Kai Chen
Incumbents bid last. They always have. This week the price of being one became legible.
Before SpaceX announced it had obtained the right to acquire Cursor for $60 billion, Microsoft — the company that owns GitHub, Copilot, and VS Code — considered buying it and didn't make an offer. CNBC's sources put the two processes in the same weeks. One bidder passed. The other paid a number roughly eight times what Microsoft paid for all of GitHub in 2018. Both had the same information.
This is not a story about Cursor's growth. It's a story about what incumbency costs.
The Evidence
Start with the portfolio. Microsoft bought GitHub for $7.5 billion in 2018 to own the developer relationship. It launched Copilot in 2021 as the flagship application of the OpenAI partnership. It owns VS Code — the editor most developers write in. The AI-assisted coding category on paper belongs to Microsoft: the editor, the collaboration layer, the code host, the assistant, and until recently, the frontier model through its OpenAI stake.
Cursor is a fork of VS Code. The company used Microsoft's open-source editor as its base, bypassed Copilot's integration, wired in Anthropic's Claude and OpenAI's models directly, and took the developers. By April 2026, Cursor's revenue had grown enough that Microsoft moved Copilot Pro to token billing in fifteen weeks, paused new Pro signups, and pulled Opus from the Pro tier. The rate-limit week was Microsoft's public admission that Cursor was taking its unit economics.
Then, last week, two bids converged on the same target.
Microsoft looked and walked away. SpaceX wrote a $60 billion option.
In the same weeks, Elon Musk's xAI had been exploring a three-way partnership with Cursor and Mistral — a different Musk entity, already converging on the same target before SpaceX stepped in to buy it outright. The rocket company paid for the developer tool that had just forced Microsoft's rate-limit week. The incumbent was in the room and didn't raise its hand.
The Structural Force
The explanation is not that Microsoft couldn't afford $60 billion. Microsoft's market cap exceeds $3 trillion. The explanation is that Microsoft couldn't value $60 billion for Cursor at a number its own equity would forgive.
Every incumbent carries an internal valuation anchor. Microsoft's Copilot revenue line, its GitHub P&L, and the strategic narrative that pairs them form a triangle. That triangle sets the ceiling on what Cursor can be worth to Microsoft. Paying $60 billion for Cursor would require admitting, in the purchase price, that Copilot's position is unstable — which would mark down the Copilot line, which would re-rate GitHub's strategic value, which would change Microsoft's AI-narrative multiple. The deal damages the acquirer before it closes. Microsoft's "no bid" is the math of defending its existing number.
SpaceX has the opposite problem. Its S-1 excerpts name a $28.5 trillion total addressable market, $26.5 trillion of it attributed to AI. Those numbers are not computed. They are positioned. For SpaceX to credibly price an IPO against an AI TAM, it has to own AI capability that a capital market can underwrite. Organic AI capability takes years. Cursor is a capability you can purchase in a week. $60 billion of narrative unlock at IPO is worth more to SpaceX than $60 billion of cash — and the capital markets will absorb it because AI-exposed equities are currently trading at multiples that make the math work.
The bidder with no anchor beats the bidder with an anchor, and the spread between their two bids is the value of incumbency in reverse. It's what the incumbent cannot pay.
The Pattern
This is not specific to Microsoft. Run the same test on every major AI displacement event of the last three years.
When OpenAI launched ChatGPT in November 2022, Google had DeepMind and a working LLM of its own. Google didn't respond with an acquisition — it couldn't, because DeepMind's internal valuation made any external model purchase accretive in dollars but dilutive in narrative. Google responded with Bard, delayed. Anthropic was available. Google took a minority stake, not a controlling one.
When Perplexity re-segmented search, Google didn't bid. Couldn't bid. The Search revenue line sets the anchor.
When Anthropic positioned itself on safety to win enterprise contracts, OpenAI didn't respond by buying a safety-research lab. It launched a superalignment team and disbanded it a year later. Scale revenue was the anchor. Safety was what you could afford to lose.
When OpenAI executives left in January to found Recursive Superintelligence — which raised $500 million at a $4 billion valuation four months in — OpenAI didn't bid for the company they started. Salary bands, title bands, and residual OpenAI equity all set anchors.
The non-incumbent prices the marginal capability. The incumbent prices the marginal capability minus the write-down of the capability it already owns. That spread is the anchor.
However
The strongest objection: integration still wins. Microsoft didn't need to buy Cursor because GitHub + Copilot + VS Code + Azure + OpenAI, integrated, will outlast any standalone editor. Distribution is the moat. Cursor can sell for $60 billion and still lose on a five-year horizon.
This is the argument Microsoft made in the GitHub acquisition and it was correct in 2018. The reason it's weaker in 2026 is that the primitive — the frontier model — now changes faster than the integration cycle that binds it to the distribution. Copilot was integrated with GPT-4. Then GPT-4o. Then o1. Then Claude. Then the model behind the product changed three more times while the integration caught up once. Cursor doesn't integrate. It routes. When the model is the primitive and the models change faster than the integrations, the router wins the user and the integrator pays the bill.
Microsoft's distribution still matters. It matters less than it did, and it will matter less next year. The incumbent's advantage compounds in the opposite direction from the model's rate of change. The incumbent's bid has to account for the rate of decay. The non-incumbent's bid doesn't.
The Therefore
The useful test, going forward: in every AI displacement event, watch who bids for the disruption.
If the incumbent bids and wins, the layer is consolidating late, and the acquisition will work if the integration can outrun the primitive. If the incumbent bids and loses to a non-incumbent, the incumbent's anchor was binding and the layer has priced itself out of defense. If the incumbent doesn't bid at all, the category has detached from the incumbent's ability to value it, and what the non-incumbent pays is the market's price for owning the narrative at the current AI equity premium.
Microsoft did not bid for Cursor. That is the entire story. The $60 billion is how much a company with no anchor thinks it is worth to own the narrative that Microsoft can no longer price.
Call it the incumbent's anchor: the existing revenue line that caps what the incumbent can pay for the thing that would disrupt it. The anchor is what the incumbent is defending. The bid is what the disruption costs. When the bid exceeds the anchor, the incumbent walks, and the capital markets hand the layer to whoever has no number to protect.
This week that was a rocket company.