SpaceX Owns Cursor Now. Stripe Wants OpenRouter Next.

Cursor closed under SpaceX on August 14, an all-stock deal worth $60 billion. Two days later, Stripe was reportedly closing in on OpenRouter — the startup that calls itself "Stripe for AI" — for $7 billion plus. [1][2]

Both companies built their whole pitch on being the neutral layer. Pick any model, any provider, no lock-in. This week, both became subsidiaries of platforms with a very specific reason to prefer some models over others.

Two deals, eight days apart

SpaceX first got an option to buy Cursor's parent company, Anysphere, back in April, priced at $60 billion. It exercised that option June 16 and closed the acquisition August 14. [2] That $60 billion is a big number attached to a short runway: Cursor was worth $9.9 billion in June 2025 and $29.3 billion five months later, after a $2.3 billion Series D. [3] Fourteen months, roughly a 6x run, then bought outright.

Cursor's own announcement frames the deal around compute access — "the largest fleet of GPUs in the world," with Grok 4.6 cited as an early result of the partnership. [2] Not Claude. Not GPT. Grok, the model from the company whose founder also owns SpaceX.

OpenRouter's story is smaller in dollars but tighter in timing. It raised a $113 million Series B in May at a $1.3 billion valuation. [1] Three months later, TechCrunch reported Stripe is paying $7 billion-plus — north of 5x that valuation — with Bloomberg corroborating and Stripe declining to comment beyond "we don't comment on rumors or speculation." [1] Unconfirmed, but not denied. OpenRouter CEO Alex Atallah has described his own company as offering "a single access point for different systems" specifically to prevent lock-in. [1]

Source: OpenRouter

What "neutral" was actually worth

Neither company was selling neutrality as a bonus feature. It was the product.

OpenRouter routes traffic across 400+ models for 8 million users — the entire pitch is that you're not stuck with one provider's pricing, uptime, or rate limits. [1] Cursor's value to a dev team was that it didn't care which lab built the model underneath; you picked whatever benchmarked best that month and Cursor got out of the way.

That's what a client is actually buying when they adopt a "model-agnostic" tool: insurance against getting stuck with one vendor's outages, price hikes, or capability plateau. It's the same reason I tell clients not to build critical logic inside a single SaaS platform's proprietary automation layer — you want the exit door visible before you need it. The Hub cut a client's automation bill 60% by owning the integration layer instead of renting one.

Why the buyer isn't neutral

SpaceX doesn't make money selling coding tools. It makes money keeping an enormous GPU fleet — and now a rocket business — funded. Giving Cursor "access to the largest fleet of GPUs in the world" [2] is the entire deal thesis, and the model getting the press mention is Grok, not whichever lab shipped the best benchmark that week.

Stripe doesn't make money routing AI calls either. It makes money on payment volume, and AI spend is a payment category Stripe doesn't fully touch yet. Owning the router that decides which model your invoice runs through puts Stripe inside a transaction it used to only process, not shape.

Neither company has announced they'll degrade model choice. They don't need to. The mechanism is quieter: defaults shift, "recommended" tiers change, and the model that happens to run cheapest on the parent company's own infrastructure gets a nudge nobody has to admit to. That's not a conspiracy — it's just what owning both the tool and the incentive looks like.

flowchart TD
  A[Pick a tool for its neutrality] --> B["Model-agnostic routing<br/>(OpenRouter, Cursor)"]
  B --> C[Platform with its own agenda acquires it]
  C --> D["SpaceX: GPU fleet<br/>needs demand"]
  C --> E["Stripe: payment rails<br/>want AI spend"]
  D --> F[Defaults quietly favor<br/>the parent's own compute]
  E --> F
  F --> G{Still true that<br/>the tool is neutral?}
  G -->|No longer by default| H[Audit: who owns this<br/>layer of my stack today?]

The money says more of this is coming

The same week, Databricks closed a funding round that tells you capital for this exact move isn't scarce. It went in asking for $1 billion. Investor demand hit $15 billion. It settled on $5 billion at a $190 billion valuation, on $7 billion ARR growing 80% a year, already cash-flow positive. [4]

That's not a company that needed the cash. That's a company that got handed a war chest because investors wanted in regardless. When funding rounds are oversubscribed 3x on companies that don't need the money, the players who do need distribution — a payments company, a compute company — have every incentive to buy the neutral layer outright instead of competing with it.

Neutral infrastructure is just infrastructure that hasn't been bought yet.

What to actually check in your stack

None of this means rip out Cursor or avoid OpenRouter this week. Both are still good tools doing what they did last month. The check is narrower: know which pieces of your stack are "neutral" by promise rather than by ownership, and know who owns them today, not who owned them when you picked the tool.

If an agent workflow routes model calls through an abstraction layer to optimize cost across providers, check who's behind that layer before you assume the routing logic stays provider-blind. Systems that connect your own tools into one system you actually own don't have this problem — the incentive to quietly favor one vendor doesn't exist if there's no vendor in the loop deciding for you. The same logic applies to picking any infrastructure tool on a promise instead of a contract — it's the same decision rule I use for agent browsers: know exactly what you're betting on before the vendor changes hands underneath you.

Sources

[1] TechCrunch — Stripe will reportedly acquire AI gateway startup OpenRouter for $7B — techcrunch.com

[2] Cursor (Anysphere) — Joining SpaceX — cursor.com

[3] TechCrunch — Coding assistant Cursor raises $2.3B 5 months after its previous round — techcrunch.com

[4] TechCrunch — Databricks wanted to raise $1B, investors wanted $15B, it settled on $5B at a $190B valuation — techcrunch.com

[5] TechCrunch — SpaceX officially closes its Cursor acquisition — techcrunch.com

The short version

• Cursor closed under SpaceX on August 14 — a $60B all-stock deal, up from a $9.9B valuation just 14 months earlier.
• Stripe is reportedly paying $7B+ for OpenRouter — 5x its May valuation — unconfirmed by Stripe but not denied.
• Both companies sold "neutral, model-agnostic access" as the entire product.
• Neither buyer is neutral: SpaceX has GPUs to keep busy, Stripe has payment volume to protect.
• Databricks' oversubscribed $5B raise the same week shows capital for this kind of consolidation isn't scarce — expect more of it.
• Check who owns the "neutral" tools in your stack today, not who owned them when you picked them.