SpaceX Offers $60B for Cursor and Redwood Materials Cuts Another 10%

The biggest story this week is SpaceX's option to buy Cursor for $60 billion. That's not a typo. SpaceX announced a partnership to build "next generation coding and knowledge work AI" and embedded a clause that lets it either pay Cursor $10 billion for the work or acquire the company outright for $60 billion later this year.

The deal gives Cursor access to Colossus, SpaceX's 200,000-GPU supercomputer, and positions Elon Musk's space company as a serious contender in the AI coding race. Business Insider notes this is part of Musk's broader push to turn SpaceX into an AI player ahead of a potential record-breaking IPO.

The $60B valuation feels aggressive for a coding tool, even one as popular as Cursor. But the strategic calculus makes sense if you squint: SpaceX needs AI tooling for its engineering workflows, Cursor needs compute at scale, and Musk gets another chip in the AI arms race. Whether the option gets exercised depends entirely on how much progress they make in the next few months—and whether Cursor's valuation holds up under scrutiny.

Redwood Materials restructures again

Redwood Materials laid off 10% of its workforce this week as it pivots harder into energy storage. The battery recycling company cut 5% back in November, so this is the second round in five months.

The timing is awkward. Redwood closed a $425 million funding round in January that pushed its valuation above $6 billion. Now it's shedding headcount to chase a new line of business. The energy storage market is real—Meta just announced plans for up to 1 GW of 100+ hour storage with Noon Energy—but pivoting into it mid-growth means Redwood is betting it can compete on a new product before recycling margins tighten.

The cuts feel like a signal that the original recycling thesis isn't scaling as smoothly as projected. That or the energy storage opportunity is big enough to justify the disruption. Either way, two rounds of layoffs in half a year isn't a confidence builder.

NeoCognition raises $40M for agents that learn like toddlers

NeoCognition landed a $40 million seed round to build AI agents that supposedly learn the way humans do. The pitch is "agents that learn like humans"—specifically by interacting with environments and building models of the world incrementally, rather than training on static datasets.

The funding round is large for a seed, which suggests investors think the cognitive architecture angle has legs. Most agent frameworks right now are brittle: they break when the task drifts outside the training distribution. If NeoCognition can ship something that generalizes better by learning on the fly, that's a real unlock for enterprise workflows.

But we've heard the "learns like a human" pitch before. The gap between a research demo and a production agent is wide, and seed money doesn't mean the hard part is solved. The interesting thing to watch is whether they can show agents improving after deployment without needing a full retraining cycle. If they pull that off, the $40M will look cheap.

Quick hits

Sequoia raised $7 billion to expand its AI bets, per TechCrunch. New leaders, same thesis: infrastructure and foundation models are where the capital is moving.

OpenAI's original San Francisco office is on the market for $1.5 million. It's Greg Brockman's former Mission District residence, where the founders worked before they had a proper office. Historical curiosity more than anything, but the price feels symbolic—early OpenAI fit in a house; now it rents skyscrapers.

A Jump Trading quant researcher is reportedly getting launch capital from Millennium, according to Business Insider. Yiming Zhang is spinning out. High-frequency trading talent spinning into hedge funds is normal, but the timing lines up with a broader shift of quant infrastructure toward AI-driven strategies.

The vibe

This week had two big-dollar moves (SpaceX and Sequoia) and two restructuring stories (Redwood and NeoCognition's pivot to agents). The SpaceX-Cursor deal is the headline, but the Redwood cuts are the more interesting data point: even well-funded climate tech companies are discovering that pivoting mid-flight is expensive.

The Cursor option will either look brilliant or delusional by year-end. If SpaceX integrates it deeply and Cursor ships major updates using Colossus compute, the $60B might pencil out. If the partnership fizzles and SpaceX walks away with a $10B bill for custom tooling, that's still a win for Cursor but a lukewarm outcome for Musk's AI ambitions.

Either way, we're past the phase where "AI coding tool" meant a GitHub Copilot competitor. The big players are now betting billions on the thesis that software engineering itself becomes a commodity once the tooling is good enough. Cursor just became the test case for whether that's true—or just another expensive experiment.