AI Circle Daily
https://aicircle.news/archive/2026-08-18?lang=en
Scan to read this page
Anthropic's annualized revenue has surged to $65 billion, adding a staggering $18 billion in just the last two months.
Take: Everyone was talking about OpenAI's dominance last year, but Anthropic is clearly catching up fast. Turns out the AI pie is big enough for multiple giants.
Nvidia has invested $1.5 billion in SoftBank's data center developer, specifically to ensure its chips power an upcoming OpenAI data center.
Take: Classic move to lock in customers. Nvidia's securing its chip sales by directly tying into the infrastructure of a major player. Smart play.
Amazon, originally a bookseller, is reportedly destroying rare and valuable books to train its AI models, as online data is already exhausted.
Take: The irony is rich here – the company built on books is now burning them for AI. Beyond copyright, this raises new ethical questions about data sourcing.
OpenAI reportedly disbanded its “preparedness team” last month, a group responsible for assessing severe model risks and developing mitigation strategies, including scenarios where AI might go rogue.
Take: Talk about mixed signals – preach safety, then axe the safety team. Not a great look, makes you wonder if they're overconfident or just prioritizing other things.
Payment giant Stripe is reportedly acquiring AI gateway startup OpenRouter for over $7 billion; OpenRouter's CEO previously called it “Stripe for AI.”
Take: If this goes through, Stripe's play in AI infrastructure gets serious. Combining payments with AI model access opens up some interesting possibilities.
ByteDance signed a memorandum of understanding with the Motion Picture Association, agreeing to rein in its AI models to protect Hollywood's intellectual property, following concerns from studios like Disney.
Take: This is an inevitable hurdle for AI content generation. Big tech companies proactively engaging with content owners is smarter than waiting for lawsuits.
Groq, formerly an AI chipmaker, raised $350 million at a $3.5 billion valuation to pivot into a “neocloud” business, expanding its Nvidia-powered data center footprint.
Take: Chips not selling well enough? Pivoting to a cloud service, even if it means using Nvidia's hardware, is a strong move for survival. Selling services on top of others' tech is a common play.
Alibaba has launched a laptop-ready AI model and released the weights for its most powerful Qwen model, intensifying its open-weight AI rivalry with Meta.
Take: Lightweight, locally runnable models are the future for true AI adoption. This open-source competition between Alibaba and Meta is great for developers.
A court has ruled that a judge who relied entirely on AI for a judicial order is still covered by judicial immunity.
Take: This ruling is a bit of a precedent, giving AI a green light in the legal system, but it leaves the question of accountability for AI errors pretty vague. Who takes the fall when the AI messes up?
An AI-generated “Autofix” feature in GitHub Copilot was found to have enabled the compromise of Snowflake's Jira system.
Take: AI tools are convenient, but security remains a huge vulnerability. Features like “autofix” could easily become backdoors if not rigorously vetted.