SpaceX’s Cursor Deal Shows Platform Control Is Being Repriced

Jordi Hays framed SpaceX’s post-market move as the dominant fact: the stock had run so hard that push notifications no longer seemed plausible at first glance. The screen showed a market-news headline saying SpaceX had overtaken Amazon in market value, with the broader U.S. indexes mixed. Hays said SpaceX was now “sitting comfortably” among the top five companies in the world.

John Coogan put numbers around the scale. Nvidia was described as being at $5 trillion; Alphabet and Apple around $4 trillion to $4.5 trillion; and Microsoft and SpaceX “neck and neck,” just shy of $3 trillion. In that context, Coogan argued, SpaceX’s $60 billion all-stock acquisition of Cursor had effectively become “free” from a market-cap perspective: SpaceX’s market cap had increased by more than four times the purchase price.

Hays said he believed SpaceX had until the fourth quarter to trigger the Cursor deal, implying the companies had built in an option window. But with the stock running, he suspected Cursor would have wanted to close as quickly as possible under whatever nonpublic terms governed the transaction. Coogan agreed with the basic incentive: if the buyer’s stock is appreciating rapidly, a seller taking stock wants to get in.

The deal was treated not just as a large acquisition, but as a category-breaking venture outcome. Coogan cited Brad Gerstner’s framing that the prior five days had produced both the biggest VC-backed IPO ever and the biggest VC-backed strategic sale ever. “We’ve never had an M&A of a VC-backed company, young startup, north of 50 billion,” Coogan said. The magnitude was easy to lose sight of because the surrounding discussion involved trillion-dollar market caps, but $60 billion was still “beyond a home run,” especially as an M&A outcome.

Hays noted that Cursor’s cap table included OpenAI, Thrive, Andreessen Horowitz, Coatue, and others. He also argued that many retail investors buying SpaceX may not have been deeply familiar with Cursor before the acquisition headlines. For those waking up to the news, the framing was straightforward: SpaceX had acquired “one of the hottest AI coding companies in the world” with billions in revenue. More institutionally aware investors, Hays said, likely already knew SpaceX had the option and would almost certainly exercise it.

The strongest skeptical read came through a Quinn Thompson post shown on screen. Thompson called the deal “brilliant corporate finance,” arguing that SpaceX was using a “newly printed, low float, retail inflated currency” to buy real businesses before lockup expiration. Coogan read the argument as a claim that the acquisition was a creative and accretive way to sell equity into an IPO pump. He also noted Nick Carter’s counter: buyers of the equity may be aware of precisely this dynamic.

That tension became the key strategic question. Coogan said Elon Musk has historically been a “build, not buy” operator, and Tesla has not behaved like Meta, Google, or Apple in its acquisition cadence. A one-off Cursor deal is one thing. But if SpaceX began doing one acquisition a month — buying “Neo Labs,” compute capacity, and “Neo Clouds,” then rolling them together — that would represent a materially different operating philosophy. Given SpaceX’s market cap and the smoothness of the Cursor transaction, Coogan said it was no longer “the craziest thing,” but it would still be a major pivot.

Hays drew on a Business Insider article about Cursor and Michael Truell to explain why the company’s story looked less like a conventional software growth curve and more like a case study in AI-platform volatility. The article shown on screen was headlined “Inside Cursor’s wild rise,” with the subhead, “How the world’s hottest AI coding company hitched its fate to Elon Musk’s rocket ship.”

The striking number, according to Hays, was that Cursor had once made up 40% to 50% of Anthropic’s revenue. He treated that as evidence of how quickly the market has been rearranging itself. At one point, Cursor looked like the front door to a large share of Anthropic’s business: user demand flowed through Cursor, and Cursor sat in the path between developers and model providers. Then the market shifted toward hyperscalers spending hundreds of millions or even a billion dollars a month.

That shift raised the deeper question of who gets to own the end customer relationship. Coogan reduced the point to that phrase. Hays connected it to Anthropic’s reported handling of Claude Code: Anthropic had allegedly told Cursor that Claude Code was “just a research effort.” Hays tied that back to Dylan Field’s question about whether Anthropic had been consistently candid. He left open two possibilities: either Anthropic had not been fully candid about its intentions, or the company genuinely believed at first that Claude Code was research, only to later conclude it needed to own that layer directly.

