Musk Predicts SpaceX AI Could Overtake OpenAI and Anthropic in Six Months, Sets First Checkpoint at 2 to 3 Months

Elon Musk spent part of Thursday, September 24, arguing on X that SpaceX's AI operation is about to leapfrog companies that have been at this far longer than it has.

Responding to a user who told him "it's over for SpaceXAI" and predicted Anthropic and OpenAI would ship stronger models within a few months, Musk laid out a three-part rebuttal. "We will keep accelerating," he wrote. "Our AI efforts are only 3 years old, vs 6 and 10 years old for Anthropic and OpenAI. If our second derivative remains strong, SpaceX will reach pole position in about 6 months."

Musk separately said he is "cautiously optimistic" SpaceX will have a model on par with Anthropic's Fable or OpenAI's GPT-6 within two to three months, according to posts cited by Token Post and TeslaNorth.

The Case Musk Is Making

Musk's argument isn't just about model quality. He pointed to hardware deployment speed as SpaceX's real edge, saying "hardware is hard" but that the company has already proven it can bring large amounts of computing capacity online quickly, per Tekedia's account of the exchange.

He also made a case that raw intelligence has diminishing value past a certain point, illustrating it with a toaster: "You don't need (and it would be cruel to put) Newton-level intelligence in your toaster." The point, as Tekedia and TeslaNorth both reported it, is that once a model exceeds what a task actually requires, chasing further gains stops being the priority.

Frontier labs do face real tradeoffs between raw capability and deployment cost, and SpaceX's ability to stand up power and compute at industrial scale, honed by building rockets and satellite constellations, is a genuine asset the AI-only labs don't have.

The Numbers Behind the Claim

SpaceX folded xAI into its business earlier this year, making AI a third revenue segment alongside connectivity (Starlink) and aerospace. According to figures reported by ABABNews, the company's second-quarter total revenue was roughly $7.8 billion, with AI contributing about $2.56 to $2.6 billion, up roughly 247% year-over-year. Connectivity brought in about $4.29 billion and aerospace about $962 million.

Musk told employees at an August all-hands meeting that AI revenue would "probably surpass the rest of the company" in September and grow further in the fourth quarter, adding that AI could eventually represent roughly 99% of SpaceX's total value within four to five years, per ABABNews. He's also set a target of 10 gigawatts of compute capacity by the end of 2027, which the company has framed as implying annual revenue in the $300 billion to $500 billion range at $30 to $50 per watt of valuation.

SpaceX's Memphis and Southaven data center cluster carries a nameplate capacity of about 1.4 gigawatts. Anthropic is leasing compute there for roughly $1.25 billion a month on a 180-day initial term that either side can cancel with 90 days' notice, a structure Musk has said SpaceX built in so it can reclaim capacity if it needs more for itself. A separate deal with Google covers roughly 110,000 Nvidia GPUs for about $920 million a month, with leasing set to begin this fall.

The Competitive Bar Isn't Standing Still

Grok 4.5 launched in June 2026 on a 1.5 trillion-parameter base and entered private beta at SpaceX and Tesla. Grok 4.7 launched this week, and Musk himself has ranked SpaceX's AI arm third in agentic coding, behind Anthropic and OpenAI, according to StockTwits.

That same week, Anthropic launched Claude Opus 5.5 and OpenAI added GPT-6 Sol and Luna, per StockTwits, meaning the targets Musk is racing toward keep moving. In an earlier roadmap post cited by StockTwits, Musk said Grok 4.7 should be roughly on par with Claude Opus 5.0 but not the newer Opus 5.1, with Grok 4.9 potentially reaching the class of OpenAI's Astra and Anthropic's Fable, and Grok 5 possibly "better than anything."

Musk's "second derivative" framing assumes SpaceX's rate of improvement holds steady while two better-funded, longer-established labs also keep accelerating. The reported valuation range on SpaceX's acquisition of the coding startup Cursor, cited by Crypto Briefing as anywhere from $6 billion to $60 billion in stock, underscores how loosely some of these figures are still being pinned down.

SPCX shares slipped 0.2% Thursday, bringing the weekly decline to 3% and putting the stock on track to snap a four-week winning streak, according to StockTwits. Pre-market indications pointed to a modest 0.2% bounce heading into Friday's session, though regular trading had not yet opened as of early Friday morning.

Musk's own two-to-three-month window is the nearer test. If a Fable- or GPT-6-class SpaceX model doesn't materialize by roughly late November or December, the six-month "pole position" claim will face its first real check against the calendar.