SpaceX Just Paid $60 Billion for an IDE. The Coding Tool War Is Over.

I uninstalled Cursor six months ago. Moved my entire workflow to Claude Code in the terminal, never looked back. At the time, people told me I was being dramatic. "Cursor is fine," they said. "It's just a VS Code fork with better autocomplete."

This week, SpaceX secured the right to acquire that VS Code fork for $60 billion.

That's more than Microsoft paid for GitHub ($7.5B), Salesforce paid for Slack ($27.7B), and Adobe attempted to pay for Figma ($20B). Combined. For an IDE. Or more precisely, for 2 million developers who use that IDE every day and the model distribution channel they represent.

I wasn't dramatic enough.

The SpaceX-Cursor Deal Nobody Is Framing Correctly

Reuters confirmed the structure: SpaceX (merged with xAI) holds an option to acquire Anysphere, the company behind Cursor, for $60 billion. The alternative is a $10 billion collaboration deal. Either way, Cursor gets access to xAI's Colossus supercomputer. Musk's stated goal: "the world's best coding and knowledge work AI."

Most of the coverage is debating the price tag. Is $60B reasonable for a company with $2B ARR? Can Cursor justify 30x revenue? Is this a bubble?

Wrong questions. All of them.

The price isn't about Cursor's revenue. It's about Cursor's distribution. Two million developers, many of them writing production code at companies that matter, have Cursor installed. They open it every morning. They trust it with their codebases, their credentials, their security posture. That daily habit is what costs $60 billion. Musk isn't buying an IDE. He's buying a channel to put Grok in front of engineers who never would have chosen it voluntarily.

The Model Swap Nobody Is Talking About

Here's the part that should make every Cursor user uncomfortable.

Cursor's magic was never the editor. It was the models. Cursor routes to Claude (Anthropic) and GPT (OpenAI) on the backend. The autocomplete, the chat, the multi-file edits that feel like having a senior engineer looking over your shoulder. That quality comes from the models, not from the VS Code fork wrapping them.

Now think about who just bought Cursor. Elon Musk. The guy who sued OpenAI. The guy whose company, xAI, is Anthropic's direct competitor. The guy who described the deal as creating "the world's best coding and knowledge work AI," meaning Grok, meaning xAI's model, not Claude, not GPT.

When does Cursor switch from Claude to Grok as the default?

Not if. When.

This already happened once at a smaller scale. Cursor shipped Composer 2 powered by Kimi K2.5 as the default model without telling users. The response from the community was immediate and brutal: "What happened to Claude?" "Why does autocomplete feel worse?" "Did something change?" Something had changed. The model powering their daily tool had been swapped, and quality dropped noticeably. Users figured it out within days.

Now imagine that swap at the scale of an acquisition. xAI didn't pay $60 billion to keep running a competitor's model as the default. Colossus isn't being offered as a collaboration gesture. It's the infrastructure for Grok to replace Claude inside Cursor. Gradually at first. Then all at once.

Three Walled Gardens for AI Coding Tools

Six months ago, the AI coding tool landscape had genuine diversity. Cursor used Claude and GPT. GitHub Copilot used GPT and Claude. Various open-source tools used whatever model you pointed them at. Developers had choices. Models competed on merit.

That era is ending.

Here's where we're headed:

Wall 1: Musk's empire. xAI builds Grok, SpaceX acquires Cursor, Cursor becomes the Grok-powered IDE. Two million developers, one model provider, no choice.

Wall 2: OpenAI's empire. Codex, GPT, the oh-my-codex ecosystem. OpenAI released a plugin that runs inside Claude Code, an aggressive move to pull developers into their orbit regardless of which tool they use.

Wall 3: Anthropic's empire. Claude Code, Claude models, the context management system I use every day. Anthropic already sent legal threats to OpenCode (134K stars) over OAuth usage in March. The message was clear: use our tool, our way.

This isn't competition. Competition means I can switch models inside my tool when one gets better. This is fragmentation. Your coding tool now comes with a political alignment, a corporate agenda, and a model you didn't choose.

I'm Inside One of These Walls

I should be honest about my own position here. I use Claude Code. I use it 10 to 14 hours a day. My publishing pipeline, my memory engine, my agent orchestration, all of it runs through Anthropic's tool with Anthropic's model.

