In April, SpaceX bought an option to acquire Cursor for $60 billion. On June 16, 2026 — four days after SpaceX’s record $75 billion Nasdaq IPO — the company exercised it. The definitive all-stock merger agreement is signed. Close is expected in Q3.
The deal was expected. The timing was a signal.
SpaceX went public on June 12 at a $1.75 trillion IPO valuation, the largest public offering in history. By June 16, after the Cursor deal was announced, SpaceX shares jumped ~16%, pushing the company past Microsoft and Amazon to the fourth most valuable in the United States at roughly $2.94 trillion. The Cursor acquisition was the first major capital deployment post-IPO, and it was made in stock — meaning SpaceX used its freshly-priced equity to buy the fastest-growing AI coding company in the market.
Cursor, for its part, is projecting more than $6 billion in annualized revenue by the end of 2026. That is a tripling of its run rate in under ten months. At $60 billion, SpaceX is paying roughly 10x forward ARR for a category leader that shows no signs of decelerating.
What the Empire Actually Looks Like#
You can now draw the full map of what Elon Musk controls in the AI coding stack:
Colossus 1 (Memphis, TN): 220,000+ NVIDIA GPUs, 300MW, the largest GPU cluster in private hands. Currently leased — in full — to Anthropic, which uses it to serve Claude. SpaceX earns infrastructure revenue from its main AI coding competitor’s primary compute provider. Musk has said this doesn’t trigger his “evil detector.” It is, at minimum, a fascinating arrangement.
xAI + Grok models: The February 2026 SpaceX–xAI merger brought Grok into the constellation. xAI’s frontier models — including Grok-Heavy, used in Grok Build — give SpaceX an in-house LLM capability that isn’t dependent on OpenAI or Anthropic.
Grok Build: xAI’s own terminal-adjacent coding agent, launched May 14, 2026. CLI-native, eight parallel agents, Arena Mode. Early benchmarks showed 70.8% SWE-bench Verified — behind Claude Code but ahead of Copilot and early Cursor models. Grok Build was xAI’s attempt to enter the autonomous coding market organically.
Cursor: 4+ million developers, the most popular IDE-first AI coding tool in the market, $6B+ ARR run rate on exit. Cursor’s Agents Window, parallel plan execution, and Composer 2.5 architecture represent the leading implementation of the IDE-embedded agent thesis.
This is not a portfolio of bets. It is a vertically integrated AI coding stack: frontier models (xAI/Grok), compute (Colossus), IDE-first product (Cursor), and terminal-adjacent product (Grok Build). The only layer SpaceX doesn’t own is a terminal-native agentic infrastructure layer — which, to be direct, is what Claude Code is.
The Landlord Problem Gets Weirder#
The SpaceX–Anthropic compute arrangement was already interesting before this deal. It is now genuinely peculiar.
SpaceX, through its Colossus lease, is the infrastructure that makes Claude Code fast enough to serve its users. The doubled rate limits Claude Code users got in May 2026? Colossus. The Mercado Libre 90% autonomous coding target across 23,000 engineers? Only plausible with Colossus-scale compute.
SpaceX just agreed to acquire Cursor, Claude Code’s primary market rival.
And yet, neither company is changing the arrangement. Anthropic’s compute contract runs through a multi-year term; SpaceX has no incentive to break it (it’s revenue), and Anthropic has no immediate alternative at that scale. So for the foreseeable future: SpaceX is simultaneously powering Claude Code and owning Cursor. The landlord is a competitor. The tenant is the market leader.
This is the strangest competitive arrangement in the history of the developer tools market, and it will eventually resolve. Whether that resolution looks like Anthropic diversifying away from Colossus, SpaceX deprioritizing the compute lease as Cursor demands more internal GPU capacity, or some other outcome is not yet clear. But the arrangement has a shelf life.
What Grok + Cursor Actually Means for Developers#
The short-term answer is: nothing changes. Cursor operates as a separate product. The Anysphere team remains in San Francisco. The deal is expected to close in Q3.
