Load up on coding: Why Elon Musk's $60 billion pre-IPO deal for SpaceX shows foresight

About 75% of OpenAI’s $2 billion monthly revenue comes from consumers, many of whom pay about $20 a month for ChatGPT. But those consumers could also be costing Sam Altman’s company somewhere between $1 billion and $3 billion thanks to the rising price of running larger and more complex AI models. OpenAI expects to burn through $115 billion of cash by 2029.

Musk is facing similar costs for Grok. Its developer xAI, which Musk merged recently with SpaceX, hasn’t disclosed the chatbot’s user numbers, instead combining them with X [formerly Twitter] to claim “600 million” monthly visitors. Even if Grok alone had 200 million regular users, based on recurrent running fees and a rough estimate that could be costing the company close to $300 million a month.

The solution has been in plain sight for some time: business customers, which AI labs can charge more. ChatGPT Enterprise, for instance, costs roughly $45 to $75 per user monthly with a minimum 150 accounts, meaning a single small or medium-sized company is worth something closer to $100,000 in annual revenue versus $240 for a single consumer.

No wonder OpenAI has shifted to aggressively targeting enterprise customers to win back some ground from Anthropic in that market. The latter’s revenue run-rate—a projection of yearly sales based on short-term data—more than doubled between February and April to $30 billion, driven heavily by Claude Code, which is popular with corporate tech teams.

SpaceX, through its ownership of xAI, has access to Colossus, one of the world’s most powerful supercomputers. It contains a cluster of tens of thousands of GPUs that, thanks to the Cursor deal, can now be put to more lucrative use.

Though the multi-trillion-dollar Generative AI boom is founded on the belief that it will transform business, that is only happening in two areas currently: coding and customer support, with coding taking the lion’s share of use.

An analysis of enterprise AI usage earlier this month by venture-capital firm Andreessen Horowitz said coding was “the dominant use case for AI by nearly an order of magnitude,” and cited the explosive growth of tools like Cursor, along with competitors Claude Code and OpenAI’s Codex.

By snapping up Cursor, Musk could not only give SpaceX more of the AI-valuation glow it already gleaned from merging with xAI, it can also boast one of the most popular AI coding tools to put him head-to-head with OpenAI and Anthropic. Musk seems to have realized in the past few months that he needed to pivot. In March, xAI hired two former Cursor executives to help refocus its coding efforts. Musk posted, “xAI was not built right first time around, so is being rebuilt from the foundations up.”

With Colossus, Cursor can also build its own generative coding model—where the AI can automatically create, suggest or complete source code—so it doesn’t have to rely on Anthropic’s Claude or OpenAI’s latest GPT model.

That distinction will be sorely needed as both the latter firms quickly encroach on Cursor’s niche of providing an integrated developer environment, tools and the slick user interface that make it feel like a ‘vibe coding’ app that can write code through natural language prompts.

Grok needed a more financially plausible use case beyond chatting and sexting with AI companions. Coding—assuming xAI’s chatbot absorbs Cursor’s features into its own interface—is one of the few that can justify the computing costs. If AI is going to eventually pay for itself, it will initially be through tools like coding assistants, but the question remains if Musk can do it cheaply enough to compete. ©Bloomberg

The author is a Bloomberg Opinion columnist covering technology.