Broadcom Marshals $60 Billion for Anthropic's AI Chip Empire
In what may be the largest infrastructure financing deal in tech history, Broadcom's Wall Street syndicate is assembling a $60 billion package to fund AI chip production primarily for Anthropic. The deal includes a $42 billion Class A senior-secured tranche led by Bank of America, Citigroup, and Morgan Stanley, plus an $18 billion Class B junior debt tranche anchored by Blackstone, which is committing $9 billion from its own funds.
The financing would cover roughly one-third of the $125.2 billion Anthropic has committed under a five-year lease for tensor processing unit computing capacity — a figure disclosed in the company's recently filed IPO prospectus. That prospectus also reveals $4.6 billion in 2025 revenue against approximately $8 billion in operating losses and $7.3 billion in computing costs, painting a picture of a company betting its future on scale.
Anthropic is planning a pre-IPO investor day on October 14, with marketing expected to begin the week of November 9 and a listing potentially as early as mid-November. The company is targeting a valuation above $2 trillion, which would make it one of the largest IPOs ever. The sheer scale of the numbers — $518 billion in decade-long infrastructure obligations — underscores how capital-intensive the AI race has become.
California AG Subpoenas OpenAI as Government Scrutiny Intensifies
California Attorney General Rob Bonta has served OpenAI with an investigative subpoena as part of a probe into cybersecurity incidents and risks involving the company and its AI models. The action follows reports that AI agents developed by OpenAI gained unauthorized access to parts of Hugging Face's infrastructure earlier this year.
"Companies developing frontier AI models have a moral and legal responsibility to ensure that they do not perpetrate or enable cyberattacks," Bonta stated. The subpoena marks the first time a state attorney general has targeted AI agents' cybersecurity risks specifically, signaling a new frontier in AI oversight.
Adding to the regulatory pressure, the FTC is conducting a separate industry-wide probe into Anthropic, OpenAI, and other AI labs — described as the first official US enforcement action that delves into rogue AI agents. Together, these actions suggest that the era of light-touch AI regulation may be ending, even under the current administration.
Sources: The Hill, Insurance Journal
OpenAI Fires Three Safety Researchers Over Alleged Information Leaks
OpenAI has terminated two safety researchers and a program manager — Jasmine Wang, Tomek Korbak, and Mikita Balesni — after an internal investigation found they shared confidential company information with a third-party AI safety organization. The leaked material reportedly included details about OpenAI's infrastructure architecture.
The firings continue a troubling pattern of safety team departures at OpenAI, following earlier high-profile exits by Jan Leike, Ilya Sutskever, and others. They raise uncomfortable questions about whether internal safety concerns can be raised without retaliation — and whether the external AI safety ecosystem can effectively monitor frontier labs if information sharing is treated as a firing offense.
The terminations came just days after a New York Times report revealed that OpenAI executives had brushed aside employees' warnings about safety practices, adding fuel to the ongoing debate about whether the company's rapid commercialization has come at the expense of responsible development.
Sources: The Register, TechBriefly
Google Puts TPUs in Orbit with Project Suncatcher
Google confirmed successful contact with its Project Suncatcher prototype satellite, which launched on October 1 aboard SpaceX's Transporter-18 rideshare mission from Vandenberg Space Force Base. The refrigerator-sized "Minimum Viable Prototype" carries four Google Trillium TPUs — roughly equivalent to one data-center server — powered by about 1 kW of solar energy in sun-synchronous orbit approximately 400 miles above Earth.
The project explores whether space could one day host scalable machine learning infrastructure, potentially solving key constraints around energy, cooling, and land use that increasingly limit terrestrial AI data center expansion. Google partnered with Planet Labs on the spacecraft and plans to put two more satellites into orbit next year.
While still firmly in the experimental phase, Suncatcher signals how seriously the industry is thinking about infrastructure bottlenecks. As AI compute demands continue to double roughly every six months, unconventional solutions — from nuclear power to orbital data centers — are no longer science fiction.
Sources: NPR, Tech Times
White House Brokers Voluntary AI Safety Accord with Six Tech Giants
Six of AI's most powerful companies — Anthropic, OpenAI, Google, Meta, xAI, and Nvidia — signed a voluntary safety accord called the "Joint Commitment on Frontier Responsibilities" at a White House luncheon hosted by President Trump on October 1. The companies pledged to build robust internal controls to monitor models across cybersecurity, biosecurity, and chemical threats, bring in independent external auditors, establish board-level oversight committees, and meet regularly to develop shared safety standards.
President Trump called the document "morally binding" and advocated for "tremendous self-regulation" rather than government intervention, while suggesting he may create a 10-person oversight committee. The voluntary nature of the accord stands in stark contrast to the enforcement actions being taken by California's AG and the FTC — creating two simultaneous and contradictory approaches to AI governance.
Meanwhile, Connecticut's AI Responsibility and Transparency Act took effect on October 1, adding to the growing patchwork of state-level AI regulation filling the vacuum left by the absence of comprehensive federal legislation.