Coogan said the situation resembled the messaging to Figma and Canva around Anthropic’s design-tool ambitions. Hays broadened the claim: every AI company is working with a general-use technology, and that makes every layer of the software stack contestable. If a company is “in the token path,” the question is whether it can maintain value as models improve. The common advice not to build a company dependent on models getting better becomes harder to operationalize if sufficiently capable models can expand into almost any product category.

The result, Hays said, is that competition is heating up everywhere at once. Every hyperscaler now has an LLM, and every SaaS company can launch adjacent features faster. AI does not just intensify competition inside a defined category; it blurs the category boundaries.

Cursor’s story, in that reading, is both exceptional and representative. It reached a scale that could justify a $60 billion strategic sale, while also depending on relationships with model providers and compute suppliers who may be motivated to move up or down the stack. Hays’s point was not that Cursor had failed to control its destiny. It was that in the current AI market, even a company growing fast enough to become one of the largest venture-backed acquisitions ever still sits inside a moving system of model providers, hyperscalers, end-user interfaces, and infrastructure owners.

The valuation discussion moved quickly from wealth-effect headlines to whether the underlying business could support the market’s expectations. Hays cited Ryan Petersen’s post saying that with SpaceX’s 20% pop, Elon Musk had made more money in a day than Warren Buffett had made in his entire career. Coogan warned that these comparisons would get increasingly strange as the numbers moved through the trillion-dollar range. On down days, he joked, the headlines would reverse just as absurdly: Musk would be said to have lost more money than some huge national or historical benchmark.

The more substantive skepticism came through a post from an account shown as jbulltard. The post argued that the move looked like many retail investors’ first IPO: SpaceX was approaching Amazon’s valuation, the float was low until lockup, nearly every retail investor wanted exposure, and “stupid moves” were inevitable until the float opened and retail holders were stuck for years. Coogan read the post aloud, noting that SpaceX had already passed the Amazon comparison.

Hays’s immediate reaction was trading-desk bluntness: “What does that mean? Sell a hundred times.” Coogan’s response was partly dismissive and partly substantive. He joked that skeptics “haven’t seen the Mass Driver demo,” then pivoted to the Cursor acquisition and compute. Cursor had relationships with Anthropic and Google and clearly wanted compute. Coogan said that after talking to Gavin Baker, he believed “a lot more terrestrial compute” could come online in the very short term.

His bull-case structure was that the less glamorous compute and neo-cloud businesses could monetize effectively now, producing the cash flow needed for more speculative ambitions later. He compared it to the Model 3 as an economic engine: a revenue-producing product that could fund the larger vision. In SpaceX’s case, that meant terrestrial compute could help finance data centers in space, a Mass Driver on the moon, and other capital-intensive projects.

Hays then cited Ed Ludlow’s valuation comparison: SpaceX’s approximate price-to-sales ratio was 150, versus Amazon at 3.6 and Microsoft at 9.2.

Coogan said people read that kind of comparison as bearish for SpaceX. His inversion was that if Amazon traded at SpaceX’s price-to-sales multiple, it would become something like a $100 trillion company. The joke carried the underlying tension: the market was not valuing SpaceX like a mature operating company. It was valuing it as a scarcity asset, a retail-demand phenomenon, and a platform for future businesses that may or may not materialize at the required scale.

Hays introduced Snap’s new Specs through Mark Gurman’s report: $2,195 all-in-one AR glasses that Evan Spiegel had called “the next computer.” Hays’s first reaction was price. At roughly $2,200, he said, the product was approaching Apple Vision Pro territory and would be a “heavy lift.”

Coogan corrected the comparison slightly: Apple Vision Pro starts at $3,499, so there is still a meaningful gap. But he agreed that Snap is trying to compete in a different product category from a living-room VR headset. Specs are being positioned as a mobile device, something naturally worn outside the home. Vision Pro had viral launch images of people walking around with it, but Coogan treated that as the exception rather than the product’s natural context.

The question became whether Snap has the consumer base, price point, and product wedge to make expensive AR hardware work. Coogan contrasted Snap with Apple and Meta. Apple has a class of loyal, affluent customers who buy each new product line: the new iPhone, high-end MacBooks, Mac Studios, Mac Minis, and even experimental products. That meant Apple could sell some $10,000 gold Apple Watches or $3,000-plus VR headsets because a subset of the installed base would try them.

