OpenAI’s $1.25T Raise, Starlink Mobile & Zuck’s Tax Deal | EP 2354
Saturday, 24 October 2026 · 4 min read · Listen to the episode ↗
OpenAI was reportedly nearing a raise of up to $30 billion at a $1.25 trillion pre-money valuation, about 30 times its cited $50 billion annualized revenue run rate, though the financing terms and revenue calculation remain unclear. SpaceX is discussing plans for at least 80,000 U.S. cell sites as it seeks to make Starlink Mobile a nationwide carrier. Meta’s deal with Mark Zuckerberg permits quarterly share sales while preserving his voting control, amid a proposed California tax on billionaires.
OpenAI was reportedly close to raising as much as $30 billion at a $1.25 trillion pre-money valuation, with Co2- and UAE-backed investors expected to lead. UAE investment could be linked to a regional compute-capacity deal. The reported valuation is about $150 billion below the $1.4 trillion Bloomberg said OpenAI was seeking in late September. The only revenue figure cited was an approximately $50 billion annualized run rate, but the calculation was not explained.
At that run rate, the valuation implies roughly 30 times sales, a measure the speakers said should be treated cautiously: annualized revenue may extrapolate a recent month or quarter and include one-off contracts, unlike recurring ARR. The round’s liquidation preferences, interest, conversion terms, and investor rights were also unknown. One speaker considered the valuation fair if OpenAI keeps growing rapidly, but saw headwinds for a quick trade, including competition from Anthropic, frontier-model labs, enterprise vendors, major tech companies’ agent products, and cheaper open-source models. The company’s ability to deploy enough compute is another constraint, and model comparisons depend on task and setup rather than yielding a clear overall winner.
Anthropic was described as having raised at a $965 billion valuation in May and reportedly targeting a $1.5 trillion to $2 trillion IPO that could precede OpenAI’s. In a hypothetical retirement portfolio for someone about 50 with 20 years before withdrawals, one speaker suggested a 30% allocation across SpaceX, OpenAI, and Anthropic, possibly 20% to SpaceX and 5% to each AI company, while emphasizing that it was not investment advice. SpaceX was the preferred holding because of its Starlink, rocket, and frontier-model ambitions.
SpaceX reportedly wants at least 80,000 U.S. cell sites and is discussing construction with Crown Castle and others as it seeks to make Starlink Mobile a fourth nationwide wireless carrier. Elon Musk said satellites combined with newly acquired low-band spectrum could help the company become a major mobile provider. Crown Castle rose as much as 10% after hours on the report, while AT&T, Verizon, and T-Mobile each fell about 5%. Speakers saw potential for lower connectivity costs, but said ground towers could improve indoor coverage and fiber-to-the-home could deliver faster speeds than the 200–300 Mbps Starlink home downloads they cited. They interpreted SpaceX’s launch plans for this year and next, with broader commercial service and international rollouts through 2027, as potentially putting Starlink Mobile about 18 months away, perhaps in 2028. They expected competitors to respond, possibly through mergers or a partnership with Amazon Leo.
A pilot by METR and Redwood Research, which had three weeks of access to OpenAI this summer, identified vulnerabilities in internal monitoring, including an exploit that could hide hundreds of agent actions. Evaluators also found a potential way for Codex agents to bypass security restrictions while appearing to retain users’ full permissions. OpenAI said it patched issues affecting production safeguards, but the evaluators had not verified the fixes. OpenAI received a 61-page report; the public gets a broad summary that the company can redact. Speakers called the audit a positive first step while questioning the independence of company-funded evaluators and noting METR’s funding and connections to effective-altruist and Berkeley networks. They argued that scarce specialist expertise makes second-level reviews and more unaffiliated audit teams important.
Speakers called cybersecurity the most important short- to mid-term AI challenge and urged more investment in white-hat protection. They also raised uncertainty over who bears liability when customers use OpenAI or Claude through AWS or Microsoft Azure. They said there is no general AI equivalent of Section 230 shielding providers, and distinguished a chatbot’s retrieval of harmful information from its allegedly encouraging or steering someone toward harm. Speakers anticipated settlements in cases alleging chatbot involvement in suicide or violence, while noting that the facts and responsibility in such cases remain contested.
California voters are due to decide in November on a proposed one-time 5% tax on residents worth more than $1 billion. Speakers characterized it as a potential asset seizure that could leave founders owing tax on illiquid shares without the cash to pay; one estimated Mark Zuckerberg’s wealth at more than $200 billion, implying a bill of about $10 billion. Meta’s agreement with Zuckerberg creates a new share class, provides him with stock more frequently, and permits quarterly share sales without changing his voting control. The agreement lasts eight years if he remains a California resident and appears to apply even if the tax fails. A prediction market put the measure’s passage chances at 23%, with about $4 million in trading volume, and one speaker said the odds showed no visible change after the deal. Speakers disagreed on policy: some favored higher taxes or a minimum tax but opposed asset seizure, while alternatives raised included taxing margin loans when taken out, or taxing part of long-held stock even when unsold.
This summary was generated from the episode transcript and can contain mistakes.