人工智慧与云端服务快速扩张，推高算力需求并拉紧供电：地面一座中型资料中心的用电量可相当于约 16,500 户家庭，更大型设施甚至接近一座小城市的耗电。为降低地面能源与散热压力，科技业开始倡议把部分 AI 基础设施搬到太空，利用稳定充足的日照以太阳能供电。
Google 于 2025 年 11 月 4 日公布「Project Suncatcher」，提出在近地轨道部署 81 颗卫星的构想，目标是在约 400 英里（650 公里）的太阳同步轨道运作，以太阳能驱动太空中的 AI 计算，并把「资料」而非电力回传地面，把大量热能留在太空真空中散去。
但低轨早已拥挤且充满碎片：碎片在低轨约以 17,500 英里/小时（28,000 公里/小时）飞行，反卫星测试与卫星解体加剧风险；Starlink 已逾 7,500 颗卫星。美国太空军可追踪逾 40,000 个大于垒球的物体，但这仍不到致命物体的 1%，大量更小碎片几乎不可见。Suncatcher 拟以半径 1 公里、节点间距小于 200 公尺的超高密度编队飞行（原型预计 2027 年初），在阻力与太空天气致密度波动下误差空间极小，一次撞击就可能连锁摧毁并产生数以百万计新碎片，推向 Kessler 连锁碰撞。2025 年上半年 Starlink 曾进行 144,404 次避碰；文中推估 Suncatcher 可能每 5 秒就遭遇一次大于沙粒的威胁。政策面上，FCC 于 2022 年要求任务结束后 5 年内离轨，也有「轨道使用费/碎片清除税」构想以资助主动清除碎片。
AI and cloud growth is driving a surge in computing demand and electricity use: a single medium-sized Earth data center can consume enough power for about 16,500 homes, and larger sites can rival a small city. To ease power and heat constraints, tech leaders are exploring space-based AI infrastructure powered by abundant, reliable sunlight.
On Nov. 4, 2025, Google unveiled Project Suncatcher: an 81-satellite low Earth orbit constellation intended to run solar-powered AI in space and beam data back to Earth. It targets a Sun-synchronous orbit about 400 miles (650 km) up, aiming to keep the heat of computation in the vacuum of space.
The main obstacle is orbital debris and congestion. Debris in LEO travels ~17,500 mph (28,000 km/h); commercial constellations like Starlink (now >7,500 satellites) add collision risk. The U.S. Space Force tracks >40,000 objects larger than a softball, yet that is <1% of lethal debris. Suncatcher’s proposed 1 km-radius cluster with <200 m spacing (with prototypes planned by early 2027) leaves little margin under drag and space-weather-driven density changes; one strike could cascade, generating millions of fragments and pushing toward Kessler syndrome. In the first half of 2025, Starlink executed 144,404 avoidance maneuvers; Suncatcher is estimated to face >grain-of-sand threats every five seconds. The FCC’s 2022 rule requires deorbit within five years of mission end, and orbital-use fees have been proposed to fund active debris removal.