1. One of the Most Expensive Early Exits in Bankruptcy History
On April 21, 2026, SpaceX announced it has secured the right to acquire Anysphere — the company behind AI coding tool Cursor — for $60 billion, or alternatively to pay a $10 billion strategic partnership fee if the full acquisition does not proceed. The deal, driven by Elon Musk's ambition to close the gap with OpenAI and Anthropic in the AI coding tools market through his xAI initiative, retroactively transformed a routine bankruptcy asset disposal into one of the most striking missed recoveries in modern corporate history. In April 2023, the FTX bankruptcy estate sold a 5% stake in Anysphere for $200,000 — the exact price Alameda Research had paid when it led the company's pre-seed round a year earlier. At SpaceX's $60 billion acquisition price, that same 5% position would be worth approximately $3 billion today, representing a return of roughly 15,000 times the sale price, realized entirely by whoever purchased the stake from the bankruptcy estate rather than by the creditors the estate was tasked with making whole.
2. How Alameda Got Into Cursor
In April 2022, Alameda Research invested $200,000 in Anysphere's $400,000 pre-seed funding round, acquiring approximately 5% of the company at an implied valuation of $4 million. Alameda was one of a small number of early backers that included Neo, the Y Combinator-affiliated seed fund, and Heroic Ventures. At the time, Anysphere was a pre-product startup founded by four recent MIT graduates — Michael Truell, Sualeh Asif, Arvid Lunnemark, and Aman Sanger — who were building what would eventually become Cursor, the AI-native code editor. The $200,000 was a small, early-stage venture bet in the context of Alameda's portfolio; the kind of speculative seed position that institutional venture funds make dozens of times a year and that most do not expect to return any value.
Eight months after the investment, FTX collapsed in November 2022 following revelations that Alameda had been drawing on customer deposits to fund its operations. Both entities entered bankruptcy, and court-appointed estate administrator John Ray began the process of liquidating assets to fund creditor recoveries. The Anysphere position was sold in April 2023 for the same $200,000 Alameda had paid — at cost, with zero realized gain, as part of what the estate characterized as disposing of non-strategic venture positions during an early phase of asset monetization.
3. What Cursor Became in the Three Years After the Sale
Cursor launched its AI coding product in early 2023, the same month the estate sold the stake. The product's growth from that launch to its current market position is among the steepest revenue trajectories in enterprise software history. The company hit $1 million in annual recurring revenue in October 2023, $100 million ARR in January 2025, $500 million ARR by June 2025, and $1 billion ARR by November 2025. As of February 2026, Cursor reported $2 billion in ARR and is guiding toward $6 billion by year-end. By April 2026, the platform had penetrated 67% of Fortune 500 companies and registered 18% usage in the Stack Overflow 2025 Developer Survey — the fastest IDE adoption in software history.
The funding rounds that followed the estate's exit reflect this trajectory directly. A Series A in August 2024 at a $400 million valuation, a Series B in December 2024 at $2.6 billion, a Series C in June 2025 at $9.9 billion, a Series D in November 2025 at $29.3 billion, and then SpaceX's $60 billion option earlier this week — each round representing the continued compounding of a product market fit that was just beginning to become visible in April 2023.
4. The Dilution Question and the Real Number
The $3 billion headline figure carries a mathematical caveat worth acknowledging. Startup equity positions are routinely diluted across successive funding rounds as new investors purchase shares, and a pre-seed position that represented 5% of the company at a $4 million valuation would have been diluted through each subsequent financing unless the holder participated in the subsequent rounds. The FTX estate, having sold the position in 2023, did not participate. Depending on how aggressively subsequent rounds diluted early holders and what anti-dilution provisions, if any, were attached to the original investment, the actual percentage of Anysphere attributable to the original Alameda stake could be meaningfully lower than 5% by the time SpaceX's option was announced. The $3 billion figure should therefore be treated as the maximum implied recovery based on the original stake percentage and the headline acquisition price — the actual forgone value depends on cap table evolution that is not publicly available.
Even at significant dilution, however, the gap between what was realized and what was forgone is enormous. A position diluted to 2% of the company at the $60 billion valuation would still represent $1.2 billion in missed recovery, compared to $200,000 received.
5. Sam Bankman-Fried's Argument From Prison
The Cursor story lands in the middle of Bankman-Fried's ongoing effort from federal prison to argue that the FTX bankruptcy administration destroyed value that should have accrued to creditors. Serving a 25-year sentence, Bankman-Fried has been active on social media and in correspondence, sharing analyses of FTX's former portfolio that he argues demonstrate the scale of the lost upside. In February 2026, he circulated a projection suggesting that FTX's net asset value would have reached $78 billion if the estate had held assets through the subsequent market recovery rather than selling in 2023 and 2024. That projection has been disputed by bankruptcy professionals who note the legal and practical impossibility of holding customer assets through a multi-year liquidation process — creditors were owed repayment, not participation in a venture portfolio.
The Cursor stake is the cleanest single example in Bankman-Fried's argument because its mathematics are simple, its timeline is unambiguous, and the trigger event — SpaceX's acquisition announcement — happened at a specific, documented moment with a specific, documented valuation. FTX creditors have been made whole in nominal dollar terms under the bankruptcy's distribution plan, receiving their claim values plus interest. What they did not receive was the upside from what the estate's assets became during the three years between the bankruptcy filing and April 2026. The Cursor stake alone represents approximately $3 billion in that forgone upside — a number large enough that it is likely to feature prominently in whatever legal proceedings or pardon advocacy Bankman-Fried pursues going forward, while prediction markets currently place the probability of a 2026 pardon at only 5%.
6. The Broader Portfolio: A Pattern the Cursor Sale Exemplifies
The Cursor position is not the only FTX estate asset whose subsequent appreciation has drawn scrutiny. The estate's sale of FTX's Anthropic stake across 2024 for approximately $1.3 billion is perhaps the other most discussed example: that position would be worth approximately $30 billion at Anthropic's February 2026 valuation, based on community analyses of the original investment size and subsequent dilution. The theoretical portfolio value of FTX's former venture positions — Anthropic, Cursor, Solana at its various price levels, Robinhood, and others — at 2026 market prices has been estimated in the range of $52 to $86 billion depending on assumptions about dilution and timing. The FTX estate realized a fraction of that through its actual liquidation timeline. The question of whether a bankruptcy estate has a legal or fiduciary obligation to hold assets for maximum long-term value recovery, versus liquidating promptly to repay creditors, is one that bankruptcy law has generally answered in favor of the latter — but the magnitude of the gap between what was realized and what was forgone in FTX's case has made it a particularly vivid case study.