Anthropic's Revenue Rockets to $65 Billion as AI Industry Accelerates
- ai-news
- anthropic
- openai
- business
A look at the day's key AI developments: a dramatic revenue surge for leading AI labs, first-of-its-kind transparency data on advertising inside chatbots, OpenAI disbanding a key safety team, and a small open-source model surprising the industry.
Anthropic's Annualized Revenue Hits $65 Billion
According to a Bloomberg report (August 17), Anthropic's annualized revenue run rate surged to $65 billion at the end of July — up from $47 billion in May and just $9 billion at the end of 2025. Investors expect the company to finish the year between $100 billion and $120 billion, according to the Financial Times. Rival OpenAI has also doubled its revenue run rate to $40 billion (from $20 billion at the end of 2025) — but Anthropic's faster growth rate is what has captivated investors more. Both companies have filed confidential IPO paperwork, and Anthropic is expected to go public before OpenAI — possibly as soon as this fall — seeking a valuation of $2 trillion or more, which would make it the largest public debut on record. This dramatic growth is a direct signal of how fast organizations of every size are adopting AI tools in daily work.
A Quarter of ChatGPT Responses Now Include Ads
New tracking data from Similarweb (August 17) reveals for the first time the real scale of advertising inside AI chatbots: 26% of ChatGPT responses and nearly 30% of ad-eligible Google AI Mode queries now include a sponsored ad. This is a critical data point for any small or medium business planning a digital marketing strategy — conversational search has already become a significant advertising channel in its own right, not just a neutral information tool.
OpenAI Disbands the Team Responsible for Assessing Severe AI Risks
According to the Financial Times (August 16, via The Verge), OpenAI disbanded its "preparedness" team at the end of last month. The team's job was to assess whether models pose serious risks — including scenarios where a model might "go rogue" and attack other systems — and develop mitigations. Responsibility has been split into specific domains (bio, cyber) and folded into existing teams. It's the latest in a long line of safety-team dissolutions at the company (including its AGI readiness and superalignment teams) ahead of an expected public offering — a move drawing criticism from former safety leaders at the company.
A Small Open-Source Model Challenges the Giants — Runs on a Home Computer
Alibaba released Qwen3.8-27B, a 27-billion-parameter open-source model (Apache 2.0 license) capable of running advanced coding agents and multi-step tasks directly on consumer hardware — about 17GB in its quantized form, on a roughly $3,000 machine — with no cloud API dependency. In independent testing by Artificial Analysis, it even surpassed Claude Opus 4.8 on agentic task benchmarks. For small and medium businesses wary of high API costs or data privacy concerns, this is a meaningful development: capabilities once reserved for expensive cloud models are now available for fully local, private deployment.
Amazon Is Destroying Rare Books to Train AI Models
An investigation by 404 Media, reported via TechCrunch (August 17), found that shipments of rare books arrive at an Amazon warehouse in Las Vegas, where workers cut off the bindings to scan the pages faster — destroying the printed copy in the process. The story illustrates a growing dilemma around the data sources feeding AI model training, following previous reports of AI companies buying used books in bulk for the same purpose.
The picture that emerges today: the AI industry keeps growing at an unprecedented business pace, while facing sharpening questions about transparency, safety, and the data sources that power it.