- Elon Musk predicts AI will outsmart humans by the end of 2026 — earlier than he first expected.
- He said in an X interview that the "world's smartest people" are entering AI as the sector booms.
- But data scarcity, a GPU shortage, and electricity demands pose some obstacles to AI development.
Elon Musk is once again betting that artificial intelligence will outsmart us all — and he expects that to occur sooner than he initially thought.
"My guess is we'll have AI that is smarter than any one human probably around the end of next year," Musk said during a live interview with Norges Bank CEO Nicolai Tangen on X, formerly Twitter, earlier this week that's now an episode of his podcast "In Good Company."
The Tesla CEO added that the "total amount of sentient compute" — a concept that may refer to AI thinking and acting independently — will exceed all humans in five years.
The tech billionaire's timeline for AI exceeding human brainpower appears to have shortened as companies like OpenAI, Google, and Microsoft race to build the best rival products. In March, Musk wrote on X that AI will be smarter than "all humans combined" by 2029. That same month, he shared another prediction on when machines will outpace human brainpower, pushing the year to 2030.
Now, the business mogul suggests that AI could surpass human intellect by 2026 partly because of the sheer number of talent in the sector.
He says that the "world's smartest people" such as physicists, have moved into AI because of how quickly the industry advances, which could speed up development. (Coincidentally, Musk says OpenAI has poached some of Tesla's AI engineers.)
"We're seeing a lot of the best talents going into AI," Musk says in the interview.
But the tech CEO says hardware constraints may get in the way of rapid advancement. Last year's constraint, Musk says, was the dwindling supply of chips needed to power AI training efforts. This year's challenge, he says, will be transitioning to a "voltage transformer supply" or getting the total amount of electricity needed to power AI systems which he implies may be less widely available in the next one to three years. And on top of that, companies are running out of quality data to train their models.
Musk expects to encounter these challenges in making Grok, an AI chatbot backed by his startup xAI, smarter than its rivals. While the tech leader claims Grok is better than OpenAI's ChatGPT, he says the current "limiting factor" to making his model even more powerful is the massive amounts of computing needed to train it.
He says he's trying to get his hands on roughly 20,000 GPUs, Nvidia's highly coveted AI chips, by May, which he claims is doable. But for Grok 3, which would be xAI's most advanced model, Musk says that 100,000 GPUs are required for additional training. That could be a major roadblock that data scarcity would amplify.
Musk pointed to synthetic and real-world video data as a potential alternative training source. But it's not just his company working to bypass data walls. OpenAI, a firm that the business leader helped start, is discussing feeding its upcoming model GPT-5 on YouTube video transcripts to prepare for a potential information shortage, sources told The Wall Street Journal. Google, too, appears to be hitting a wall. In February, the search giant struck a $60 million dollar-a-year deal with Reddit, a public forum, to train its AI models on user posts which critics say could lead to troubling outputs.
But despite the training obstacles, Musk seems convinced that AI is advancing at a fast enough rate to reach artificial general intelligence in the next few years, though some experts don't think that will happen anytime soon.
Daphne Koller, a former computer science professor at Stanford and MacArthur "genius," said on a January panel that machines won't be as smart as humans for a long time, as they still have a lot to learn.
"We're only starting to scratch the surface of the data that are available," Koller said.
Musk didn't immediately respond to Business Insider's request for comment before publication.