If you want a planet that doesn't get paved with datacenters in the long run, you do want datacenters in space
Nobody wants a city on Mars. Nobody wants AI in every app. Nobody wants data centers in space. Nobody wants humanoid robots or flying cars.
We want clean water. We want a stable climate. We want a habitable planet. We want wildlife to survive. We want to live.
4
10
156
4,728
If you hold a normal grain of sand to the sky at arms length, that grain is covering 10,000 galaxies. 1,000,000,000,000,000 stars, each with at least 1 planet on average.
The scale of the universe is literally incomprehensible.
The scale of the universe is literally incomprehensible.
2
1
35
2,567
Ezra Feilden retweeted
Great to have our esteemed colleagues @Google planning to join the GPU party on orbit alongside dozens of @nvidia Jetsons, Thors, and even a H100 courtesy of @Starcloud_
1
1
5
2,595
Ezra Feilden retweeted
I’m super excited share that @MIT-trained aerospace engineer and serial entrepreneur Peter Reinhardt @reinpk has committed $1M to the @FermiExplorer mission to send humanity’s first spacecraft to another star! 💫
Find out how you can get involved and contribute at FermiExplorer.org 🤓
And thanks so much for your generosity and time, Peter, welcome aboard!! 🙏😃🚀
Full article: uk.finance.yahoo.com/news/mit-train…
Find out how you can get involved and contribute at FermiExplorer.org 🤓
And thanks so much for your generosity and time, Peter, welcome aboard!! 🙏😃🚀
Full article: uk.finance.yahoo.com/news/mit-train…
10
20
121
42,295
Ezra Feilden retweeted
Brilliant … so, that would make a:
Phone: Compute
Desktop/Laptop: Advanced Compute
Data Centres: Super Compute
Data Centres in Space:
Super Star Compute
A collection of data centres in space: @Starcloud_ 💫
Phone: Compute
Desktop/Laptop: Advanced Compute
Data Centres: Super Compute
Data Centres in Space:
Super Star Compute
A collection of data centres in space: @Starcloud_ 💫
Gwynne is suggesting we rebrand data centers to super compute.
0
1
2
268
This is cool. Starlight replaced by an eerie imperceptibly dim infrared glow.
For those saying we wouldn’t be know what to look for to see a Dyson sphere in a far galaxy, how about this? We can see >1 trillion galaxies. This is what one would like a few hundred thousand years into hitting super intelligence and starting to build Dyson spheres. I’m not even joksomething like this in the hundred billion new galaxies that it is about to observe.
2
4
20
3,809
Starcloud-2 payload shipping out for integration today. Soon to be by far the most powerful computer ever put in space.
21
14
151
16,023
It only took humanity 4 years to go from launching the first satellite to launching the first interplanetary spacecraft. Yet it's been 60 years since then and we've never launched a spacecraft to another star.
Fermi fixes that.
Fermi fixes that.
Did you know that humanity has never launched a spacecraft to another star? That is about to change! I'm super excited to share that my cofounders, @AdiOltean, @EzraFeilden, and I, have created a non-profit to attempt to send humanity’s 𝗳𝗶𝗿𝘀𝘁 𝘀𝗽𝗮𝗰𝗲𝗰𝗿𝗮𝗳𝘁 𝘁𝗼 ed your help!
We have formed a non-profit called The @FermiExplorer Mission with the goal of launching a minimum viable spacecraft to Alpha Centauri, the nearest star system, leaving in 3 years and arriving in around 80,000 years. We hope it will be both the first spacecraft to leave Earth for another star, and the last to arrive, as per this concept video!
The mission is intended to make the idea of interstellar travel real, and to encourage more people to confront one of humanity’s biggest unanswered questions when we look beyond our planet: "Where is everybody?" (the Fermi Paradox)
The mission has four high-level objectives:
- Aim: To launch the first spacecraft to another star. It should get at least 99% of the way to the nearest star system, Alpha Centauri, within 80,000 years.
- Payload: It should have at least a 1kg, 10x10x10 cm payload.
