SpaceXAI has launched a multi-billion dollar business that will let them make $100 billion in 2027 and trillions by 2030.
Huge sustainable profits that will be many multiples of Amazon and Coreweave.
XAI renting 300MW to 445 MW of colossus 1 data center.
Dedicated AI data centers rent for ~2 to 2.4 X the cost to build. This would be $8-16B/year.
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
The other commenters are missing the bigger picture, and to be fair, this wasn’t called out in the original piece. SpaceX has two pillars in the AI space, the infrastructure business and the intelligence business. They’ve been able to deploy capacity at a scale and efficiency noone else has matched. Their plan is to continue to do so, then leverage their space technology to build orbital and eventually solar infrastructure that no competitor can approach. We already see that dynamic playing out with Starlink.
SpaceX already reached escape velocity from the rest of the industry in communications, and will pull ahead. There really is no viable competitor within the next 20-30 years and only after that if SpaceX takes their foot off the accelerator.
The intelligence component is competitive but lagging at the moment. That may change, but the field as it operates now is becoming commoditized which is remarkable for such an immature market.
There are several product innovations at the human interface level coming over the next two years which will likely accelerate AI adoption and impact. The issue is no longer the models but the scaffolding and how we interact with them. Its not clear who will pull ahead on this, but it will be a wild ride.
“They’ve been able to deploy capacity at a scale and efficiency noone else has matched”
This is not necessarily true. They for sure are over capacity. It doesn’t seem to be done with particular efficiency. To the point that their infrastructure is not justified by their own usage.
The all part of datacenter in space has already been debunked elsewhere: the issue is you either spread the computation on a swarm of satellites to have better heat dissipation (which in vacuum is terrible) but you have enormous computation lag because your datacenter Is effectively delicaluzed over the whole earth surface (in orbit but is low earth orbit so the surface of the orbital shell is comparable to the planet) or you try to reduce the distance between nodes so you end with bigger computational clusters but then heat dissipation is abysmal and limits your computation)
There is also the impossibility of doing maintenance and zero resel value because everything gets incinerated but these are peanuts compared to the physical limits
You forget the real reason to build in space. No endless decade long permit hell and NIMBY chaos
Sending stuff to space claiming is an amazing and profitable business choice is a very different proposition than doing it because of nymbism. If it is an amazing business choice and you have the tech, then you have a real competitive advantage. If it is a stupid business choice (which it is because of physics) and you are doing it for nymbisim, even if you have the tech, you do not have any significant advantage, and your very costly endeavour is exposed to the competition of countries and industries that
A) have not a bad reputation, therefore triggering less nymbism
B) have different regulations where these kinds of projects cannot be blocked
C) have a different geography where you can put the infrastructure without really angering anyone
It sounds more “we built an infrastructure but our model is not justifying the cost” solution.
Either grok is not used enough or is not cost effective enough.
It is not a good result for an AI company.
Pump & dump before the IPO most likely.
If they were on the offensive in this business they would have used the compute to accelerate the development of GROK and their own models.
Either, they gave up on GROK or they try make the numbers looks good before the IPO.
Didn’t the people who got rich in the California gold rush sell picks and shovels?
Honestly I doubt those claims. Looking at the whole picture, with electricity and other costs looks like 2 – 7 years of renting to become profitable.
That’s a great idea, especially considering SpaceX/xAI has proven they can scale data center infrastructure much faster than the industry standard. However, does this imply that future Grok improvements might be delayed since they are leasing out space originally earmarked for Grok’s expansion ?