SpaceX shattered every IPO record this week, but the real story was capital flooding into AI from every direction — Bezos betting $12 billion on physical-world intelligence, Apple finally reinventing Siri, and Washington trying to write the rules before the money moves too fast. The line between tech company and infrastructure play has all but disappeared.
The week's unmistakable centrepiece was SpaceX going public. Trading under the ticker SPCX on the Nasdaq, Elon Musk's rocket-and-satellite empire raised $75 billion at a $1.77 trillion valuation — dwarfing Saudi Aramco's previous record by a factor of three. The stock closed its first day at $161, up 19% from the $135 offering price, buoyed by an unusually retail-friendly structure that reserved 30% of shares for everyday investors. Revenue climbed 33% last year to $18.7 billion, though a $4.9 billion net loss underscored how much of the thesis still rests on Starlink's trajectory and the bet that AI compute may eventually follow the satellites into orbit.
Apple used its WWDC keynote on Monday to unveil Siri AI, a ground-up rebuild of its voice assistant powered by the latest Apple Intelligence stack in partnership with Google. The new Siri holds natural back-and-forth dialogue, reads on-screen context, and edits images and text — capabilities that felt routine on competing platforms but had been conspicuously absent from the iPhone. Availability will be limited to devices with at least 12 gigabytes of unified memory, and China is excluded entirely while Apple works through regulatory requirements. The move signals that the smartphone AI race is no longer about who ships a model first, but who embeds it deeply enough that users stop reaching for third-party apps.
The physical-AI thesis gained another enormous backer when Prometheus, the startup co-founded by Jeff Bezos and Vik Bajaj, closed a $12 billion round at a $41 billion valuation, bringing total funding past $18 billion. Bezos described the goal as an "artificial general engineer" — software that designs and simulates complex physical systems from jet engines to drug compounds, pointedly insisting it has nothing to do with robotics. Meanwhile, Nvidia found its own way around Washington's GPU export curbs by pitching Arm-based Vera CPUs to Chinese clients, with shipments targeted for August; because export controls primarily restrict high-end accelerators rather than general-purpose processors, the Vera line opens a side door back into the world's second-largest compute market. And Super Micro Computer announced a $7 billion equity raise to fulfil roughly $39 billion in AI server orders received in recent weeks — a staggering figure that nonetheless sent shares tumbling nearly 20% on dilution fears, a reminder that the infrastructure buildout demands capital on a scale that tests even public-market appetite.
Washington, for its part, is scrambling to keep pace. President Trump signed an executive order on June 2 establishing a "Secure Frontier Model Deployment" framework that asks AI developers to give the federal government early access to frontier models for up to 30 days before wider release — framed as voluntary, but carrying the implicit weight of White House attention. Three days later, a National Security Presidential Memorandum laid out four pillars — Adoption, Adaptation, Assurance, and Accountability — for AI procurement across the defence and intelligence apparatus. At the state level, Colorado quietly replaced its landmark AI Act before it ever took effect: Governor Polis signed a softer replacement bill in May, pushing the effective date to January 2027 and stepping back from the EU-style obligations the original imposed on high-risk AI deployers. The pattern is now unmistakable — federal policy leans toward access and acceleration, while states experiment and often retreat.
What connects SpaceX's record debut, Bezos's physical-AI wager, and Supermicro's scramble for capital is the sheer gravitational pull of AI infrastructure spending. Half of all global venture funding now flows into AI-related companies, and the four largest venture rounds in history — OpenAI, Anthropic, xAI, and Waymo — all closed within the past five months. For businesses watching from the sidelines, the practical signal is that compute, connectivity, and physical-world intelligence are converging into a single investment thesis, and the companies positioning themselves at those intersections are attracting capital at a pace that makes last year's mega-rounds look quaint.