OpenAI reportedly cuts funding target to $20bn, Intel shares crash
OpenAI's smaller funding round, Intel's crash, a fake Claude Code update in Cursor, Google's AI satellite, Apple Intelligence in China, Anthropic's credit rating, SemiAnalysis's new test and OpenAI's biology model.
OpenAI cut its funding target to $20bn, the FT says That’s down from $30bn, and the first chunk of money is expected to be smaller still, the FT reported on Friday, with UAE investors leading. The Wall Street Journal says OpenAI has $11bn lined up so far and some big investors were lukewarm. The last valuation talked about was $1.25tn. Source
Intel’s shares crashed almost 29 per cent, worst day since 1974 It happened on Friday, after Intel forecast weaker sales than expected for the last three months of the year. Its boss Lip-Bu Tan blamed a PC slowdown caused by high memory prices and short supply, and AI data centres are soaking up that memory. The Nasdaq fell almost 3 per cent, with Nvidia, AMD and Arm all down too. Source
A fake Claude Code update hit up to 5,000 Cursor users Someone published a tampered version of Anthropic’s Claude Code extension on Cursor’s marketplace, and it was live for about 11 hours on Thursday. It added the attacker’s key so they could log in to your machine and sent back details about it, and uninstalling may not clear it. Anthropic says its own systems weren’t touched, and Cursor is contacting the people affected. Source
Google’s first data centre satellite is running AI in orbit It’s a prototype called Suncatcher, with four of Google’s own AI chips that run one at a time and its Gemma model on board. After three weeks of testing, the first thing engineers asked it was to write a haiku about being in space. Google said on Friday it’s working normally, and two more satellites are planned for 2027. Source
Apple Intelligence is finally coming to China, with Alibaba’s Qwen It arrives with iOS 27.3 in December, and developers can test it from 4 November, Apple said on Friday. Alibaba’s Qwen plugs into Siri, Writing Tools and Visual Intelligence, but China doesn’t get the new Siri Apple showed off this year. Source
Anthropic faces a junk rating after its IPO, the FT says Blackstone, which anchored the $63bn deal to pay for Anthropic’s chips, has told investors in it to expect a double-B rating, the top end of junk, which makes borrowing more expensive. It had earlier said Anthropic would be investment grade before 2027. The FT reported it on Friday, and Anthropic and Blackstone wouldn’t comment. Source
Top AI model scores 12 per cent on SemiAnalysis’s new test It’s called GDPx and gives models jobs of up to eight hours across 11 fields, from chip design to biology to law, and pretty much every model failed the finance and legal tasks outright. SemiAnalysis put it out on Monday and says Claude Opus 5.5 is the best all-rounder, though safety filters can drag down scores in areas like biology and cyber. Source
OpenAI’s new biology model roughly doubles the old one’s scores GPT-Rosalind-6 Research is GPT-6 Astra with fewer refusals on biology questions, and it came out on Friday for approved researchers only. Its biggest jump was on genetics, from 17 to 38 per cent. OpenAI rates its biology and chemistry risk as High, one step below its top level. Source