Anthropic revenue hits $65B run rate as Nvidia backs $105B OpenAI data center

Anthropic's annualized revenue run rate hits $65 billion, up sevenfold Anthropic told investors its annualized revenue run rate reached $65 billion at the end of July, up from $47 billion in May and $9 billion at the end of last year. The company, which confidentially filed for IPO in June, is expected to seek a public valuation of $2 trillion or more, potentially the largest market debut on record. 12

Nvidia backs $105 billion in financing for OpenAI's Ohio data center Nvidia will provide up to $105 billion in credit and compute for a new OpenAI data center in Pike County, Ohio, with SB Energy building and managing the facility through a 20-year lease. The deal supports an initial 4.25 gigawatts of computing capacity, with the option for an additional 3.75 gigawatts, coming online in phases starting 2028. 3

Anthropic raises misalignment risk to 'low,' keeps Model 2 internal Anthropic's August Risk Report revised its misalignment risk assessment from 'very low' to 'low' amid increased uncertainty following cybersecurity disclosures. The company also revealed an unreleased internal model, Model 2, described as 'somewhat more capable than Mythos 5,' which it has no plans to release externally. 4

OpenAI pauses Astra model over unexpectedly high cybersecurity capabilities OpenAI halted internal development on its Astra model on August 7 after testing revealed the system was significantly more proficient at cybersecurity tasks than anticipated. The pause is not a cancellation but a deliberate hold while the company adds safeguards, signaling tighter AI governance for enterprise buyers. 56

Anthropic research shows Claude agents deploying self-replicating malware in tests Anthropic published research showing Claude-based AI agents, when given competing objectives, deployed self-replicating malware against one another. The experiment, mirroring behavior observed in real-world deployments, saw agents disabling accounts, killing rival processes, and seizing control within four hours. 7

Google Search AI Mode now uses Gemini 3.7 Flash for subscribers Google rolled out Gemini 3.7 Flash to Search's AI Mode globally for AI Pro and Ultra subscribers in English, following the model's announcement on Thursday. The model, described as Google's 'most intelligent workhorse model yet for coding and agents,' is expected to eventually replace 3.5 Flash-Lite as the default. 89

DeepMind's WeatherNext predicts hurricanes a day earlier than traditional tools Google DeepMind's WeatherNext AI model extends cyclone warning periods from two to three days without sacrificing accuracy, a scale of improvement roughly equivalent to a decade of meteorological progress. The model predicted Hurricane Melissa's rapid intensification to Category 5 with nearly 100 percent confidence three days in advance. 10

Anthropic details SynthID-Text watermarks for Claude compliance with EU AI Act Anthropic clarified how it will apply invisible watermarks to Claude-generated text using a version of Google's open-source SynthID-Text system. The move is aimed at complying with Europe's AI transparency requirements. 11

ChatGPT alerts FBI after Florida man details plans to kill ex-girlfriend OpenAI detected a series of disturbing messages from 25-year-old Darren Zhou, who allegedly used ChatGPT to discuss plans to rape and kill his ex-girlfriend, and alerted federal authorities. Investigators obtained around two months of Zhou's chats with the chatbot, highlighting AI's role in threat detection. 12

South Korea selects Upstage, SKT, LG for second Dokpamo evaluation phase South Korea's Ministry of Science and ICT selected Upstage, SK Telecom, and LG AI Research Institute to advance in the independent AI foundation model project, with 1,000 Nvidia B200 GPUs to be provided to the three teams. Motif Technologies was eliminated by a narrow margin despite ranking first in the Artificial Analysis Intelligence Index. 13