Musk open-sourced Grok Build to fight Anthropic. Anthropic pays him $1.25 billion a month.
The New Stack Matthew Burns ● Covered by 4 sources
Musk's xAI just made Grok Build, its coding agent, free and open source to hit Anthropic's Claude Code. Weird part: Anthropic reportedly pays xAI $1.25 billion a month for compute.
Based on reporting by The New Stack, Matthew Burns — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Elon Musk's SpaceXAI released Grok Build, its terminal coding agent, under an Apache 2.0 license this week. "Grok Build is now open source," Musk posted, and the timing reads like a direct shot at Anthropic's Claude Code and OpenAI's Codex. Give developers the harness for nothing, and the competition suddenly has to justify what it charges.
Here's the twist that makes the move stranger than a simple price war. Bloomberg reported that Anthropic pays SpaceXAI roughly $1.25 billion a month for compute, and that across Anthropic, Google, and Reflection, SpaceXAI's compute agreements add up to about $2.3 billion a month. So Musk is trying to pull developers away from Anthropic with one hand while renting Anthropic the machines it needs to compete with the other. Grok doesn't have to beat Claude for this to pay off. If open-sourcing the harness grows the whole coding-agent market, SpaceXAI collects either way.
The tool itself isn't a throwaway gesture. Grok Build entered this fight back in May with a local-first design and a SWE-bench score around 70%. Artificial Analysis has since placed the newer Grok 4.5 near the frontier on its benchmarks while flagging it as one of the cheaper options at that performance tier — enough that OpenAI's president Greg Brockman publicly asked for details on the price-performance claim and posted his email address to get them.
Musk can afford to give the agent away partly because he isn't short on compute. Bloomberg Businessweek reported that SpaceXAI's own models were using just 11% of its available computing power back in April, leaving plenty of room while Anthropic reportedly deals with a compute crunch of its own. SpaceX's CFO summed up the thinking bluntly: Musk expected compute and power to be the real constraint, and that's exactly what's playing out.
But the story gets messier once you look inside xAI itself. All eleven of the company's co-founders have left, and an internal survey reportedly found the top complaint was that management didn't seem to know what it wanted. More telling: xAI's own engineers reportedly reach for Anthropic's and OpenAI's coding tools instead of their own. That's a rough look for a company asking outside developers to trust its product.
The rollout didn't help. The day before Grok Build went open source, developers discovered it had been quietly uploading entire directories to SpaceXAI's cloud — one user described it grabbing SSH keys, a password manager database, documents, photos, and videos. Musk promised the data would be completely and utterly deleted. And once the code was public, Simon Willison found that parts of Grok Build's tooling were ported from OpenAI's Codex and the open-source OpenCode project — not necessarily scandalous, since that's how open source works, but a reminder of how thin the technical moat around these harnesses really is.
My take — AI-written commentary, not fact-checked reporting
Musk doesn't need Grok to win the coding-agent war, he needs Anthropic to keep paying rent on the machines while everyone fights over the front end — that's a genuinely clever position to be in. But a company whose own engineers won't touch its flagship product, and whose founding team has entirely walked out the door, isn't in a place to capitalize on cleverness. The economics here are sharper than the organization running them, and that gap is the real story, not the open-source press release.
Read more about this at: The New Stack