Keeping flying satellites connected with lasers

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: Keeping flying satellites connected with lasers · Channel: Google · 02:07 · Watch on YouTube

Topics: AI infrastructure and hardware, Robotics and autonomous systems

Summary

Project Suncatcher is exploring data centers in space, using abundant power to scale AI. Connecting the satellites requires high-bandwidth, low-latency communication, which the project aims to achieve with laser links. The team is adapting ideas from space and terrestrial communications to support very high bandwidth over shorter distances. A proof-of-concept test used optical transceivers and two telescopes to demonstrate a bidirectional link of 800 gigabits per second, and the team says there is no known theoretical reason the concept cannot work.

Key points

Project Suncatcher is a Google moonshot to build data centers in space and scale AI.
The satellite constellation needs high-bandwidth, low-latency communication.
The project uses lasers to transmit information between neighboring satellites instead of physical connectors.
Laser links require precise targeting while the satellites are moving.
Existing space laser systems are generally designed for lower bandwidth over larger distances.
The team is combining approaches from space technology and terrestrial technology.
The proof of concept tests whether slightly modified terrestrial data center hardware can work in space.
Off-the-shelf optical transceivers designed for fiber were used to drive two telescopes.
The demonstration achieved a bidirectional link of 800 gigabits per second.

Chapters

Start	Chapter	Summary
00:00	Project Suncatcher and the connectivity challenge	The project aims to build data centers in space and needs fast, low-delay communication to connect the satellites.
00:32	How satellite laser links work	Lasers encode information in properties of light and must be precisely aimed between moving satellites.
01:03	Adapting laser communications	The team is combining space and terrestrial technologies to achieve high bandwidth over shorter distances.
01:35	Proof-of-concept test	A telescope-based demonstration with optical transceivers showed bidirectional communication at 800 gigabits per second.

Quotes

Project Suncatcher is a Google moonshot — 00:00

Instead of using physical connectors in space, we use lasers. — 00:32

we are playing with all sorts of different approaches — 01:03

I have 800 gigabits per second both directions. — 01:35

Concepts and entities

Project Sun Catcher, Google, comunicação a laser no espaço, transceptores ópticos, data center espacial

Full transcript (original language)

00:00 - Project Suncatcher is a Google moonshot to try to build data centers in space taking advantage of basically free and abundant power, and scale AI to deliver something really amazing. (gentle ambient music) One of the big questions we need to answer is “How can we keep the constellation connected?” If you think about a data center on Earth, it’s a very large number of chips that are working together as efficiently as possible. So lots of data with no delay. If we’re going to do that in space, we need to have really,

00:32 really high bandwidth, low latency communication between the satellites, as well as within a single satellite. (gentle ambient music) Instead of using physical connectors in space, we use lasers. We encode information in the different properties of light, amplitude, frequency, phase, and we fire a laser from one satellite to its neighbor, and collect the information on the other side. (gentle ambient music) The level of precision is broadly equivalent to trying to hit a coin-sized target from kilometers away, while you and the target are also moving.

01:03 Technology to use laser communications in space is well-understood, and exists at the moment. The challenge for us is that most of the state-of-the-art systems are optimized to do something very different. They’re trying to do relatively low bandwidth at very large distances, whereas we’re trying to do very high bandwidth at small distances. And in order to do that, we are playing with all sorts of different approaches, trying to get the best of space technology, and the best of terrestrial technology, and combine them. (gentle ambient music) - What we’re trying to do with this project is the first stage is proof of concept.

01:35 And the whole point was, can we just take a data center hardware that’s literally running in terrestrial data centers, modify it slightly, and see if it’s going to work in space. We’ll use off-the-shelf optical transceivers that’s used for a fiber, but instead of driving a fiber, we’ll drive two telescopes. I have two telescopes, and it was like bidirectional link. And when I stick my hand in, I blocked both lights, so there’s no communication. When I remove it, data was flowing both ways, and what I’ve seen is I have 800 gigabits per second both directions.

02:07 It was simple test, but they show proof of concept. Now we can actually pushing this data race even higher. - We’ve done enough research to say that there isn’t a theoretical reason that we know of that this isn’t going to work. There’s available power, there’s tons of space in space, and I think that it’s not outrageous to say that AI would never be the same if we can make this happen. (gentle ambient music ending)