SpaceX AI Could Outpace Anthropic Within Half a Year, Musk Says

SpaceXAI CEO Elon Musk predicts AI dominance within six months, leveraging compute hardware to bridge the gap with competitors.

SpaceXAI CEO Elon Musk recently stated on X that his company aims to lead the frontier AI landscape within approximately six months.

Musk expressed cautious optimism, suggesting SpaceXAI could match Anthropic’s Fable or OpenAI’s GPT-6 within two to three months.

SpaceXAI currently provides 17 models, each varying in intelligence, performance, and pricing.

According to Artificial Analysis benchmarks, SpaceXAI’s top models are Grok 4.7 iterations, which currently trail Anthropic, OpenAI, and Meta’s Muse Spark 1.3 (max).

If our second derivative remains strong, SpaceX will reach pole position in about 6 months

Elon Musk, SpaceXAI CEO

Elon Musk, CEO of Tesla and SpaceX. Credit: Getty Images

Key facts

• Grok 4.7 scores 46 on Artificial Analysis intelligence benchmark
• Claude Opus 5.5 leads intelligence benchmark with score 58
• OpenAI’s GPT-6 Astra (max) scores 53 on intelligence
• Grok 4.7 (xhigh) ranks 16th across general benchmark
• Grok 4.7 outperforms Alibaba’s Qwen3.8 Max and Kimi K3 on a general benchmark.

Frontier speed prediction

Musk made his prediction on X, stating: “We will keep accelerating.”

He noted that SpaceXAI’s AI efforts are only three years old, compared to six years for Anthropic and ten for OpenAI. “SpaceX will reach pole position in about 6 months,” he added.

His argument centers on hardware and intelligence capabilities, noting that current intelligence suffices for many tasks. “Once you far exceed the caliber of intelligence needed for a class of tasks, additional intelligence is pointless,” he said.

“You don’t need (and it would be cruel to put) Newton-level intelligence in your toaster!”

The hardware bottleneck

Musk explained that “Hardware is hard” and “bringing massive compute online rapidly is incredibly difficult”. He added, “SpaceX has demonstrated exceptional ability in this regard and will only get better”.

The company is entering semiconductor manufacturing via the Terafab project in Texas, essential for its mission: to build “systems and technologies necessary to make life multiplanetary”, “understand the true nature of the universe”, and “extend the light of consciousness to the stars” as stated in an SEC filing.

The project aims to produce 1 terawatt of compute annually. SpaceX and Tesla are investing an initial US$16.8 billion into this semiconductor facility.

Musk previously indicated that global production capacity cannot meet his companies' internal needs.

Speaking in Austin, Texas, Musk explained: “To give you a sense of what we are talking about, the current output of AI compute is roughly twenty gigawatts per year…. All of the rest of the output from Earth is about 2% of what we need."

We will keep accelerating

Elon Musk, SpaceXAI CEO

Elon Musk announcing the TeraFab. Credit: X/Elon Musk

Grok’s performance

According to Artificial Analysis benchmarks, SpaceXAI’s top models are Grok 4.7 iterations, scoring 46 on the intelligence benchmark at the highest level.

Anthropic’s Claude Opus 5.5 (max with fallback), currently the top model, scores 58, while OpenAI’s GPT-6 Astra (max) scores 53. Excluding cost, this 7- to 12-point intelligence gap represents the performance deficit Musk’s models must overcome.

In a general benchmark ranking over 250 AI models across intelligence, price, performance, speed, and context window, SpaceXAI’s top iteration, Grok 4.7 (xhigh), ranks 16th, behind various OpenAI and Anthropic models, as well as Meta’s Muse Spark 1.3 (max).

Grok 4.7 currently outperforms some top Chinese providers, such as Alibaba’s Qwen3.8 Max (0902) and Moonshot’s Kimi K3, who often release open-weight or open-source models.

AI providers are increasingly competing on cost to keep up with Chinese firms. Both Anthropic and OpenAI have recently released models offering core capabilities of advanced models at significantly lower costs, likely to prevent enterprise customer loss to Chinese rivals.

For example, Grok 4.7 (xhigh) costs US$3.74 per task, while Xiaomi’s advanced MiMo-V2.6-Pro, with similar capabilities, costs just US$0.13 per task.

Elon Musk has told SpaceX employees he plans for xAI data center capacity to reach 10GW of AI compute by 2027. Credit: SpaceX

Key partners

SpaceXAI's key partners include:

• NVIDIA: American technology giant specializing in graphics processing units and AI chips that provide core compute power for training Grok models.

• Tesla: Elon Musk's own electric vehicle and clean energy company collaborating on joint supercomputing infrastructure, shared AI research and autonomous software development.

• Dell Technologies: Global IT infrastructure vendor assembling server racks and liquid-cooled data centre hardware to power SpaceXAI's massive AI factory supercomputers.