As international demand for computing continues to blow up, the carbon footprint of information facilities is a rising concern. A brand new research outlines how internet hosting these amenities in area may assist slash the sector’s emissions.
Information facilities require monumental quantities of energy and water to function and funky the tens of millions of chips housed inside them. Present estimates from the Worldwide Power Company peg their electrical energy consumption at round 415 terawatt hours globally, roughly 1.5 p.c of complete consumption in 2024. And the Environmental and Power Research Institute says that giant knowledge facilities can use as a lot as 5 million gallons per day for cooling.
With demand for computing assets rising by the day, specifically for the reason that speedy adoption of resource-guzzling generative AI throughout the financial system, this threatens to change into an unsustainable burden on the planet.
However a brand new paper in Nature Electronics by scientists at Nanyang Technological College in Singapore means that internet hosting knowledge facilities in area may present a possible resolution. By counting on the considerable photo voltaic vitality out there in orbit and releasing waste warmth into the chilly vacuum of area, these amenities may, in precept, change into carbon impartial.
“Area affords a real sustainable atmosphere for computing,” Wen Yonggang, lead creator of the research, mentioned in a press launch. “By harnessing the solar’s vitality and the chilly vacuum of area, orbital knowledge facilities may rework international computing.”
To validate their proposal, the researchers used digital-twin simulations of orbital computing methods to mannequin how they might generate energy, handle warmth, and keep connectivity. The workforce investigated two potential architectures: one designed to scale back the footprint of information collected by satellites themselves and one other that will obtain knowledge from Earth for processing.
The primary mannequin would contain integrating knowledge processing capabilities into satellites geared up with sensors—for instance, cameras for imaging the Earth. This could make it attainable to hold out costly computations on the information on board earlier than transmitting simply the outcomes again to the bottom, quite than processing the uncooked knowledge in terrestrial knowledge facilities.
The opposite strategy includes a constellation of satellites geared up with full servers that would obtain knowledge from Earth and coordinate to hold out advanced computing duties like coaching AI fashions or working giant simulations. The researchers word that this type of distributed knowledge heart structure—versus assembling a big, monolithic knowledge heart in orbit—is technologically possible with at the moment’s satellite tv for pc and computing applied sciences.
The workforce’s evaluation means that the appreciable carbon footprint of launching {hardware} into area might be offset inside 5 years of operation, after which the amenities may run indefinitely on renewable vitality.
Important technical and logistical hurdles stay. Laptop chips are weak to radiation, an ever-present hazard in area, which might necessitate the usage of specialised radiation-hardened processors. Lengthy-term upkeep of the amenities would additionally require in-orbit servicing applied sciences that don’t but exist. And as computing applied sciences quickly enhance, chips depreciate in only a few years. Protecting orbital knowledge facilities stocked with the newest and biggest might be pricey.
However the NTU workforce isn’t the primary to drift the thought of shifting computing amenities into area. Final 12 months, French protection and aerospace big Thales printed a research exploring the feasibility of the thought. And subsequent month, the startup Starcloud will launch a satellite tv for pc carrying an Nvidia H100 GPU as a primary step in the direction of making a community of orbital knowledge facilities.
Whereas realizing the imaginative and prescient is prone to require technical breakthroughs and an enormous quantity of funding, one resolution to computing’s ever rising carbon footprint could also be above our heads.
