Anthropic IPO buyers weigh $1.1 trillion; Meta pauses its terms

Some potential buyers of shares in Anthropic's coming IPO now think the company could be valued as low as $1.1 trillion, Bloomberg reported, against the roughly $2 trillion investors were said to expect about a month ago. The same day, Bloomberg reported that OpenAI has raised $11 billion so far in a round that values it at $1.25 trillion. The round could reach about $20 billion, the reduced target first reported last week; OpenAI had hoped for as much as $30 billion at a valuation of $1.4 trillion. Both moves follow falling tech stocks and midterm results that ran against data center projects.

Companies used more AI tokens than ever in August and September, but spending on them grew just 1.5 percent across the two months, according to new data from the corporate card company Ramp. In July, spending had grown 42 percent from the month before. Ramp economist Ara Kharazian said nearly all AI spending comes from venture-backed tech companies, and that the heaviest users, who spend close to 10,000 dollars per employee each month, dipped slightly. A typical company spends 12 to 15 dollars per employee per month, so the gap between the two groups is enormous.

Meta has shelved the commercial terms that would have let it crawl and train on any website linked from a Facebook, Instagram or WhatsApp business account. "We're pausing the changes," a spokesperson told Reuters, adding that the existing terms remain in effect. Meta had already pushed the start date from October 30 to November 20 after publishers objected, and courts in Brazil and Norway issued preliminary injunctions blocking the terms. Nearly every news publisher keeps a Facebook page, so the change would have reached far beyond Meta's own platforms.

The Republican chairs of the House committees on China, foreign affairs and intelligence have asked Nvidia chief executive Jensen Huang to appear at a joint closed-door briefing on November 18. They want to know what security protocols Nvidia uses against foreign espionage, after reports that US intelligence officials believe Huang was targeted through an executive at Megaspeed, a Singapore cloud company that bought billions of dollars of Nvidia chips. Nvidia said its regional cloud customers receive "zero information relevant to national security," and would not say whether Huang will attend.

Crusoe confirmed that Google has acquired its data center site in southeast Wyoming, a project planned to scale to 10 gigawatts with the energy infrastructure company Tallgrass Energy. Crusoe had tried and failed to sign Google as a customer there, and Google reportedly raised concerns about cost and timing under Crusoe's management, TechCrunch reported. Google now plans to invest $17 billion in the campus. The two companies still work together: Crusoe is building another data center for Google near Amarillo, Texas.

Internet-scanning firm Censys found 1,465 servers running an exposed console for Artex, an open-source agent that automates attacks from reconnaissance to data theft. About 650 of those sit in identical hosting blocks, leaving roughly 815 distinct deployments, and 32 servers expose open directories holding implants and attack kits. CrowdStrike has tied Artex to attacks on South Korean banks, and Microsoft has linked it to a group that compromised ConnectWise customers. Censys said many operators are likely authorized testers, but the tool has no need to be reachable from the internet at all.

OpenAI released BioBasicsBench, 100 lab-analysis tasks built with the enzyme-design company Ligo from scientists' own raw data: gel images, colony plates, microscopy and instrument readouts. Tasks ask agents to interpret results, spot errors and recognize when the evidence cannot support a conclusion. At its highest reasoning setting, GPT-6 Astra earned 66.7 percent of the available credit, up from 43.8 percent for GPT-5.6 Sol. OpenAI says models still fall short on reading images and checking an experiment's controls.