SpaceX Is Going Public on an AI Story It Hasn't Written Yet

The technology industry's appetite for trillion-dollar narratives is running ahead of the physics — SpaceX scrubbed its Starship V3 launch moments before liftoff, but the more consequential development is in the IPO filing: Musk is going public on an AI pitch while the AI product he controls, Grok, is losing ground to every serious competitor.

SpaceX's IPO Is an AI Bet with a Voting Lockout

The SpaceX IPO filing reveals Musk will hold more than 50% of voting power in the public company — a control structure that goes beyond what Zuckerberg or Page maintained at comparable stages. That matters because the IPO thesis, per Ars's reporting on SpaceX's AI ambitions, leans heavily on orbital data centers as a differentiator in the AI infrastructure race. The pitch: low-latency compute delivered from space. The problem: Grok, the AI product Musk actually controls, is widely regarded as trailing GPT-4o and Claude by a meaningful margin. Investors who can't vote out management are being asked to trust a CEO whose AI track record is mixed. The Information's valuation analysis puts the company at $700 billion — not the $1.75 trillion implied by recent secondary markets — a gap that suggests the AI narrative is doing a lot of heavy lifting.

AI Agents Are Printing Money for Some and Writing Exit Clauses for Others

Workday's stock jumped 10% after the company disclosed meaningful gains from AI agents embedded in its HR and finance platform. That's the bull case made concrete: agents that automate repetitive knowledge work inside existing SaaS workflows, generating productivity lift that justifies contract renewals. But a quieter counter-movement is forming: sophisticated enterprise buyers are negotiating contracts with explicit escape hatches — clauses that let them exit if AI capabilities don't materialize or if a purpose-built agent displaces the incumbent suite entirely. The companies winning on agents today are buying goodwill; the ones that overpromise are writing their own termination terms.

Google Isn't Really Google Anymore

TechCrunch's roundup of search alternatives lands the same day as a widely-circulated essay making the IBM-ification argument about Google: a company once defined by engineering excellence, now bureaucratic, slow, and willing to degrade its core product in service of ad revenue. The AI Overview rollout has accelerated this. Google is converting its search results page into an answer engine that discourages clicks — which works until the summaries are wrong, and which makes every query a small bet on AI accuracy. Switching costs remain high for most users. But the conversation has shifted from "why would you use anything else" to "here are six alternatives worth trying" — and that shift in posture matters.

Waymo's Edge Cases Are Getting Wet

Waymo has expanded its service pause to four cities after its robotaxis repeatedly navigated into flooded roads. The specific failure mode is instructive: the vehicles detected the road ahead as passable because their sensor and decision stack wasn't calibrated to recognize flood conditions from available inputs. That's not a software bug you patch overnight — it's a gap in the training distribution. The AV industry has spent years arguing that rare edge cases are solvable with more data. Standing water on city streets is not a rare edge case. It rains.

The Gap Between Narrative and Reality

The SpaceX IPO will probably be oversubscribed regardless, because the AI story is irresistible and Musk's voting structure means the market's only real lever is price. Between $700 billion and $1.75 trillion lies a trillion-dollar question about whether SpaceX can execute on an AI narrative it's only just begun to construct — and whether public investors, locked out of governance, will have any recourse if it can't.