OpenAI Files Confidential S-1, Targeting Historic $1 Trillion IPO
OpenAI has taken the first formal step toward going public, confidentially filing its S-1 prospectus with the SEC as of May 22. The company is working with Goldman Sachs and Morgan Stanley and targeting a September 2026 listing that could value it at approximately $852 billion to $1 trillion — making it potentially the largest technology IPO in history.
The filing comes on the heels of OpenAI closing a massive $122 billion funding round in March at an $852 billion post-money valuation. The company has hit $25 billion in annualized revenue, driven largely by surging enterprise adoption of its API and ChatGPT Enterprise products. A key legal overhang was also lifted on May 18, when a U.S. jury unanimously rejected Elon Musk's claims against the company, clearing a major risk for prospective public investors.
If the IPO proceeds on schedule, OpenAI would join a remarkable wave of AI-adjacent public listings: SpaceX filed its own S-1 on May 20 targeting a $1.75 trillion Nasdaq debut in June, and Anthropic is reportedly eyeing an October 2026 listing at a $900 billion valuation.
Anthropic Projects First-Ever Quarterly Profit on Soaring Revenue
Anthropic shared financial projections showing $10.9 billion in expected Q2 2026 revenue — a staggering 130% increase from the $4.8 billion it generated in Q1. More significantly, the company projects approximately $559 million in operating income, marking its first-ever quarterly operating profit.
The turnaround is driven by improving compute economics: Anthropic's cost ratio dropped from 71 cents per revenue dollar in Q1 to 56 cents in Q2, helped by its expanded infrastructure deal with SpaceX's Colossus data centers. That deal, revealed in SpaceX's S-1 filing, is worth a jaw-dropping $1.25 billion per month through May 2029 — a total commitment of roughly $45 billion covering access to over 220,000 NVIDIA GPUs across Colossus 1 and Colossus 2 facilities in Memphis.
Enterprise adoption continues to accelerate, with the number of customers spending over $1 million annually doubling from 500 to 1,000 between February and April. Anthropic's annualized run rate has reached $43.6 billion, up from roughly $9 billion at the end of 2025.
Chinese AI Models Now Dominate OpenRouter, Intensifying Price War
In a dramatic illustration of how quickly the competitive landscape is shifting, Chinese AI models now account for more than 60% of usage on OpenRouter, the largest third-party AI routing platform — up from approximately 1% in 2024, according to Databricks CEO Ali Ghodsi.
A CNBC investigation published today found that for a standard enterprise workload, Claude costs $4,811, ChatGPT costs $3,357, while Chinese alternatives like DeepSeek run at $1,071, Kimi at $948, and Zhipu's GLM at just $544. That makes the leading American models up to 9x more expensive than the cheapest Chinese alternatives.
The price gap raises uncomfortable questions about the sustainability of $800+ billion valuations for US AI companies. However, enterprise buyers appear to be adopting an "advisor model" pattern — using premium American models for high-stakes reasoning tasks while routing simpler workloads to cheaper Chinese alternatives.
White House Reverses Course, Embraces AI Oversight
In a significant policy pivot, the Trump administration is now embracing AI oversight measures it had previously rejected. The shift was catalyzed by national security concerns over Anthropic's "Mythos" model, which demonstrated the ability to identify and exploit cybersecurity vulnerabilities.
A planned executive order — postponed on May 21 due to internal disagreements — would establish a 90-day pre-launch voluntary review framework involving NSA-classified testing. The renamed Center for AI Standards and Innovation (CAISI, formerly Biden's AI Safety Institute) has already signed agreements with Google, Microsoft, and xAI to evaluate models before public release, and has completed over 40 model evaluations including unreleased frontier systems.
Unlike the Biden administration's focus on AI ethics and existential risk, the Trump approach frames oversight through the lens of immediate national security threats — cyberwarfare, infrastructure vulnerabilities, and strategic competition with China. Anthropic's own policy paper, released this month, warns that US models hold merely a "several months" lead over Chinese counterparts in capabilities.
Google Unveils Gemini Spark Agent, Omni World Model, and Gemini 3.5 Flash
Google announced a wave of AI products this week, headlined by Gemini Spark — a new general-purpose AI agent that can reason across connected apps within the Gemini ecosystem. Spark is launching in beta for trusted testers and Google AI Ultra subscribers.
The company also revealed Omni, a new world model, and Gemini 3.5 Flash, a lighter-weight model that Google claims beats Anthropic's Opus 4.7 on five coding and agentic benchmarks while costing significantly less — priced at $1.50/$9 per million input/output tokens. Flash does trail on long-context and knowledge-breadth tasks, suggesting it is optimized for the fast-growing developer and agent market segment.
Additionally, Google announced a no-code agent builder for Google Workspace, a redesigned developer platform with over 200 models (including third-party options such as Claude), and Project Mariner, a web-browsing agent. The moves reflect Google's strategy of leveraging its massive distribution advantage — Search, YouTube, Android — to compete with Anthropic and OpenAI on the consumer and enterprise fronts simultaneously.