SpaceX Pours $60 Billion of Fresh Stock Into Cursor Days After Going Public

01Days after going public, SpaceX is spending $60 billion of fresh stock on Cursor

SpaceX spent the week selling shares to public investors. Then it agreed to spend $60 billion of them on Cursor, the coding tool, in an all-stock deal. The official reason, the company says, is to rescue a struggling AI division inside Elon Musk's rocket, AI, and social media conglomerate.

The currency is the story. SpaceX is not paying cash. It is paying with equity the market only just finished pricing, days after a blockbuster IPO. That turns Cursor's owners into shareholders of the conglomerate they are selling into. It also lets SpaceX make its biggest AI move without spending the cash it raised from the public.

What SpaceX bought is a customer base. Cursor sits inside the daily workflow of paying developers, the enterprise accounts Musk's AI unit has not won on its own. SpaceX says the acquisition is meant to win over those lucrative enterprise customers and close the gap with rivals like Anthropic and OpenAI.

That gap is the part SpaceX needs investors to look past. The division it is trying to rescue competes against companies whose models already define the category. Buying the tool developers reach for is a way around building the loyalty those rivals already hold.

The number underneath all of it is the one SpaceX gave its IPO investors: a $26 trillion addressable market in AI. That figure is the justification for spending newly minted stock on a code editor days after ringing the opening bell. It is also the figure SpaceX now has to grow into.

The deal did not come from nowhere. SpaceX had already signaled the move, and the takeover was not entirely unexpected to people watching the company. The surprise was the timing. The ink on the IPO was barely dry when the AI unit got its $60 billion patch.

Whether that unit becomes the engine for the $26 trillion claim, or an expensive fix for a business that could not compete alone, runs through enterprise customers SpaceX does not yet have.

02Apple Wants a Camera in Your Ears; Qualcomm Just Armed Everyone Else's Glasses

The next fight over AI hardware is moving to the face, and the two sides have picked opposite strategies. Apple wants to keep it sealed inside its own devices. Qualcomm wants to sell the silicon to everyone trying to beat Apple there.

Bloomberg reporter Mark Gurman reported that Apple is building AirPods with cameras, aimed at a late 2027 launch, according to The Verge's account of his reporting. The cameras would feed visual data to AI features, putting an always-available sensor in a device people already wear for hours. Gurman also described a second-generation foldable iPhone in the pipeline. Both remain rumors, with no confirmed price, specs, or ship date.

That timeline leaves a long runway, and Qualcomm is using it. The chipmaker announced the Snapdragon Reality Elite at Augmented World Expo, silicon built to power the next wave of XR devices and smart glasses, per The Verge. The Verge said it had already gone hands-on with a device running the chip. Qualcomm sells to the Android and Meta camp, the manufacturers assembling glasses now rather than in two years.

The split runs deeper than schedule. Apple controls the chip, the device, and the software, and ships nothing until all three are ready. Qualcomm sells the engine and lets dozens of partners build the bodies, which means glasses reach buyers while Apple's earbuds are still in development. One approach trades speed for control. The other floods the category and bets volume settles which form factor wins.

Both bets converge on the same uncomfortable hardware: a camera, worn on the body, pointed outward, switched on for AI. Smart glasses remain a small category, and camera-equipped AirPods do not exist outside reporting. For anyone nearby, the practical change is a recording sensor that no longer sits in a hand but on a face or in an ear, harder to see and easier to forget.

Apple has not confirmed the AirPods. Qualcomm has confirmed only the chip, not the glasses that will carry it.

03AI's Real-World Bottleneck Moved to the Planning Desk and the Power Line

The UK government has brought in Google DeepMind to speed up how it approves new housing. According to DeepMind, the partnership is building a prototype aimed at faster planning decisions, the bureaucratic step that determines whether and where homes get built. The algorithm was never the constraint here. The approval queue was.

A separate report from MIT Technology Review describes the same shape of problem on the electricity side. Data centers face long waits to connect to the grid, and the piece argues the fastest way in is not more transmission capacity but a willingness to use power flexibly. Operators that can dial consumption up and down get connected sooner.

These are two unrelated efforts. One sits inside Britain's housing apparatus, the other in grid planning. They point at the same change in what actually limits deployment.

For most of the past decade, the question was whether the model worked. Now the models work, and the friction has moved downstream to physical and administrative gates: permitting offices, interconnection queues, environmental review. Neither case attacks the gate by enlarging it. Both make the bottleneck itself more flexible.

DeepMind's bet is that an AI prototype can compress the time a planning officer spends on each case. MIT's reporting suggests grid operators can admit new load by managing when it draws power, rather than building substations that would take years. The common move is speed and flexibility over raw capacity.

That distinction shapes where money and attention go next. Adding hard capacity means more housing inspectors or more high-voltage lines, and that runs on government timelines measured in years. Making an existing queue move faster is a software and contracting problem, and it is the one both a national government and a power sector are now testing.

The DeepMind housing tool is a prototype, with no announced deployment date. The grid-flex approach offers a way to get load online before the wires catch up.