If Anthropic ships a bad model update tomorrow, my CI breaks. If they decide to lock down the API further, like they did with OAuth in January, I eat whatever constraints they impose. If Claude Code's pricing doubles, I pay it or I rewrite everything.

I picked a wall. I picked it because Claude is the best model I've used for the kind of work I do. But I picked a wall nonetheless. And I'm not going to pretend that Anthropic's walled garden is somehow more virtuous than Musk's just because I live in it.

The difference, for now, is that I chose Claude Code knowing what model it ran. Cursor users are about to have that choice made for them.

What $60 Billion Actually Buys

Here's what's strange about the valuation. Cursor has $2B in annual revenue but runs negative margins on consumer accounts. Every free-tier user costs them money. The product's retention is strong, but retention of a subsidized product isn't the same as retention of a profitable one.

So what does $60B buy?

It buys the habit. Engineers don't switch tools easily. I know this because it took me three weeks to fully transition from Cursor to Claude Code, and I was motivated. Most engineers aren't. They'll grumble about Grok being worse than Claude. They'll write angry Hacker News comments. Then they'll keep using Cursor because switching costs are real and their muscle memory is in that editor.

It buys the data. Every keystroke, every codebase indexed, every autocomplete accepted or rejected. That's training data for Grok's code model. Two million developers generating continuous feedback on what good code looks like, what edits get accepted, what suggestions get rejected. That dataset is worth more than the revenue.

It buys the enterprise foothold. Cursor's security agents were one of their strongest differentiators for enterprise buyers. Now those security features come with xAI branding. Enterprise decisions about AI coding tools were already hard. Now they're political.

The Uncomfortable Question

I've been writing about tool dependency risks for months. The theme is always the same: the tools are getting better, but the dependency is getting deeper. Your workflow, your productivity, your ability to ship, all of it sits on top of decisions made by companies whose interests don't always align with yours.

This acquisition is the most vivid example yet.

Two million developers built their daily practice around Cursor. They learned the shortcuts. They configured the settings. They let the token anxiety loop train them into checking autocomplete before thinking. And now the company that owns their daily tool is the same company that runs X (formerly Twitter), has a CEO who starts public fights with regulators, and views Anthropic and OpenAI as enemies to be defeated.

Nobody asked those developers if they wanted Grok. Nobody consulted them on the acquisition. The tool they chose is about to become a tool that chose them.

What I'd Watch For

I don't know when the model swap happens. Could be six months. Could be a year. But there are signals.

Watch Cursor's model selector. Right now you can pick Claude, GPT, or other models manually. If that selector starts defaulting to Grok, or if Claude disappears as an option entirely, that's the inflection point.

Watch the terms of service. Acquisition-driven ToS changes are how companies quietly expand data rights. Every codebase Cursor has indexed becomes relevant when the new owner builds training pipelines.

Watch the governance gap. Most companies have no policy on which AI coding tool their engineers use. That was fine when all the tools ran roughly the same models. It's not fine when your tool is owned by a company with specific geopolitical positions and your competitor's tool is owned by a company with different ones.

And watch what happens to Roo Code, which shut down the same week this deal was announced. Indie AI coding tools can't compete with companies spending $60 billion on distribution. The market is consolidating around three players, and if you're not backed by one of them, you're dead.

Where This Leaves Me

I'm still in Claude Code. I'm still betting on Anthropic's model quality over the alternatives. That bet has been right so far.

But I'm also building with more awareness that my workflow depends on a vendor who could change direction at any time. Anthropic's current strategy is vertical integration. Claude Code is their biggest product at 54% of enterprise revenue. They have every incentive to keep it good.

Incentives change. Companies get acquired. Strategies shift. I wrote about this yesterday: the tools are not neutral. They come with dependencies, assumptions, and risks that most developers don't think about until the rug gets pulled.

SpaceX just paid $60 billion for the right to pull that rug under 2 million developers. The coding tool war isn't about features anymore. It's about who owns the channel between a model and the engineers who use it.

I picked my channel. If you're still on Cursor, you should think about whether Elon Musk picked yours for you.