Level 1 — Absolute Beginner
Computers that run AI need a lot of power and a lot of space.
Some big companies now want to build computer centers in space, not on Earth.
In space, sunlight can power the computers all the time.
Companies like SpaceX and Google are working on this new idea.
- computer
- a machine that stores and processes information
- power
- energy used to make something work
- space
- the area beyond Earth's air, where planets and stars are
- sunlight
- light that comes from the sun
- satellite
- an object that orbits a planet, sometimes made by people
- company
- a business that makes or sells things
- idea
- a plan or thought about how to do something
- orbit
- the curved path an object follows around a planet
Level 2 — Elementary
As artificial intelligence uses more and more electricity, some of the world's biggest tech companies are exploring a bold new idea: building AI computer centers in space instead of on Earth.
The idea is that satellites in orbit could use solar panels to collect sunlight almost constantly, giving them a steady source of clean energy without needing large power plants or cooling systems on the ground.
SpaceX has filed plans with regulators to eventually launch as many as one million small satellites designed to work together as a giant computing network in orbit. Google, working with satellite company Planet Labs on a project called Project Suncatcher, is testing solar-powered satellites equipped with its own AI computer chips.
Still, many experts point out serious challenges. Keeping computers cool, protecting them from radiation, avoiding space debris, and the high cost of launching hardware into orbit all remain major obstacles before space-based data centers can compete with those on Earth.
- electricity
- a form of energy used to power machines and devices
- solar panel
- a device that turns sunlight into electricity
- network
- a group of connected things that work together
- regulator
- an official group that makes and enforces rules
- radiation
- energy that travels through space, which can be harmful in large amounts
- debris
- scattered pieces of broken or discarded material
- obstacle
- something that blocks progress or makes a task harder
- compete
- to try to be more successful than others in the same activity
Level 3 — Intermediate
As the energy demands of artificial intelligence continue to climb, a growing number of major technology companies are pursuing an ambitious alternative to building ever more data centers on Earth: moving AI computing hardware into orbit.
The appeal lies in near-constant access to solar energy. Satellites positioned in the right orbits can capture sunlight for the vast majority of their orbital period, potentially reducing dependence on the power plants and elaborate cooling infrastructure that terrestrial data centers require.
SpaceX has filed an application with regulators proposing to eventually deploy up to one million orbital data center satellites at altitudes between 500 and 2,000 kilometers, arguing that launching roughly one million tonnes of satellites annually could ultimately generate around 100 gigawatts of AI computing capacity. Separately, Google has partnered with Earth-observation satellite maker Planet Labs on Project Suncatcher, testing solar-powered satellites equipped with its own Tensor Processing Unit AI chips and demonstrating optical links capable of transmitting data at 1.6 terabits per second in laboratory conditions.
Despite the momentum, analysts caution that orbital computing faces formidable obstacles, including thermal management in the vacuum of space, radiation-induced hardware degradation, collision risk from orbital debris, and launch costs that remain far higher than the cost of building comparable capacity on the ground.
- ambitious
- having a strong desire to achieve something difficult
- terrestrial
- relating to or existing on Earth
- infrastructure
- the basic physical systems needed for something to operate
- deploy
- to put into position or use for a particular purpose
- altitude
- the height of an object above a reference point, such as sea level
- capacity
- the maximum amount something can produce or hold
- thermal management
- the process of controlling the temperature of a device or system
- degradation
- a gradual decline in quality or performance
Level 4 — Advanced
As the energy footprint of artificial intelligence workloads continues its steep ascent, a widening cohort of major technology firms is pursuing an audacious alternative to the relentless proliferation of terrestrial data centers: relocating AI computing infrastructure into orbit.
The rationale hinges on near-continuous access to solar irradiance. Satellites positioned in favorable orbits can harvest sunlight for the overwhelming majority of each orbital period, a characteristic proponents argue could substantially attenuate reliance on the power plants and elaborate thermal infrastructure that ground-based data centers demand, even as skeptics note that the vacuum of space introduces its own formidable cooling constraints.
SpaceX has filed a regulatory application proposing the eventual deployment of up to one million orbital data center satellites at altitudes spanning 500 to 2,000 kilometers, projecting that launching approximately one million tonnes of satellite mass annually could ultimately yield roughly 100 gigawatts of AI compute capacity; Elon Musk has publicly asserted that within two to three years, the lowest-cost avenue for generating AI compute will be orbital rather than terrestrial. Google, meanwhile, has partnered with Earth-observation satellite manufacturer Planet Labs on Project Suncatcher, validating radiation-hardened Tensor Processing Units and demonstrating optical interlinks capable of 1.6 terabits per second under laboratory conditions.
Notwithstanding this momentum, industry analysts caution that orbital computing confronts formidable, unresolved obstacles, including thermal dissipation in vacuum, cumulative radiation-induced hardware degradation, collision exposure from proliferating orbital debris, and launch economics that remain substantially uncompetitive relative to terrestrial buildout, factors that collectively suggest orbital compute is unlikely to rival ground-based capacity within the timeframe most critical to the current AI race.
- footprint
- the total impact or resource consumption of an activity
- audacious
- showing a willingness to take bold risks
- proliferation
- a rapid increase in the number of something
- irradiance
- the amount of radiant energy, such as sunlight, received per unit area
- attenuate
- to reduce the force, effect, or amount of something
- interlink
- a connection that links two systems or components together
- dissipation
- the process by which energy, such as heat, is spread out and lost
- uncompetitive
- not able to compete effectively due to unfavorable cost or performance
Talk about this article with KatePractice this story's vocabulary in a real spoken conversation with your AI tutor