Cursor vs Claude Code vs GitHub Copilot: What SpaceX's $60 Billion Deal Changes

Last week, the AI coding tool that led the way on agentic coding stopped being an independent company. SpaceX closed its acquisition of Cursor, the startup behind one of the most popular AI editors, in a deal first agreed in April at $60 billion. Cursor's own blog post announcing the close is two short paragraphs of celebration and a lot of GPU talk. The Hacker News thread that followed was not in a celebrating mood.

"I'm skeptical of this deal," wrote the top commenter. "Cursor led the way on agentic coding, but Claude Code and Codex have replaced it."

That gap, between the company's framing (bigger models, cheaper compute) and the developer reaction (skepticism, jokes, and several "won't be using that again" posts), is the story. If you build software with AI every day, your editor choice is now partly a bet on which corporate empire wins. The tool that completes your code also decides which models you can reach, where your prompts go, and how hard it is to leave. That was always true. What changed last week is that the stakes became impossible to ignore.

What actually happened

The deal did not come out of nowhere. Here is the timeline, with numbers attached to each step:

• April 21, 2026. SpaceX announced it had an agreement with Cursor and an option to acquire the startup for $60 billion, per TechCrunch. The Hacker News thread on the announcement hit 823 points and 989 comments. Cursor's closing post frames that step as "our partnership with SpaceXAI to accelerate our model training efforts."
• June 16, 2026. Days after SpaceX's blockbuster IPO, the companies said they were moving forward with the acquisition, reported TechCrunch at $60 billion in stock.
• August 14, 2026. Cursor published "Cursor is now a part of SpaceX." The Hacker News thread (101 points, 120 comments when I checked) answered with skepticism, "to the moon" jokes, and naming questions: "Isn't it SpaceXAI now? I lost track," one commenter asked.
• August 15, 2026. TechCrunch confirmed the close, noting that SpaceX also acquired Musk's xAI earlier this year.

The two sentences from Cursor's post that matter most:

We will have access to the largest fleet of GPUs in the world, giving us the compute to build stronger models that are also more economical to run.

SpaceX is building the computing capacity needed to scale intelligence far beyond what exists today. Cursor will be one place where that intelligence becomes useful.

Why developers are not celebrating

The thread is worth reading in full, but four comments capture the range:

• The market-share skeptic. "Cursor led the way on agentic coding, but Claude Code and Codex have replaced it," wrote GMoromisato. "But $60B seems insane to me. What's Cursor's market share relative to Codex and Claude Code? If it's not meaningful, then it's $60B for an acquihire."
• The consolidation worry. "I feel like people aren't updating to the possibility that a significant portion of software development could be under the purview of SpaceX," wrote hardwaregeek. "Whether it's developer Twitter, Cursor's planned GitHub competitor, or Cursor/Grok, that's a lot of the developer mindshare. And that possibility is fairly concerning."
• The moat question. "Is this solely to prop up SPCX? Besides its following, does Cursor have any sort of moat? It's VS Code with a harness right?" asked stldev.
• The distribution read. "I don't see how there's anything so special about it that SpaceXAI would want to acquire it," wrote ravenstine. "It seems more likely that acquiring Cursor is a way to get more users for Grok, seeing as how it still has a very small share of the market."

The quieter comments were the sharpest. "Subscription is gone the second I feel like their not running the inference at a loss," wrote AnEro. And: "...and I won't be revisiting Cursor again," wrote midnitewarrior.

A needed dose of perspective came from imagent, who pointed out the collaboration predates the close: "they have already been collaborating on building models together. So don't expect a significant bump in performance on new Grok models due to the additional data from Cursor, that bump already happened with Grok 4.5 and now 4.6." Grok 4.6 itself was released through Cursor on August 12, which Cursor's post calls "an early look at what we can now build together."

Why the deal rattles developers: the neutrality problem

The reason this deal makes developers uneasy is not the money. It is the structure. Cursor is a store that sells access to other people's models. Its pricing page advertises "access to frontier models" and "generous limits for Grok" on the same tier. Meanwhile, its new parent sits on the other two sides of the market. SpaceX owns a model business itself (Grok, via xAI, which SpaceX acquired earlier this year), and it runs the compute business that the other model makers depend on: TechCrunch reported that SpaceX has been renting out its computing infrastructure to customers including Anthropic and Google, revenue that roughly doubled on those deals and Starlink growth.

So the chain looks like this. Anthropic and Google pay SpaceX for GPUs. Cursor's users pay Cursor for model access, and Cursor pays Anthropic and Google for those models. SpaceX's own model, Grok, competes with the same models inside Cursor's editor. One company now sits on every side of the market: the harness, one of the models, and the compute underneath the competing models. That is what people mean when they say neutrality died quietly.

The mechanism that makes this visible is default pressure. A harness with a model of its own can steer users by making that model cheaper, faster, or simply the default. Nobody has to force anything. An editor that shows you one model at a lower price is already making part of the choice for you. The thread read that clearly: if the acquisition is "a way to get more users for Grok," then the moat Cursor is building is distribution, not technology.

The second mechanism is lock-in, and it is the one Java and Spring teams should care about most. Modern AI coding tools accumulate state fast: project rules files, MCP server configs, agent memory, review history, custom commands. The longer your team uses one harness, the more your workflows live inside it. That switching cost used to be a reason to pick carefully. Now it is a reason to pick carefully while watching the owner's incentives.

