OpenAI Raises $122 Billion While the Vibes Get Weird

OpenAI closed the largest private funding round in history, but its CFO doubts it's IPO-ready, Sora is dead, and Anthropic is quietly flexing a model it won't release. The money is flowing — but the power map is shifting.

The Money Story Hides a Messier One

OpenAI just closed $122 billion in funding at an $852 billion valuation — the largest private tech financing in history. And yet, as The Verge put it bluntly this week, "the vibes are off at OpenAI." The company's own CFO told The Information she doesn't believe they're ready for an IPO, citing uncertainty about whether revenue growth can support spending commitments. That's a remarkable admission from a company valued higher than most countries' GDP.

Meanwhile, Q1 2026 saw AI companies collectively raise $297 billion (per the New York Times), with OpenAI and Anthropic leading the charge. Jeff Bezos's stealth AI venture, Project Prometheus, is poaching engineers from both OpenAI and xAI. The talent wars are as telling as the funding wars — where the best researchers choose to work signals where the real breakthroughs will happen.

Codex vs. Claude Code: The Agent Showdown Heats Up

OpenAI launched Codex as a full coding agent — available in ChatGPT, VS Code, and the terminal. A coding agent is AI that doesn't just suggest the next line of code, but plans, builds, tests, and iterates across entire projects autonomously. Anthropic fired back with Claude Code Desktop, which now includes computer use on Windows — the AI can navigate your screen, click buttons, and complete tasks across applications. Google's Gemini Code Assist added MCP support (Model Context Protocol — an open standard that lets AI tools plug into external data sources and services). Even Cursor, the popular AI code editor, launched its own agent experience to compete.

The pattern is clear: every major vendor now treats autonomous coding agents as table stakes. The differentiator is shifting from "can the AI write code" to "can it operate your entire development workflow."

Anthropic's Mythos Flex

Perhaps the most interesting signal this week was what Anthropic chose not to release. The company announced Claude Mythos Preview, a model so capable at finding software vulnerabilities that they restricted it to a handful of tech companies. NBC News reported Anthropic called it a "cybersecurity reckoning." This is a strategic play — demonstrating frontier capability while positioning as the responsible vendor. Whether that framing holds depends on whether the security community validates the claims.

The Sora Collapse

OpenAI shut down Sora, its AI video generation app, less than two years after its splashy debut. The BBC reported the $1 billion Disney deal was also cancelled. Video generation turns out to be extraordinarily expensive to run at scale, and Sora never found product-market fit. For anyone building on vendor APIs: this is a reminder that even the best-funded labs will kill products fast.

What to Watch

The bottom line: The AI vendor landscape is splitting into two tiers — companies with enough capital to burn through product failures (OpenAI, Anthropic, Google), and everyone else racing to find a defensible niche before the window closes.

Sources