It’s done.
$6 billion in the bank.
And it’s going to take every dollar…
This last Sunday was a big day in the race towards artificial general intelligence (AGI).
Elon Musk’s latest new venture, xAI, announced the closing of its $6 billion series B funding round at a post money valuation of $24 billion.
It’s a massive raise for a company that has been around for less than a year. And it comes with an even more remarkable valuation.
It’s the kind of valuation that is only possible because of Musk’s demonstrable track record of building teams capable of pulling off things that have never been done before.
This raise won’t be a surprise to regular Outer Limits subscribers. I’ve written about xAI a few times, most recently in a couple of past AMAs. Now that the round has closed, the money is in, and the race to AGI has become even more interesting.
Prior to the raise, xAI had been doing a lot with limited resources.
Since its launch in July of 2023, it released Grok-1, its large language model (LLM) in November of last year. It subsequently open-sourced that same release, to critics’ surprise.
And by March 2024, xAI released Grok-1.5, which added advanced reasoning and problem solving to Grok.
It’s performance is impressive, and not far off in terms of benchmarks compared to major foundational LLMs like Gemini Pro 1.5 (Google), GPT-4 (OpenAI), and Claude 3 Opus (Anthropic).
The reason that it’s impressive is that to date, xAI was funded by Musk with $750 million of seed capital, and about $250 million of in-kind investment from X (Twitter) in compute resources.
That’s “only” about $1 billion in total, and xAI is already a player in the race to AGI.
And that’s what the additional $6 billion is for.
It takes billions of investment to purchase enough computational power to build advanced large language models and train an artificial general intelligence. And it takes money and reputation to pull in the very best talent to make it happen.
And fortunately, Musk and his company now have both.
xAI also has a mission that stands sharply apart from the rest of the field:
It’s hard to believe that a statement like this would be controversial.
For an endeavor as important to humanity as the creation of an AGI, the “rigorous pursuit of the truth” and the absence of pushing political narratives embedded within an AGI is the most logical path forward for a non-tyrannical, non-Orwellian future.
A brief poll of nearly 100,000 people demonstrated that most agree with that vision. OpenAI was a distant second, with Meta the least desirable of the four included in the poll.
And yet, one of “godfathers” of AI, Yann LeCun, chose to pick a fight with Musk’s comments above.
It was an interesting exchange because it dramatically highlighted the difference between an academic and a builder.
It also highlights the remarkable contradictions, emotionally driven comments, and corrupt political positions even intelligent people can have.
This was a terrible mistake, which resulted in LeCun embarrassing himself, contradicting himself, and ultimately revealing his heavy bias.
To which Musk replied:
And Musk was 100% correct.
LeCun’s pronouncement is saying that unless you tell us what you’ve created, published it, and allowed “us” to criticize it, it’s not science at all. That is complete and utter nonsense.
It’s an insult to every scientist, entrepreneur, builder, engineer, and inventor that has created products and services that the world values. Just because they chose not to publish their inventions, and instead focused on bringing them to life in a company, they are somehow not doing “science”?
And the most ironic part is that LeCun works for Meta as its Chief AI Scientist — the same company that has programmed its AI to push political narratives, influence, and manipulate its users, and actively pushed misinformation — disguised as “science” — during the pandemic, as well as suppressed peer reviewed scientific research. And Meta continues to do so.
As if that wasn’t enough embarrassment already, it got worse.
LeCun then tried to claim that the secret to Tesla’s autonomous driving technology is from someone else’s invention.
To which Musk disappointed LeCun, revealing that Tesla’s secret used a different approach to solving vision (note: “tbh” stands for “to be honest”).
As much as LeCun would like to know the answer of how Tesla solved one of the most challenging engineering problems — full autonomy based on vision alone — he will have to wait a lot longer to find out. The answer is proprietary to Tesla, and arguably worth more than a trillion dollars.
I would even argue that Tesla’s autonomous software, its AI, is likely the single most valuable invention in history.
As I’ve written before, Tesla’s AI is not just about self-driving cars.
It’s the key to autonomy on just about any kind of mobile hardware we can imagine.
Tesla is already focused on semi-trailers for hauling goods from point A to point B. This technology alone can solve the vast labor shortages in the trucking industry.
And as we know, Tesla has already leveraged this autonomous software as the foundation for Optimus, its bipedal humanoid robot.
The technology can be adapted to drones, delivery robots, eVTOLs, aircraft of any kind, and other home robotic assistants. The possibilities for this technology are nearly endless. And the positive impact on the global society will be immense.
Musk has always had an enormous vision for the projects that he embarks on. And he and his teams share a sense of urgency to achieve things that were previously thought impossible.
The reality is that many in the industry are now shaking in their boots. Musk now has the capital — and has no problem — attracting the best talent to build the world’s first AGI.
He won’t attract the pontificators and the hypocrites, he’ll attract the doers and the builders who are interested in the “rigorous pursuit of the truth” and won’t shy away from hard work.
Yesterday, OpenAI felt the need to make it public that it is already “training its next frontier model.” This is GPT-5, or whatever new name that it is given. Irrespective of what it will be called, it will be remarkably better than the current GPT-4o, which is already impressive.
Not surprisingly, Musk and his team at xAI are already building the “Gigafactory of Compute,” xAI’s supercomputer, which will be used to develop an unbiased artificial general intelligence.
The stated plan is to have 100,000 specialized semiconductors all connected into a massive supercomputer to achieve that task. The assumption is that the core of the supercomputer will be built on NVIDIA’s H100 GPUs, and likely the forthcoming new NVIDIA GPU Blackwell architecture.
The news of the $6 billion financing sent NVIDIA shares up about 7% yesterday, an incredible jump for a company of that size.
Musk forecasts that the Gigafactory of Compute will be completed by fall 2025. When finished, it will be about four-times the size of what Meta uses to train its own AI models.
In the short-term, xAI has the ability to rent compute from Oracle’s cloud services, X (Twitter), and Tesla, which has its massive Dojo supercomputer, which is already one of the largest GPU clusters in the world.
This competition to AGI is immensely important, and the field has already narrowed down to just five companies — six if we count Microsoft:
Apple is of course a wild card. It clearly has the financial resources to build and compete, but it is at least 18-24 months behind the crowd listed above.
As consumers, citizens of the world, and those who believe in freedom, this competition is healthy and good. Fortunately, there are enough resources for several impressive companies to take on a massive engineering challenge like AGI.
And more than one will be successful. That’s the point.
A world where one single company, owned and controlled by a very small number of people, with the only AGI on the planet — and that company programs bias and politics into its AGI — would be disastrous for global society.
But having a few companies that have created effective AGIs with immense utility — that create entirely new industries, jobs, and solve problems that could otherwise not be addressed — will affect the best possible outcome for an AGI-enabled future.
Competition and truth will weed out the foundation models that are trying to manipulate us. And the world will be better off for it — ignoring the hypocrites, naysayers, and narrative pushers. It will simply be a lot more fun to focus on enjoying the ever increasing quality of life.
Let’s go!
We always welcome your feedback. We read every email and address the most common comments and questions in the Friday AMA. Please write to us here.