Meta’s Ray-Ban strategy, by contrast, lowered the consumer leap. The Ray-Ban Meta glasses are still recognizable sunglasses, with a camera added for a few hundred dollars more than a familiar non-camera version. Coogan said that is still giftable, and Meta can effectively afford to lose money on devices indefinitely. Hays agreed: if camera sunglasses end up unused in a drawer, the buyer feels less burned than if a much more expensive device fails to become useful.

The hosts brought in Tyler Cosgrove for an impromptu consumer test. Asked whether he would buy Snap Specs for $2,200, Tyler said they looked “really cool” and that some demoed game features, including ping pong, seemed interesting. But he also called $2,200 personally pricey. Coogan and Hays then turned the device into a challenge: if they bought a pair, would Tyler give up his phone and live with the glasses exclusively for a period?

Tyler immediately asked the practical questions: could he text, could he call, and did the glasses need a phone to do anything? Coogan looked up the battery life and said it was up to four hours, enough — in the running joke — to watch “Lawrence of Arabia” with an intermission. Tyler suggested 48 hours without a phone was not very hard, essentially a weekend digital detox; he even said he could do a week. The bit was comic, but it surfaced the actual adoption issue: if the device cannot replace core phone behaviors, the “next computer” framing will be tested against ordinary communications, battery life, and dependency on the phone.

Hays later returned to Spiegel’s product positioning. Spiegel had called the glasses both “very, very wearable” and “highly capable,” which Hays said places them between two existing categories: large, clunky headsets that are powerful, and lightweight glasses that do not do much. On weight, Hays cited Microsoft’s first HoloLens from 2016 at 580 grams, Snap Specs at 132 grams, and Meta’s Ray-Ban Display at 70 grams. Snap’s device is roughly half the weight of HoloLens-class hardware but nearly twice the weight of Meta’s glasses.

Hays said the product appears to check “a lot of the boxes” on field of view, but he emphasized the developer ecosystem as the unresolved question. He hoped a startup, perhaps a YC company, would build exclusively on top of the device and find a breakthrough interface or killer use case. But he also stressed the user-base problem: even one million units sold would not be a large platform for developers, and at $2,200 per device, getting there would require a lot of consumer spending.

The discussion of Snap’s hardware risk led into a broader comparison with Meta’s core business. Coogan read a post from Sean Frank, who said that while people were doubting Meta — advertisers frustrated by bugs and model changes, investors worried the company was missing AI or repeating the AR saga — Meta was still working for large customers. Frank described himself as a “large(ish)” customer who would spend around $100 million with Meta that year, and said no ad space came close to Meta’s scale.

Coogan’s reaction was to compare market prices rather than product categories: the fact that one could buy “one SpaceX or two Metas for the same price” was, to him, unbelievable and not something most people would have guessed a year earlier. Hays agreed that the situation was strange, while also offering the emotional explanation. Rockets are inspiring. Social media is a fantastic business, but it does not produce the same narrative force.

That distinction mattered because Snap’s AR push depends on a story about the next computing platform, while Meta’s current value still rests heavily on a massive advertising machine that practitioners continue to defend. Hays said people probably needed a “reality check” from someone like Frank, who is “in the trenches” and understands the business. The implication was not that Meta’s AR investments are irrelevant. It was that the market’s fascination with new hardware and frontier narratives can obscure the durability of an advertising platform that still works at enormous scale.

Near the end, Coogan surfaced Daniel Gerke’s provocative post that it would be “much more socially transformative” to ban social media for people over 65. Hays framed it as the inverse of the U.K.-style push to restrict social media for users under 16.

His analysis was economic rather than moral. A ban on over-65 users would, he argued, “ruffle some feathers” and likely draw more resistance from social media companies because the purchasing power of older users is high. The Boomer generation holds substantial wealth, and losing those users would be damaging for Facebook, YouTube, and even TikTok.

For younger users, Hays said platforms have an incentive to acquire people as early as possible and keep them. But if regulation sets a minimum age, the companies may mainly want a level playing field: everyone competes to acquire users when they become legally available. He described that as Connor Hays being able to “duke it out” with every other social media site as soon as someone turns 16. The details, he said, would depend on how the guardrails are set.