SpaceX Gets First Dibs on Cursor. What It Means for xAI.

On April 21, 2026, SpaceX announced that xAI and Cursor are now working closely together. The deal has two parts: xAI will supply tens of thousands of GPUs from its Colossus data center to train Cursor's next model, Composer 2.5. And SpaceX has secured an option to acquire Cursor outright for $60 billion later this year, or pay $10 billion for the partnership work.

SpaceXAI and @cursor_ai are now working closely together to create the world's best coding and knowledge work AI.

The combination of Cursor's leading product and distribution to expert software engineers with SpaceX's million H100 equivalent Colossus training supercomputer will allow us to build the world's most useful models.

Cursor has also given SpaceX the right to acquire Cursor later this year for $60 billion or pay $10 billion for our work together.

— SpaceX (@SpaceX) April 21, 2026

This is not a routine cloud compute deal. This is SpaceX, weeks away from the largest IPO in history at a $1.75 trillion target valuation, buying an option on the fastest-growing B2B software company ever built. The timing, the structure, and what it reveals about xAI's position in the coding AI race all deserve a closer look.

#Cursor's Trajectory

Cursor, built by Anysphere, has become the defining AI coding tool of 2026. The numbers are staggering: $100 million ARR in January 2025, $500 million by June, $1 billion by November, and $2 billion by February 2026. That trajectory makes it the fastest-scaling B2B software company on record, ahead of Slack, Zoom, and Snowflake.

The company is in talks to raise $2 billion from Andreessen Horowitz and Thrive Capital at a $50 billion valuation, with Nvidia participating as a strategic investor. They are forecasting over $6 billion in ARR by end of 2026.

This is the company SpaceX just got an option to buy for $60 billion.

#The Deal Structure

The announcement contains two distinct components.

First, xAI will provide dedicated GPU capacity from its Colossus infrastructure to train Cursor's upcoming Composer 2.5 model. Colossus currently houses approximately 200,000 Nvidia GPUs, with plans to scale to one million. Cursor will use tens of thousands of these for training, a scale that typically requires weeks of parallel computation across trillions of tokens.

Second, SpaceX has secured an option to acquire Cursor for $60 billion later this year, with a $10 billion payment as the alternative if the acquisition does not proceed. This is not a letter of intent or a handshake. This is a structured option with a defined price and a defined breakup fee.

The $60 billion figure represents a 20% premium over Cursor's expected $50 billion post-money valuation from their current fundraise. The $10 billion alternative is effectively the price of the partnership itself: compute access, strategic alignment, and the right of first refusal.

#Why xAI Needs This

To understand why SpaceX is making this move, you need to understand where xAI actually stands.

In March 2026, Elon Musk publicly acknowledged that xAI "was not built right first time around" and announced a rebuild from the foundations up. This came after 9 of 11 original co-founders departed the company in the weeks following the SpaceX merger. Only Manuel Kroiss and Ross Nordeen remain.

The specific weakness Musk identified was coding. Grok Code Fast 1, xAI's dedicated coding model, was falling behind Claude Code and OpenAI's Codex on benchmarks that professional developers use. In response, xAI hired Andrew Milich and Jason Ginsberg from Cursor, two engineers who scaled the product to $2 billion in revenue, reporting directly to Musk.

That was six weeks ago. Now SpaceX is buying an option on the whole company.

#The Grok 4.3 Question

While this deal was being structured, xAI quietly released Grok 4.3 Beta on April 17. No press release. No blog post. No model card. No benchmarks. It appeared in the model selector on grok.com flagged as "Early Access" and locked behind the $300/month SuperGrok Heavy tier.

Musk later clarified the model is 0.5 trillion parameters, not the 1 trillion previously speculated. The new features include native video and audio APIs, batch image generation, and the ability to generate PDFs and spreadsheets from conversation.

What Grok 4.3 does not include: published benchmarks, a model card, or any independent evaluation. For a $300/month product from a company preparing for the largest IPO in history, the absence of performance data is notable. Independent benchmarks are not expected until May 2026.

Compare this to the competition. Claude Opus 4.7 leads SWE-bench Verified at 87.6%. Claude Sonnet 4.6 scores 72.7% and powers GitHub Copilot's coding agent. GPT-5.4 scores 74.9%. Grok's position on these benchmarks is unknown because xAI has not published the numbers.

When a company drops a flagship model with no benchmarks while simultaneously buying an option on the leading AI coding tool, the signal is clear: the internal product is not where it needs to be.

