AI news, 18 August: Anthropic revenue run rate tops $65 billion

Anthropic’s annualized revenue surges past $65 billion. According to Reuters and Bloomberg, Anthropic’s revenue run rate topped $65 billion by the end of July—up from $47 billion in May and about $9 billion at the end of 2025. The Claude developer shared the figure with investors as it prepares for a possible IPO later this year. TechCrunch reported that investors expect full-year 2026 revenue in the $100–120 billion range, while OpenAI’s run rate was recently put near $40 billion. Anthropic was valued at $965 billion after a major funding round in May; strong enterprise demand for its coding and assistant tools is widely seen as the main driver.

Update: Nvidia commits up to $105 billion guarantee for OpenAI’s Ohio data center. Nvidia said it will invest $1.5 billion in SoftBank-owned SB Energy and provide a guarantee of up to $105 billion tied to OpenAI’s long lease of a huge data-center campus in Pike County, Ohio, Reuters reported. Nvidia will be the exclusive chip supplier for a site that could eventually reach about 8 gigawatts, with initial capacity eyed for 2028. CEO Jensen Huang said the deal secures long-lived infrastructure for Nvidia compute and is not “circular financing.” The package marks a fresh, concrete step beyond earlier talks and underlines how deeply chipmakers are now financing the power and buildings their customers need.

ByteDance signs AI copyright pact with Hollywood. ByteDance and the Motion Picture Association agreed on Monday to strengthen copyright safeguards on ByteDance’s Seedance video and Seedream image models, according to Reuters. The deal follows a February cease-and-desist push by the MPA and concerns from studios including Disney that the tools could generate protected characters and likenesses without permission. Newer model versions include tougher intellectual-property protections, ByteDance said; both sides pledged to keep working on safeguards as the technology evolves.

Music publisher sues Anthropic and Suno over training data. Independent publisher Round Hill Music filed California federal lawsuits alleging that Anthropic and AI music company Suno used lyrics from at least 500 of its songs—including material linked to James Brown, The Kinks, and the Goo Goo Dolls—to train their systems, Reuters reported. Round Hill said it may add thousands more compositions and is seeking damages that could approach or exceed $1 billion. Spokespeople for Anthropic and Suno did not immediately comment. The cases join a widening wave of copyright suits by rights holders against AI firms.

Amazon scanned rare books for AI training, investigation finds. 404 Media reported—and TechCrunch amplified—that Amazon has bought rare books, cut off their spines, and scanned them at a Las Vegas facility for AI training. Amazon said it purchases books through commercial channels to improve products and services. Rare and out-of-print texts are attractive because they often cannot be found online and predate generative AI, reducing the risk of models training mainly on synthetic text. The story has fueled fresh public debate over how far companies should go to feed data-hungry systems.

ECB blog: AI-fueled tech rally likely faces a correction. A European Central Bank staff blog argued that a correction in elevated U.S. tech valuations driven by AI optimism is likely, citing patterns from past technological revolutions and investor over-exuberance, according to Reuters. The post does not represent formal ECB policy. It noted substantial euro-area household and institutional exposure to major U.S. tech stocks and warned that, unlike the dot-com era, policymakers now have less room to cut rates or use fiscal tools if a sharp drop coincides with broader market stress. Timing, the authors stressed, cannot be known in advance.