$122B for OpenAI, $30B for Anthropic, and Claude Finds What Humans Missed
Tue, Apr 7, 2026 · 9 stories
A few weeks ago, OpenAI's GPT 5.4 was the first AI to solve an open Ramsey-style combinatorics problem from FrontierMath that humans hadn't cracked. This week, Claude found a Linux vulnerability that had been hiding in plain sight for 23 years. Looking at the scoreboard, we're seeing a 1-1. The AI duopoly is forming in real time. OpenAI closed the largest private funding round in history at $122 billion while Anthropic tripled its revenue to an astounding $30 billion run rate. These two companies are pulling away from the competition. With money being the true endgame here, the race to IPO is going to need a photo finish.
Top Stories
Bloomberg Technology
OpenAI closed a $122 billion funding round valuing the company at $852 billion, with Amazon, Nvidia, and SoftBank among the major contributors. Meanwhile, private market sentiment is shifting -- Next Round Capital says OpenAI shares are flooding secondary markets while Anthropic emerges as the stronger financial opportunity.
Why this matters:
The largest private funding round in history tells you less about OpenAI's strength than about how much capital the AI arms race now demands. When secondary market insiders say Anthropic looks like the better bet despite a fraction of the headline valuation, it signals the market is starting to price execution over hype.
Stratechery
Anthropic's revenue run rate has surged from $9 billion at end of 2025 to $30 billion, a 3.3x increase in roughly four months. The company simultaneously locked in a compute partnership with Google and Broadcom to supply TPU chips, securing the infrastructure runway to sustain this growth.
Why this matters:
Revenue tripling in a quarter while simultaneously locking down a strategic compute pipeline is the kind of execution that turns a "promising challenger" into a market leader. The Google/Broadcom deal isn't just about chips -- it's Anthropic using its biggest investor's hardware advantage as a competitive moat against OpenAI's Nvidia dependency.
mtlynch.io
Claude Code discovered a security vulnerability in Linux that had evaded human review for 23 years. The same week, Claude generated a full FreeBSD remote kernel RCE exploit with root shell access (CVE-2026-4747), demonstrating AI's growing capability in security research.
Why this matters:
Benchmarks show what models can do in theory. Finding a real vulnerability humans missed for two decades shows what they can do in practice. For anyone building or investing in AI-assisted security tooling, this is the proof point -- and for everyone running production infrastructure, it's a wake-up call about what attackers will soon have access to as well.
Quick Hits
DeepMind Blog
Google launched Gemma 4, their most capable open models yet, designed for on-device and agentic workflows -- already running on iPhones via Google AI Edge Gallery.
The Verge
Mustafa Suleyman, Microsoft's first-ever CEO of AI, says renegotiating the OpenAI contract "unlocked" the company's ability to pursue superintelligence internally -- a pivot he'd been planning for nine months.
The Verge
Starting April 4, Claude Code subscribers must pay separately for OpenClaw and other third-party tool integrations, effectively pricing out the popular open-source harness.
Bloomberg Technology
Perplexity faces a California privacy lawsuit alleging it shared user personal information with Meta and Google, a significant test case for AI search companies' data handling practices.
TechCrunch
Anthropic acquired stealth biotech startup Coefficient Bio in a $400M stock deal, marking its first major expansion beyond pure AI into biological research and drug discovery.
Reddit/LocalLLaMA
Mistral dropped Voxtral TTS, an open-weight text-to-speech model that clones any voice from 3 seconds of audio in 9 languages, beating ElevenLabs Flash v2.5 with a 68.4% human preference win rate.