#What This Means for the SpaceX IPO

SpaceX has filed for what could be the largest IPO in history, targeting a $1.75 trillion valuation and aiming to raise up to $75 billion. The filing is expected to list on Nasdaq as early as June 2026.

The AI layer is central to the valuation thesis. SpaceX is not just a rocket company anymore. After the xAI merger, the pitch to investors is: rockets plus AI plus data centers, eventually in space.

But the AI layer has problems. The co-founders left. The coding product is behind. Grok 4.3 shipped without benchmarks. The company Musk said was "not built right" is being pitched as a pillar of a $1.75 trillion valuation.

The Cursor option solves multiple problems simultaneously. If SpaceX exercises the $60 billion option, it instantly acquires the number one AI coding tool, $2 billion in ARR growing to $6 billion, a world-class engineering team, and a product that developers actually use. If SpaceX pays the $10 billion instead, it still gets the GPU revenue, the strategic partnership, and the narrative that xAI's compute infrastructure powers the best coding AI in the world.

Either outcome strengthens the IPO story.

#The Coding AI Landscape

The broader context matters. AI coding is now one of the highest-stakes races in technology.

Cursor has the product and the growth. Claude Code has the deepest context understanding, consuming 33K tokens where Cursor needs 188K for the same benchmark task. GitHub Copilot has the distribution through Microsoft and VS Code. OpenAI has Codex.

Grok is not in this conversation. Not because xAI lacks compute (they have more GPUs than most competitors), but because the model and the tooling are not competitive. Developers on the Cursor forums discuss Grok Code's efficacy with measured skepticism. The consensus is that Grok is fast for quick scripts but lacks the deep reasoning and multi-file understanding that professional developers need.

This is the gap the Cursor deal is designed to close. If xAI cannot build the best coding AI internally (and the co-founder exodus and silent Grok 4.3 drop suggest they cannot, at least not yet), then buying or deeply partnering with the company that already has it is the rational move.

#Developer Sentiment

The developer community reaction to the SpaceX-Cursor partnership has been mixed.

The bullish case, articulated by some on X: an acquisition would give Grok the number one coding agent harness overnight, Cursor gets access to 500 million potential users through the X platform, and Musk acquires a team that actually ships.

The bearish case: Musk's track record with acquired engineering teams is not encouraging. The xAI co-founder exodus happened weeks after the SpaceX merger. Cursor's culture (small team, fast shipping, developer-first) is the opposite of the Musk management style that drove those departures. And Cursor already has better options for compute (they were in talks with multiple cloud providers before this deal).

There is also the question of what happens to Cursor's model-agnostic approach. Today, Cursor supports Claude, GPT-5.4, Gemini 3.1, and Grok. If SpaceX acquires Cursor, does it stay model-agnostic, or does Grok become the default? For the developer community that chose Cursor precisely because it lets them pick the best model for the job, a Grok-locked Cursor would be a dealbreaker.

#The $10 Billion Question

The most interesting number in this deal is not the $60 billion acquisition option. It is the $10 billion alternative.

$10 billion for a compute partnership and strategic alignment is an extraordinary sum. For context, that is more than the entire annual revenue of most enterprise software companies. It values the partnership itself at roughly 5x Cursor's current ARR.

This suggests SpaceX believes the compute deal alone, training Composer 2.5 on xAI's infrastructure, creates enough strategic value to justify a $10 billion price tag. It also means Cursor has extraordinary leverage: they get either a $60 billion exit or $10 billion in partnership value, plus they keep their independence.

For xAI, the $10 billion scenario still accomplishes critical goals: it generates massive revenue to offset the cash burn from Colossus, it ties Cursor to xAI's infrastructure, and it gives SpaceX a compelling AI narrative for the IPO roadshow.

#What Happens Next

The most likely near-term outcome: SpaceX pays the $10 billion, uses the partnership for IPO optics, and revisits the acquisition after the IPO when it has public market currency to deploy.

The acquisition option likely has an expiration window, probably Q4 2026. If Cursor's $2 billion fundraise closes at $50 billion as expected, exercising the $60 billion option represents a 20% premium, which is modest by tech acquisition standards. But if Cursor hits its $6 billion ARR target and the next valuation mark is $80 billion or higher, the $60 billion option becomes a bargain.

Musk is not buying a coding tool. He is buying time. Time for xAI's rebuild to produce results. Time for Grok to close the gap with Claude and GPT. And time to go public with an AI story that has substance behind it, even if that substance was built by someone else.

The question for developers: does this make Cursor better, or does it make Cursor Grok?