Anthropic filed confidentially for a public offering today, marking a significant step toward going public for one of AI's largest independent laboratories. The move comes as the company rolls out Claude Opus 4.8, which delivers a 4x improvement in code reliability over its predecessor. Anthropic has raised roughly $7 billion to date and reached a $60 billion valuation in its most recent funding round.
Meanwhile, NVIDIA accelerated its AI infrastructure push at GTC Taipei, launching Cosmos 3, a video generation model designed for robotics and industrial applications. The company also moved Vera Rubin, an inference optimization platform, into production and packaged a 1-petaflop AI system for developer laptops. This positions NVIDIA to capture more of the inference market as enterprises move beyond training into deployment.
Google deprecated Gemini 2.0 Flash today, signaling shifts in its model lineup as competition intensifies. The company continues to reshape its AI portfolio in response to Claude's gains in reasoning and coding tasks.
On the regulatory front, California's SB 867 cleared the state Senate, banning AI companion chatbots designed for children's toys. The legislation addresses concerns about children forming parasocial attachments to AI systems. Illinois attempted similar data-center regulation, but the measure stalled in committee, reflecting the fragmented approach states take to AI governance.
The contrast between lab velocity and legislative pace remains stark. Anthropic filed for IPO status while advancing code generation capabilities. NVIDIA shipped production inference tools and packed supercomputer performance into consumer hardware. Google consolidated its model strategy. All of this happened in days or weeks.
State legislatures, by contrast, worked months on AI regulation and produced mixed results. California moved a narrow restriction through one chamber. Illinois couldn't advance data-center rules at all.
This disparity shapes AI's trajectory. Labs