OpenAI's Self-Funded Buyback and the Executive Exodus

2 min read
openaiexecutive-turnoveripoai-safety
View as Markdown
Originally from youtube.com
View source

My notes

Summary

OpenAI self-funded a $7 billion employee share buyback at its March 2026 valuation of $852 billion, with no outside buyers, days before its COO of eight years and its CRO (eight months in the role) both resigned. The video argues this pattern - a decade of senior departures, three dissolved safety units, and a pre-IPO valuation nobody outside the company has tested - signals deeper instability than the “routine reshuffle” framing OpenAI offers publicly.

Key Insight

  • Self-funded tender = avoided price discovery: the October 2025 tender ($6.6B) had outside buyers (Thrive, SoftBank, Tiger); the August 2026 one ($7B) was funded entirely from OpenAI’s own March round cash, at the same $852B mark. No external buyer had to agree the price was still right - the valuation was asserted, not tested.
  • Every “adult supervision” hire has since left: Denise Dresser (ex-Slack CRO, 8 months), Fidji Simo (ex-Instacart, head of Applications, 14 months, cited health reasons), Brad Lightcap (COO, 8 years), and Sarah Bidal - hired specifically as the operating bench under a founder who’d already been fired once (Nov 2023 board ouster). Median tenure of these “grown-up” hires is well under two years.
  • Every person who formally challenged Altman’s trustworthiness is now gone: Ilya Sutskever (52-page memo on Altman’s conduct, per unsealed Musk-litigation depositions) left to found Safe Superintelligence; Jan Leike (Superalignment co-lead, cited safety “taking a backseat to shiny products”) now runs safety research at Anthropic; Mira Murati, Helen Toner, and Tasha McCauley are all gone. Altman is the only one still there.
  • Three safety units dissolved since 2024: Superalignment (May 2024), Mission Alignment (Feb 2026), and the Preparedness team assessing catastrophic risk (disbanded end of July 2026, after safety lost its independent reporting line in a July reorg). Concurrent structural shifts: removed the 100x investor profit cap, converted to a Delaware public benefit corp, softened “safely” in the charter language.
  • The circularity behind the IPO stakes: Amazon, Alphabet, Microsoft, and Meta spent a combined $165B in capex in one quarter; 92% of Nvidia’s revenue is data-center chips, much of it sold back into rounds Nvidia itself invested in (e.g., its $30B stake in OpenAI’s March round). One MIT study cited in the video found 95% of enterprise generative-AI pilots showed no measurable profit impact - the infrastructure bet rests on demand growth that hasn’t yet shown up in returns.
  • Confidential S-1 filed 8 June 2026 (Goldman Sachs, Morgan Stanley, JP Morgan leading); bankers reportedly discussing $830B-$1T, but the listing timeline has already slipped from late 2026 toward 2027, with Altman said to be pushing for a $1T/September window against CFO pushback.