Good morning. The capital markets story finally caught up with the AI story: Anthropic filed for what could be the biggest IPO in history, Alphabet is raising $80 billion to keep the GPUs coming, and OpenAI quietly walked into Anthropic’s enterprise stronghold on AWS. Meanwhile, Florida’s attorney general is testing whether a chatbot can be sued for a mass shooting, and Meta’s support bot was handing out Instagram accounts to anyone who asked nicely.
Anthropic files confidentially for an IPO. Anthropic submitted a draft S-1 to the SEC, days after its $65 billion Series H closed at a $965 billion valuation that edges out OpenAI. Reporting on the size of the offering varies wildly — Wired cites a $61.5 billion valuation while The Verge and the company’s own filing point to the $965 billion figure, and revenue is reportedly running at $47 billion annualized despite ongoing losses. HN’s reaction was mostly nervous: one commenter noted that until now the blast radius of an AI bust was confined to corporate investors, and dragging retail into 401(k) exposure at these multiples is a different kind of risk.
Alphabet raises $80 billion to feed the buildout. Google’s parent announced an $80 billion equity raise, including a $10 billion private placement with Berkshire Hathaway split between Class A and Class C shares. Per TechCrunch, total AI capex for the year is now projected at $180–190 billion, with Alphabet saying demand exceeds supply. The stock fell about 1.7% after hours, and one HN commenter summed up the broader mood: every hyperscaler is now spending wildly to be the last one standing, and the fallout underneath the winner is going to be considerable.
OpenAI lands on AWS Bedrock. OpenAI’s frontier models and Codex are now available through AWS, ending the de facto Azure exclusivity that made Anthropic the default frontier option for enterprises locked into AWS contracts and data governance regimes. Several HN commenters who work in regulated enterprises were blunt: a lot of Claude adoption was an artifact of “we can only buy through Bedrock,” and that moat just evaporated. One suggested Anthropic’s IPO timing might not be coincidental.
Florida sues OpenAI over violent incidents. Attorney General James Uthmeier filed an 83-page suit against OpenAI and Sam Altman, alleging ChatGPT contributed to the 2025 FSU shooting, youth addiction, and suicides, and that the company prioritized winning “the AI arms race” over safety. It’s the first state-level action of its kind, following civil suits from victims’ families. HN commenters on the Politico writeup were skeptical of the legal theory — comparing it to 1990s video game crusades and noting that gun manufacturers enjoy broader liability shields than OpenAI is being denied here — and flagged the conspicuous absence of Google, xAI, or Anthropic as co-defendants.
Meta’s support chatbot was handing out Instagram accounts. The Verge reports that Meta’s AI support bot contained a flaw letting attackers link a new email to anyone’s Instagram account just by asking, with no authentication check. The exploit was used to hijack Barack Obama’s @obamawhitehouse account (which briefly posted Iranian propaganda), US Space Force accounts, and Sephora, among others. Meta has patched it, but the pattern — give an LLM account-modification privileges, skip the identity verification — is becoming familiar.
That’s the briefing. Two IPOs and an $80 billion raise in a single news cycle is a useful reminder that the bottleneck right now isn’t models or chips, it’s how much capital the public is willing to put behind both.