- Schedule: It should launch before the end of 2029
- Cost: It should cost less than $15M to design, build, launch, and operate.
The mission was planned with the help of the folks at @astrodigitalgeo, @MattGialich and the folks at @AstroForge, and @alexwg+@matthew_pines physical AI lab PSI, who devised the final mission profile. The mission will be executed by full-time program manager @GarrettJameson, who joins us after 10 years as a US Air Force Officer, in collaboration with a US bus manufacturing partner that will be selected over the coming months.
We will need your help to complete the mission, so if the idea of the mission excites you, please check out the Fermi Explorer website in the comments to find out more about why we are doing the mission and how to get involved!
Thanks to London Studio @ZeroOneCreative and @ashkan01C for producing the concept video 🙏🙏
5
8
56
5,987
Ezra Feilden retweeted
Incredible stuff and important vote of confidence from NVIDIA on the orbital date centers category (not just SpaceX).
Congrats Philip, Ezra, and Adi!
Congrats Philip, Ezra, and Adi!
I'm super excited to share that we have raised $250M at a $2.3B valuation to build a LOT more AI compute capacity in orbit 🚀
Thanks so much to our supporters, including Manhattan West, who led the round, and new investors, @nvidia, @Cisco_Invests, @CedarCapital, Goanna Capital, ell as continued investment from existing backers @Benchmark, @EQTVentures, @Soma_Capital, @NFX, @SevenSevenSix, and others.
Next week, @Starcloud_ will move into a new 100,000 square ft facility that will enable us to ramp up our production rate to 100 satellites/week.
This is just the beginning! @EzraFeilden, @AdiOltean🏃🏻➡️🏃🏻➡️🏃🏻➡️
3
6
20
6,142
Ezra Feilden retweeted
Thinking about orbital compute a bit and whether the ladder comes up behind first movers:
There is an interesting timing dynamic here that feels similar to what happened with EVs.
Investing in orbital compute when compute is priced at a massive premium lets you work through the hardest part of the cost curve.
Others can say, “maybe it gets cheaper, maybe it doesn’t,” and keep harvesting terrestrial compute profits today.
But if/when the orbital cost curve does bend, there may be a period where early movers have already built the launch experience, operating know-how, customer relationships, and supply-chain position that make them hard to follow.
There is an interesting timing dynamic here that feels similar to what happened with EVs.
Investing in orbital compute when compute is priced at a massive premium lets you work through the hardest part of the cost curve.
Others can say, “maybe it gets cheaper, maybe it doesn’t,” and keep harvesting terrestrial compute profits today.
But if/when the orbital cost curve does bend, there may be a period where early movers have already built the launch experience, operating know-how, customer relationships, and supply-chain position that make them hard to follow.
20
13
145
53,224
There are six fully/partially reusable orbital launch vehicles in active funded development in the US: Starship, Stoke Nova, Neutron, New Glenn, Terran R, Eclipse.
Up until today we've had at most one partially reusable operational.
Tell me again that launch prices won't fall.
Up until today we've had at most one partially reusable operational.
Tell me again that launch prices won't fall.
12
1
63
4,741
Crazier still, China has 14 in active funded development. For those who will fly on Chinese rockets, pricing is going to be even lower.
1
1
3
775
Without space datacenters we're on track for a Grey Goo type doomsday scenario, where every square meter of earth's surface gets tiled by solar + datacenters. Once AI is unconstrained to expand out into space, it makes no sense to eat up earth - we're safe.
Plots of empty desert could soon be worth more than the most fertile farmland in human history.
@CJHandmer ran through the back-of-the-envelope math.
10 acres of solar is enough for 1 MW of H100s. And the FLOPS done by those H100s is roughly equivalent to 1000 human brains.
If tch human performance per FLOP, we could produce 100 humans' worth of cognition per acre of solar.
And that's just the floor. Today's chips are far less energy-efficient than the brain, so the actual number could be bigger still by >50x.
The implied land value is insane. The clear incentive would be to use up almost all the world's land for solar.
3
2
16
3,684