Elon Musk plugs xAI into Tesla and Twitter

Elon Musk is wasting no time connecting the dots across his empire for his newly formed startup, xAI.

Speaking in a Twitter Spaces audio session on Friday, Musk laid out an ambitious game plan: xAI will work closely with Tesla on both silicon and AI software, while tapping Twitter's firehose of public data to train what he calls a maximally curious artificial intelligence.

Raiding Twitter's feed and Tesla's silicon

Musk made no apologies for turning to Twitter for training data, pointing out that virtually every AI company on Earth had already scraped the platform illegally. Since multiple entities were hammering the site and bringing the system to its knees, prompting recent rate limits, Musk figured xAI might as well use public tweets, images, and videos just like everyone else.

He did not say whether xAI would actually pay Twitter for that access, an interesting omission given that Twitter is saddled with 13 billion dollars in debt.

On the hardware front, Musk pointed toward Tesla's custom silicon and its Dojo supercomputer. While he wouldn't call Tesla's custom chips GPUs, he noted they offer GPU-equivalent power. Musk also claimed Tesla's Hardware 4, currently rolling out in cars, is three to five times more capable than Hardware 3, with a Hardware 5 planned in a few years to leap another four to five times.

Musk argued that frontier AI cannot just live on the web; it must understand the physical world. That is where Tesla's vast repository of driving video clips comes into play. As for the humanoid robot Optimus, Musk admitted it is still in the early stages, adding that the team still needs to ensure humans can easily switch it off.

He also took a swipe at AI safety guardrails, reiterating that xAI will avoid being politically correct and could offer answers that people find controversial even if they are true. Long term, he hopes xAI can train via self-play without relying solely on human data, similar to DeepMind's AlphaZero.

Why it matters

Musk is doing what almost no other founder in tech can do: stitching together an entire ecosystem of social data, proprietary automotive vision, and in-house silicon to fast-track an AI lab from a standing start.

We think this cross-pollination gives xAI a unique angle against OpenAI and Google. While language models trained purely on internet text are hitting data walls, xAI gets a live feed of human culture from Twitter and real-world spatial reasoning from millions of Tesla cameras.

However, the setup creates massive corporate governance headaches. Tesla is a public company answerable to shareholders, while Twitter and xAI are private ventures steered entirely by Musk. Transferring compute insights, engineering focus, and proprietary driving data across these corporate borders will inevitably raise questions about conflicts of interest.

It looks like Musk is willing to risk those messy corporate entanglements to win the AI race. If xAI successfully blends physical-world perception with raw social sentiment, it could become a formidable contender, but the regulatory and shareholder scrutiny along the way will be intense.