Elon Musk Is Building 'Colossus', A New AI Supercomputer

[Lynyrd Skynyrd]

Musk's new project,  dubbed Colossus and built by his new AI startup, xAI, has been created to train the latest version of the GROK language model, known as GROK-3. It's powered by no fewer than 100,000 Nvidia H100 GPUs.
If that's not enough for you, in an X post Musk says Colossus will double in power "in a few months" thanks to the addition of another 50,000 H200 Nvidia chips, which each pack roughly twice the AI acceleration performance of an H100 GPU.
It's not clear how much this is all costing Musk and xAI. Estimates of pricing for Nvidia's H100 GPUs vary from $20,000 to as much at $90,000 a pop. Presumably, Musk managed to get a comparatively decent deal buying 100,000 of the things in one go.
But even at the lower estimate, you're looking at $2 billion for the Nvidia chips for phase one, let alone building the datacenter, all the relevant infrastructure, staffing up, and doing all the work involved in setting up training for an advanced LLM. Oh, and whatever those other 50,000 H200 are costing on top as a little light frosting.
Indeed, it was only a few weeks ago that xAI launched GROK-2 as an exclusive-access thing for X subscribers. GROK-2 apparently made do with a piffling 15,000 H100 chips for training, the poor deluded little AI dear. And yet by some measures, GROK-2 ranks second and only behind ChatGPT-4o in the LLM league tables.
So, even the first phase will be six to seven times more powerful than GROK-2, only to supposedly double in power a few months later. Clearly, Musk has his sights set on building the most powerful LLM out there.



​​https://www.yahoo.com/tech/elon-musks-monster-wakes-xai-135759910.html

[midicon]

Musk's new project,  dubbed Colossus and built by his new AI startup, xAI, has been created to train the latest version of the GROK language model, known as GROK-3. It's powered by no fewer than 100,000 Nvidia H100 GPUs.
If that's not enough for you, in an X post Musk says Colossus will double in power "in a few months" thanks to the addition of another 50,000 H200 Nvidia chips, which each pack roughly twice the AI acceleration performance of an H100 GPU.
It's not clear how much this is all costing Musk and xAI. Estimates of pricing for Nvidia's H100 GPUs vary from $20,000 to as much at $90,000 a pop. Presumably, Musk managed to get a comparatively decent deal buying 100,000 of the things in one go.
But even at the lower estimate, you're looking at $2 billion for the Nvidia chips for phase one, let alone building the datacenter, all the relevant infrastructure, staffing up, and doing all the work involved in setting up training for an advanced LLM. Oh, and whatever those other 50,000 H200 are costing on top as a little light frosting.
Indeed, it was only a few weeks ago that xAI launched GROK-2 as an exclusive-access thing for X subscribers. GROK-2 apparently made do with a piffling 15,000 H100 chips for training, the poor deluded little AI dear. And yet by some measures, GROK-2 ranks second and only behind ChatGPT-4o in the LLM league tables.
So, even the first phase will be six to seven times more powerful than GROK-2, only to supposedly double in power a few months later. Clearly, Musk has his sights set on building the most powerful LLM out there.



​​https://www.yahoo.com/tech/elon-musks-monster-wakes-xai-135759910.html

You have got to hand it to Musk, he is like one of those comic super heroes! He doesn't just think big, he goes and does it!
I don't really understand all the financial aspects of his projects but he is certainly a rare individual with a lot to offer humanity. We can't say that about many people. I think he is an amazing guy.

[Quantum12]

Yes, you have that right! He does what he says. That is rare!

You have got to hand it to Musk, he is like one of those comic super heroes! He doesn't just think big, he goes and does it!
I don't really understand all the financial aspects of his projects but he is certainly a rare individual with a lot to offer humanity. We can't say that about many people. I think he is an amazing guy.

[Maxmars]

Oooh, oooh!  I've never had a contextually-relevant place to post this thought before...

Such a theoretical AI, one with the kind of processing speed as Grok3 might sport, could be used to all manner of absolutely fantastic tasks...

For example:   

A complete analysis of the tax-law code, including times and conditions, to demonstrate that the IRS' application of "Tax" law is perfectly fair and balanced , that there are no inconsistencies or schemes embedded within it for 'some' to use and to be denied (or not even made available) to others...  using symbolic logic and, and an LLM once could find exactly "when" and "where" things got "unjust" and or "exploitative"... knowing that you can deduce the "who" ushered the code, and then you could infer the "why."

or 

A complete and comprehensive analysis of US juris prudence, the where's and when 'exceptions' were employed for a similar end as above... or perhaps the actual practices of 'prosecutions' to be represented by the tools of legal contrivance, the posture of prosecutorial judgement... it would lay bare the reality of how and in what way the law has become a tool of something other than justice.

or

A complete and comprehensive analysis of US legislation, it's architecture and evolution.  The analysis of "content" and "design" alone would reveal exactly the chain of authorship, and it would be easy to see, in detail, the evidence of an architect "outside" the electorate.  The legislatorial process has undeniably degraded over the last 10 generations of political "party" identities.  Such a large scale distillation of 'proposals,' 'content,' 'timing,' and 'effect' would be very telling about exactly what it might be that has "gone wrong."



Any of several "legal" domains could benefit from an unbiased "logic enema," such as could be carried out be a sober and cogent tool... contract law, patent law, the "law" and "medicine."  There are very many who might agree if I were to proclaim that juris prudence is broken, now... 

But who am I kidding... the system isn't 'broken.'  It is doing exactly what they want it to do (it is their design)…  And they would rather spend billions to "insert" their own biases into an "AI" model to conceal it, rather than face the hard cutting fire of a real analysis.

It's a goddamn shame that "AI," such as it will manifest, will be exactly what the wrong people want it to be.  And it will be "used" in the manner they plan.

[Velvet Elvis]

I'm really disappointed he's not calling it Deep Thought. What a missed opportunity. Why Elon why?

[Karl12]

The chap looks to be 'positioned in place' and there are some uncomfortable facts for fans here:


https://corbettreport.com/musk/