The medium-term is more interesting. SpaceX now has incentive to route Colossus compute toward Cursor inference and toward training Cursor’s next model generation. The April framing was explicit: Cursor had been compute-constrained — unable to train frontier models independently. That constraint is now structurally resolved.
Cursor’s Composer 2.5, released May 2026, benchmarked at 79.8% SWE-bench Multilingual at $0.50/$2.50 per million tokens — genuinely competitive with Opus 4.7 at a fraction of the API cost. That model was trained before Cursor had full Colossus access. The next Cursor model iteration will have it.
There is also the Grok Build question. xAI now has two coding products: the IDE-first Cursor and the CLI-adjacent Grok Build. This is not unusual in large tech companies — Microsoft has both Copilot and Azure AI Foundry — but xAI is not Microsoft. The internal model competition between Grok Build and Cursor will need resolution. The most likely outcome is differentiation by user segment: Grok Build for developers who want terminal-native autonomy with xAI-model integration; Cursor for developers who want the IDE experience with multi-model flexibility. But there will be redundancy, and organizational friction is inevitable.
The Ceiling That Didn’t Move#
The most important thing to understand about this deal is what it does not change.
Cursor’s architectural thesis is that the IDE is the right place to coordinate agentic workflows. That thesis is embodied in every product decision Cursor has made: the Agents Window, Design Mode, the /best-of-n command, the cloud development environments. It is a coherent, well-executed implementation of that thesis.
Changing ownership does not change the architecture. SpaceX acquiring Cursor does not make Cursor terminal-native. It does not give Cursor access to the file system, CI pipelines, test runners, git, or deployment scripts in the way a terminal-native agent can access them. The structural gap between what an IDE-embedded agent can touch and what a terminal-native agent can touch is not a capability gap — it is an architectural gap. It does not close through M&A.
This is worth stating plainly, because the combination of a $2.94 trillion parent company, Colossus-scale compute, and Grok-frontier models will produce a version of Cursor that is significantly better than what exists today. The benchmarks will improve. The inference will be faster. The model quality will increase.
Better IDE-embedded AI is still IDE-embedded AI. The ceiling moves up. It does not disappear.
What the Market Looks Like Now#
As of June 18, 2026, every major AI coding tool is owned by a tech conglomerate:
| Tool | Owner | Architecture |
|---|---|---|
| GitHub Copilot | Microsoft | IDE-embedded |
| Claude Code | Anthropic | Terminal-native |
| Codex + Windsurf | OpenAI | Terminal + IDE |
| Cursor | SpaceX/xAI | IDE-first |
| Grok Build | xAI/SpaceX | CLI-adjacent |
| Antigravity | IDE + Terminal |
The independent AI coding tool era is over. What replaces it is a market where your choice of coding agent is also a vote on which hyperscaler gets your compute spend, your usage data, and your workflow telemetry.
That is not inherently bad. Large companies provide stability, compliance infrastructure, and the capital to actually train frontier models. But it means the evaluations matter more than ever. Choose the tool with the architecture that matches your actual development workflow — not the one with the biggest parent company.
For teams doing serious autonomous work — agents that run overnight, multi-agent pipelines, CI-integrated review loops, spec-driven workflows — the terminal-native architecture remains the right choice. That hasn’t changed because Cursor now has a $2.94T parent.
What has changed is that Cursor will be meaningfully better in twelve months than it is today. Evaluate accordingly.
Sources:
- SpaceX to acquire the AI coding startup Cursor for $60 billion — CNBC
- SpaceX Seals $60 Billion Cursor Acquisition Four Days After Record IPO — TechTimes
- SpaceX Agrees to $60 Billion Cursor Acquisition Days After Record IPO — Bloomberg
- SpaceX Surpasses Microsoft at $2.94T Valuation After $60B Cursor Acquisition — KuCoin
- SpaceX’s $60 billion Cursor deal more than doubles its young cofounders’ net worths — Quartz
- Why SpaceX just made a $60 billion bet on AI coding ahead of historic IPO — Teslarati
- SpaceX Acquires AI Coding Startup Cursor for $60 Billion in All-Stock Deal — MLQ.ai