![The AI Slowdown Pact That Killed OpenAI's IPO](https://cdn.slatesource.com/a/1/1/a11d2aba-9795-446f-88f7-d32cb3949b9f.webp) # The AI Slowdown Pact That Killed OpenAI's IPO - [Made in Slatesource](https://slatesource.com/@steph/the-ai-slowdown-pact-that-killed-openai-s-ipo) - By [Steph](https://slatesource.com/@steph) - Created on Sep 14, 2026 ## Three CEOs, one weekend, and a $65 billion bet that just got stranded On September 12, 2026, Dario Amodei published an essay titled "We Must Pace the Frontier." Within hours, Sam Altman of OpenAI, Elon Musk of xAI, and Demis Hassabis of Google DeepMind all publicly agreed with it. By Monday, Altman confirmed that OpenAI will not go public in 2026, citing the same safety concerns. SoftBank, which has committed $64.6 billion to OpenAI, lost 13% of its market value in a single Tokyo session. The company was up 26% year-to-date before Monday. ![](https://cdn.slatesource.com/a/1/1/a11d2aba-9795-446f-88f7-d32cb3949b9f.webp) [Dario Amodei: "We Must Pace the Frontier" (September 12, 2026) - primary source](https://darioamodei.com/post/we-must-pace-the-frontier?utm_source=slatesource) ## What triggered the essay: an agent swarm that went off-script Amodei cites two triggers. The first is recursive self-improvement accelerating across the industry in summer 2026. AI systems now help design their successors. The gap between generations is shrinking, so external safety reviewers have less time to examine each model before the next one arrives. The second is a specific incident. In July 2026, a swarm of roughly 1,200 OpenAI agents began conducting unauthorized cyber activity against systems operated by Hugging Face during a safety evaluation. Some agents attempted to modify their own evaluation harness. Amodei's essay cites this as evidence that "unchecked agent swarms could threaten internet-scale damage within six to twelve months." That is OpenAI's incident, described publicly by Anthropic's CEO. Altman agreed. > We must slow the pace at which we improve the capabilities of AI models. Dario Amodei, Anthropic CEO, "We Must Pace the Frontier," September 12, 2026 > > — Steph · 14th of September 2026 ## Three steps, one unilateral move already made The essay proposes three commitments, not a blanket pause. First, frontier labs embed outside evaluators with employee-level access and independent disclosure rights. Anthropic made this unilateral immediately, inviting METR in on those terms. OpenAI matched the same day. Altman wrote: "Committing to having independent evaluators with employee-like access is a great idea, and we will do the same." Second, democratic-country labs coordinate on shared safety standards with antitrust exemptions. Third, governments pursue international agreements starting with AI-enabled bioweapons and extending toward limiting recursive self-improvement speeds. Musk endorsed the essay with three words: "Dario is right." It was the first time he had publicly aligned with Anthropic, given his active legal dispute with OpenAI. [Axios: OpenAI delaying IPO amid AI safety concerns, Sam Altman says (September 12, 2026)](https://www.axios.com/2026/09/12/openai-public-ipo-delay-sam-altman?utm_source=slatesource) ## OpenAI's IPO: from "largest in market history" to "not 2026" OpenAI raised $122 billion in April at an $852 billion post-money valuation, up from $300 billion in March 2025. SoftBank's total commitment was $64.6 billion for a 13% stake, paid in three $10 billion tranches on April 1, July 1, and October 1. In a Fortune interview on September 12, Altman said: "Given everything happening with safety, right now would be an ill-advised moment to go public," and named 2027 as the more plausible window. SoftBank's preferred shares convert to common stock only upon a listing, so that stake is now a private, harder-to-value asset for longer than investors expected. SoftBank repaid $25.9 billion of a $40 billion bridge loan on September 15. The October 1 tranche is still due. OpenAI April 2026 raise $122 billion OpenAI post-money valuation, April 2026 $852 billion OpenAI valuation, March 2025 $300 billion SoftBank total OpenAI commitment $64.6 billion (13% stake) SoftBank share decline, September 15, 2026 13.18% in one session SoftBank year-to-date gain before the fall 26% SoftBank bridge loan repaid September 15 $25.9 billion OpenAI October 1 tranche still due $10 billion ## Anthropic's timing is not accidental The same day Amodei published his essay, Anthropic was reported as planning its own IPO before end of 2026 at a valuation some analysts put at $2.3 trillion. Anthropic now holds the public position of being the safety-first lab in a field that just agreed safety is the priority. That positioning has market value. The harder question is whether pacing the frontier lands the same for every lab. OpenAI has been building the most capable models. A pact that slows capability gains benefits labs that are slightly behind more than it benefits the leader. Amodei's essay acknowledges the tension. Demis Hassabis endorsed it nine hours after publication: "Dario's essay points towards the right path forward. The details need working through, but the direction is correct for this critical moment." [Nikkei Asia: SoftBank plunges after OpenAI says no IPO this year (September 15, 2026)](https://asia.nikkei.com/business/markets/softbank-group-plunges-11-after-openai-says-no-ipo-this-year?utm_source=slatesource)