Google Suncatcher的原型空间数据中心是在线的，本月早些时候发布之后与空间X猎鹰9号火箭上的其他100多个航天器一起与地面通信。
该卫星载有4个Google Tensor处理单元(TPU)并产生高达1千瓦的电力，目的是为未来在轨道上进行大量推断工作的近距离飞行卫星网络铺平道路。 然而，TPU的动力需求意味着它每次只能开启一个约20分钟，使已经很棘手的空间对地通信业务复杂化。
经过大约三个星期对卫星的子系统进行测试的试运行过程，航天器已准备好工作。 当地面的工程师本周送来他们的第一个提示时，他们要求预先装入卫星的模型，Google DeepMind的Gemma 4，写一个有关太空的haiku. 因此：
蓝色大理石下面
静态数据流
星光为我的核心。
Google 称用这些以空间为主题的石灰石测试模型(他们本应该派一位诗人？)，也许更重要的是，公司要求航天器报告卫星周围的轨道状况。
发言人告诉TechCrunch，卫星名义上运行。
正在进行的试验活动将审查芯片在热力压力下和空间辐射环境中的表现。 Google在地球上的粒子加速器中广泛测试了它的TPU，但现在需要看看辐射对其在太空中的硬件的真正影响如何影响它们的处理器。 辐射可造成“位翻转”——在计算机内存中产生存储值——这反过来又可能破坏数据而不触发典型的监测工具。
实地的试验表明，辐射的影响对于推论工作量是最小的，对于培训一个模型来说，问题更多，但现在Google小组将评估这些动态，而不是模拟。
如果这个航天器上的TPU证明是有能力的，下一步将是在2027年再发射两颗配备先进芯片的发布卫星。 这些航天器将支持更高的发电，并展示卫星间光学联系。
Google的Starcloud项目在规划卫星之间的高速联系以支持一致分配的工作量方面，在空间数据中心的愿景中大多是孤立的；Space QQs数据中心预计也会使用激光链接，但也会运行得太远，无法有效共享处理。 星云号和牛仔太空号等公司预计会在单个卫星上进行推论载荷。
Google和SpaceQQ的视野紧密相连。 Alphabet,Google的母公司，在SpaceX拥有重要的投资股份。 而对于发布的轨道数据中心，两家公司将要求星舰(Starship)，即位于埃隆·穆斯克太空视野中心的仍在开发中的重力运载火箭，完全和快速地重新使用。
在本月早些时候发表的同行评审分析中，Suncatcher研究者估计SpaceX需要飞1,800次星舰，以达到与它早先的飞行器相同的学习驱动成本降低。
Suncatcher是一个经典的Google moonshot——公司的重点是设计20世纪30年代经济的基础设施。 然而，能否到那里，可能取决于公司其他成员如何在近期内对大赦国际的压力和机会作出反应。 Google的母公司Alphabet本周报告，将AI插入搜索业务
校正：Suncatcher原型卫星机上装有4个TPU，但一次只能操作一个。
本文由 AI图鉴自动抓取并翻译，版权归原作者所有；如需查看原文请点下方链接。
Google Suncatcher’s prototype space data center is online and communicating with the ground after launching earlier this month alongside more than 100 other spacecraft on a SpaceX Falcon 9 rocket.
The satellite, which carries four Google Tensor Processing Units (TPUs) and generates up to 1 kW of electricity, is intended to pave the way for future networks of close-flying satellites performing large inference workloads in orbit. However, the power demands of the TPU mean it can only turn one on for about twenty minutes at a time, complicating the already tricky business of space-to-ground communications.
After a commissioning process that tested out the satellite’s subsystems over about three weeks, the spacecraft was ready to work. When engineers on the ground sent their first prompt this week, they asked the model pre-loaded onto the satellite, Google DeepMind’s Gemma 4, to write a haiku about being in space. And thus:
Blue marble below
Data streams in silent grace
Starlight feeds my core. ✨
Google said it tested the model with these space-themed limericks (they should have sent a poet?) and, perhaps more importantly, the company asked the spacecraft to report on the orbital conditions around the satellite.
A spokesperson told TechCrunch that the satellite was operating nominally.
The ongoing test campaign will examine how the chips perform under thermal stress and in the space radiation environment. Google has extensively tested its TPUs in particle accelerators on Earth, but now needs to see how radiation’s real effects on their hardware in space affect their processor. Radiation can cause “bit flips”—causing a stored value in computer memory—and that in turn can corrupt data without triggering typical monitoring tools.
Testing on the ground suggested the impact of radiation would be minimal for inference workloads and more of a problem for training a model, but now the Google team will get to assess those dynamics in reality, not a simulation.
If the TPUs onboard this spacecraft prove capable, the next step will be launching two more satellites equipped with the advanced chips in 2027. Those spacecraft will support higher power generation and demonstrate inter-satellite optical links.
Google’s Starcloud project is largely alone among visions for space-based data centers in planning on high-speed links between the satellites to support coherent distributed workloads; SpaceX’s data centers are also expected to use laser links but also to orbit too far apart to share processing effectively. Companies like Starcloud and Cowboy Space expect to perform inference loads on individual satellites.
Google and SpaceX’s visions are closely linked. Alphabet, Google’s parent company, holds a significant investment stake in SpaceX. And to launch their orbital data centers, both companies will require Starship, the still-in-development heavy-lift rocket at the center of Elon Musk’s space vision, to be fully and rapidly reusable.
In a peer-reviewed analysis published earlier this month, Suncatcher researchers estimated that SpaceX would need to fly Starship 1,800 times to achieve the same learning-driven cost reductions it saw with its earlier vehicles.
Suncatcher is a classic Google moonshot—the company’s focus is on designing infrastructure that will be economical in the 2030s. Getting there, however, may depend on how the rest of the company responds to the pressures and opportunities of AI in the near term. Alphabet, Google’s parent company, saw its stock plunge this week after reporting that efforts to insert AI into its search businesses are hurting ad revenue growth.
Correction: The Suncatcher prototype satellite has four TPUs onboard, but can only operate one at a time.
Auto-fetched and translated by AI Atlas. All rights belong to the original author.