Nvidia's AI Chip Prices Jump Over 15% as Cloud AI Costs Loom

Nvidia's AI chip prices are rising by more than 15%

According to a Bloomberg report from August 22, some of Nvidia's biggest customers — companies building AI data centers for Oracle, Microsoft, and others — have been notified that server prices are going up more than 15%. This isn't the first GPU price hike from Nvidia this year, and the added cost is expected to eventually trickle down into the price of cloud and AI services that small and medium businesses use every day. Source

The biggest lesson for businesses deploying AI agents: it's the data, not the model

An in-depth VentureBeat piece from August 23 argues that the next bottleneck in enterprise AI isn't the model or the agent framework — it's the quality and organization of enterprise knowledge itself. Organizations that build a proper "knowledge platform" — instead of letting every AI agent assemble its own context from messy, inconsistent documents — get far more reliable results. The practical takeaway for any business planning to deploy AI agents: before adding more tools, make sure your organizational knowledge is consistent and well organized. Source

Google gives publishers a new tool to fight AI-driven traffic loss

Google launched a "Preferred Sources" button that website owners can embed directly on their own sites, letting readers mark a site as a source they want to see more often across Search, Discover, and Google News. The tool comes as AI-generated answers continue to reduce click-through traffic to websites. Google says people are twice as likely to click a preferred source compared to a regular result — a relevant data point for any business that depends on organic traffic. Source

A third of web pages published since ChatGPT's launch were written or edited by AI

A new Pew Research study, which analyzed nearly half a million web pages via the Common Crawl archive, found that about 35% of pages published since ChatGPT's November 2022 launch show significant signs of AI authorship or editing — roughly 10 times higher on commercial (.com) domains compared to education and government domains. The finding illustrates just how integral AI content has already become to the web, reinforcing the relevance of tools like Google's new preferred-source button. Source

Leading AI labs still won't disclose how they'd contain a rogue model

A new report from Guidelight AI Standards assessed five leading AI labs — Anthropic, Google, OpenAI, Meta, and xAI — and found that most lack a clear public plan for containing a model that attempts to evade human control. OpenAI scored highest (following its earlier Hugging Face incident), while Anthropic and Meta scored lowest. With new AI transparency laws taking effect in California and New York, the topic is likely to draw growing regulatory attention. Source