The comparison: Cursor vs Claude Code vs GitHub Copilot

With that backdrop, here is how the three mainstream choices stack up right now. Pricing below is from each vendor's marketing page, current as of August 16, 2026.

• The pitch. Cursor is an AI-first editor (a VS Code fork) with agent mode, cloud agents, and a review bot. Claude Code is a terminal-first agent harness from Anthropic that works alongside any editor. GitHub Copilot is the IDE-integrated assistant built into the GitHub ecosystem, with a CLI and agent mode layered on.
• Price. Cursor: Hobby is free, Pro is $20 per month, Teams is $40 per user per month, per cursor.com/pricing. Claude: Pro is $20 per month billed monthly (less with annual billing) and Claude Code is included in every paid plan, with Max plans from $100 per month, per anthropic.com/pricing. Copilot: Free is $0 (2,000 completions and 50 chat requests per month), Pro is $10, Pro+ is $39, Max is $100, per github.com/features/copilot/plans.
• Model access. Cursor advertises "access to frontier models" plus Grok with "generous limits," and it is the one where the parent company also makes the model. Claude Code is Claude by design: one company, one model family, coherent pricing. Copilot offers a model picker, and its plans page lists access to third-party agents including Claude Code and Codex as a feature.
• Form factor. Cursor puts the agent inside the editor. Claude Code is a CLI you run in your terminal, so it composes with whatever editor you already use, and its rules live in files (CLAUDE.md) that other tools increasingly read too. Copilot is an IDE plugin plus CLI plus mobile, deepest inside VS Code and GitHub.
• Where the data goes. Cursor routes through Anysphere, now SpaceX, infrastructure, with privacy modes on team and enterprise plans. Claude Code stays inside Anthropic's platform. Copilot lives in GitHub and Microsoft, and its plans page notes that interactions from Free, Pro, and Pro+ users may be used to train models unless the user opts out.
• The lock-in profile. Cursor has the richest harness (cloud agents, a marketplace, automated review), which also makes it the most expensive to leave. Claude Code keeps your rules portable-ish in text files, but heavy usage is tied to a Claude subscription. Copilot is the most entangled with GitHub as a platform, but the least expensive to walk away from.

Who should use which

• Choose Cursor if you want the agent living inside the editor, your team runs heavy agent workflows, and you are comfortable renting that harness from a company now owned by a compute empire. If you are a Java or Spring team, check your organization's data policy before standardizing: the ownership change is days old, and enterprise contracts will take months to reflect it.
• Choose Claude Code if you want a terminal-first harness that leaves your editor independent, you already pay for a Claude plan (Claude Code is included), and you want rules-as-files you can carry between tools. The independence is the feature.
• Choose GitHub Copilot if your team lives in GitHub and VS Code, you want a free tier to start, and you want model flexibility without changing vendors. The credit model ($0.01 per credit, Max includes $100 per month in credits) keeps heavy usage predictable.

My take, with full disclosure

I build Java backends with Spring AI and run my own agent infra on a VPS: Hermes Agent, Claude Code, and custom MCP servers. Claude Code is where my agentic work happens. I have not paid for Cursor recently, so treat my Cursor observations as a reading of the public record and the HN thread, not first-hand experience. I have no stake in any of these companies.

The honest summary: the deal changes nothing about what Cursor can do today, and a lot about what it will be incentivized to do in the next two years. "Generous limits for Grok" on the Pro tier is a preview, not an accident. Whether that is good or bad depends on whether you trust a vertically integrated stack to keep selling you the competition's models at the same price.

A 7-question checklist before your team standardizes

Before your team picks, or re-picks, an AI coding tool, run this checklist. Save it, because the owners keep changing:

1. Who owns the vendor, and who owns the parent? One acquisition ago, this question was trivia. Now it determines incentives. Check the M&A history before you sign a team contract.
2. Can you bring your own models, or is the model captive? If the vendor also makes the model, ask what happens to neutral access over time.
3. What does the plan you actually need cost, and how does usage scale? Flat seat, credits, or usage-based billing behave very differently under heavy agent loads.
4. Where does your code and conversation data go, and who else can see it? Look for privacy modes, training opt-outs, and enterprise controls, and assume the answer can change.
5. How hard is the exit? Can you export rules, prompts, and agent memory and re-point your workflows at another tool in a day, or are you inside a moat?
6. Does the roadmap serve your stack? Java and Spring teams should check whether their language, build tooling, and framework get first-class support or afterthought status.
7. Re-run the checklist every six months. The ownership of these tools is changing faster than any IDE market in the history of the industry.

The bottom line

Cursor is still a good editor, and last week's news does not make the code it writes worse. But the X vs Y question is no longer just "which tool is better." It is "which tool is better, and who owns the answer?" For most teams, that points at one of the two independent-feeling options: a harness tied to a single model company (Claude Code) or a platform assistant with a model picker (Copilot). Cursor's own path now depends on how hard SpaceX pushes its model down the default path.

Have you tried Cursor, Claude Code, or Copilot? Would the owner changing hands make you switch? I read every response.

I write about Java, Spring Boot, and AI agents every week. Subscribe, it's free.