Elon Musk now says xAI is delaying the Grok 5 launch till next year

By William Gavin

Musk has said the AI chatbot will 'feel sentient,' and there's a chance it will achieve human-level intelligence

Elon Musk's founded his artificial intelligence startup in spring 2023. Its primary product is Grok, a large language model.

Elon Musk said Friday that the next version of his artificial intelligence chatbot will blow all of its rivals out of the water - after some extra training time.

"Grok 5, I think, will be the smartest AI in the world by a significant margin in every metric - without exception," the xAI founder said Friday in an appearance at Baron Capital's annual investment conference.

Musk had said in August that he planned to debut what he called the "crushingly good" next version of Grok by the end of 2025, but said Friday that it will instead come out within the first three months of next year. Grok 5 will launch with twice as many parameters - the settings that influence the behavior of large language models - as the current model, he said.

The new model will be both "extremely" intelligent and fast, Musk said, noting that he thinks there is a roughly 10% chance for the chatbot to achieve human-level intelligence. He added that xAI has been training Grok to be better with using tools, like searching the internet, and understanding live videos.

"I've never thought that before," Musk said of the possibility Grok could achieve what experts call artificial general intelligence. "It's really gonna feel sentient," he added.

Grok has primarily been used alongside Musk's companies, most prominently in X, which xAI acquired in March. The chatbot has also been implemented in Tesla's (TSLA) electric vehicles and powers customer-support features for SpaceX's satellite internet service. Musk also wants Grok to be embedded in Tesla's humanoid robots.

Scaling AI development comes down to a few things, including attracting the right talent and installing a lot of hardware as quickly as possible, according to Musk. As the industry has found out, that can get very expensive.

xAI's Colossus data center in Memphis, Tennessee, has at least 200,000 graphic processing units already installed, but the company plans to eventually use one million units in the facility. The startup has been burning through about $1 billion each month as it spends heavily on developing its AI models, Bloomberg reported in June.

See More: Big Tech needs a staggering $1.5 trillion to fund the AI boom. This is the complex playbook it's using to get it.

Meta Platforms (META) alone plans to have as much as $72 billion in capital expenditures this year, with most of that cash going toward AI infrastructure. That's expected to continue in 2026, although the tech giant also expects to spend a large chunk of change to compensate employees.

Earlier this year, Meta offered up pay packages worth as much as $300 million to poach top-level researchers from rival OpenAI, according to Wired. In June, Meta invested $14 billion in Scale AI and hired CEO Alexandr Wang to lead its superintelligence lab. The company cut some 600 positions in its AI unit last month, Axios reported.

Alphabet (GOOGL) expects capex of between $91 billion and $93 billion in 2025, while Amazon.com (AMZN) plans to spend $125 billion and be "very aggressive" in building data centers. Microsoft, which has partnered with OpenAI, had $34.9 billion in capex in its fiscal first quarter, $10 billion more than analysts had expected.

-William Gavin

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

11-14-25 1751ET