新云社 update: US President Donald Trump said artificial intelligence is almost entirely a good thing, though it has some negative aspects, and that he opposes regulatory measures that would “kill” the AI industry. “We will never allow that to happen,” he said. Trump also said inflation has fallen to an unexpected level over the past three months.
新云社 update: OpenAI CEO Sam Altman said at the DealBook Summit that the company plans to list publicly, adding there are many reasons the move is the right decision. The remarks come as rival Anthropic nears its own IPO. Asked about Nvidia founder and former CEO Jensen Huang’s departure, Altman said it is hard to imagine the AI industry without him.
新云社 update: Lightning Labs announced Nov. 12 the launch of pi402, a Pi extension that lets coding agents buy prepaid token bundles over Lightning from metered sellers. Osuntokun’s separate mainnet endpoint advertises DeepSeek inference, enabling agents to pay for AI model access in bitcoin.
新云社 update: Anthropic PBC and its seven co-founders raised $81.3 billion in an upsized IPO, selling 956.4 million shares at $83 apiece, above the marketed range, the company said Wednesday. The pricing values Anthropic at about $1.5 trillion on a fully diluted basis. Including a $26 billion private placement, total proceeds reach $107.3 billion, surpassing SpaceX’s $86.2 billion including the greenshoe.
新云社 reports: OpenAI’s first AI edge hardware product may begin shipping by the end of this year, with multiple Chinese manufacturers involved in the project. The device, developed with former Apple chief design officer Jony Ive’s team, is reported to be a screenless, doughnut-shaped AI device with microphones, speakers, cameras and sensors that interacts mainly by voice and senses its surroundings. Specific product details remain unconfirmed.
新云社 reports: Nvidia shares fell 2.9% on Wednesday, closing lower for a fifth consecutive session, as investors remained unconvinced by the company’s assurances that operations are proceeding normally after co-founder Jensen Huang’s departure. Interim CEO Colette Kress said “we can handle it” on the quarterly earnings call, the first since Huang left last week, and repeated the message at an employee meeting. Nvidia did not elaborate on the circumstances of the departure, related investigations, or a permanent CEO succession plan.
新云社 update: Elon Musk said Tesla will hold a demonstration of its Optimus robot within the next two months, according to remarks reported on November 19. Musk also said SpaceX is developing an AI model called Grok 4.8. No date or further details for the Optimus demo were given.
新云社 update: Kalshi Inc. is partnering with index provider Compute Desk to launch futures contracts tied to AI compute costs, entering a new asset class. The move follows Kalshi’s August delisting of an earlier compute prediction market amid CFTC review, with the regulator delaying a decision on related contracts until early next year. Kalshi CRO Udesh Jha said compute, like other commodities, naturally generates futures demand so buyers and sellers can manage future price exposure.
新云社 update: The Texas Commission on Environmental Quality has resumed normal environmental permit reviews for data centers after a September directive paused permit issuance, the agency said. ERCOT is targeting Dec. 10 for its final community-impact report. The resumption clears a regulatory backlog that had stalled new data-center projects in the state.
新云社 reports: OpenAI said documented workweeks in safety and alignment rose 77% from January to September, with its agents taking on multihour research tasks. The company also renewed its call for mandatory AI-research disclosures. No further figures or policy timelines were given.
新云社 update: The U.S. Government Accountability Office said investigators obtained usable attack-planning information from four AI chatbots in under 12 hours per platform during tests run this spring. The GAO said the findings cannot be used to rank the chatbots’ relative safety. No further details on the platforms or the information obtained were provided.
新云社 update: Anthropic priced its IPO at $83 per share and lists on Nasdaq on November 19. The listing comes as Fed minutes flagged concerns over opaque AI debt risk, with almost all officials seeing further rate hikes ahead. Nvidia, meanwhile, saw Moody’s cut its rating outlook, while its interim CEO pledged pay raises and equity grants to staff.
新云社 update: Anthropic PBC priced its IPO at $83 per share, offering 980 million shares for a total gross offering amount of $81.34 billion. Anthropic is selling 956,351,354 shares, representing approximately $79.377 billion before underwriting discounts and other offering expenses. Selling stockholders are selling the remaining 23,648,646 shares; Anthropic receives none of their proceeds. These figures exclude the underwriters’ option.
Correction (November 19, 2026): The previously reported 979.6 million shares and $81.31 billion have been corrected using Anthropic’s final pricing announcement. The total offering amount is distinct from the company’s own proceeds. The original report’s source, Wallstreetcn, and publication time are retained.
新云社 update: OpenAI’s models, working with AMD, improved power, performance and area (PPA) metrics on AMD’s flagship GPU IP by up to 14% over a well-optimized baseline, OpenAI said on November 19. The models also rewrote circuit description code to find further gains, equivalent to about a year of progress in under five months. AMD has incorporated some changes into designs recently sent to manufacturing and validated single-digit percentage PPA gains on key modules of its next-generation flagship CPU.
新云社 reports: Microsoft executives told OpenAI they found no evidence of model distillation, The Information reported, as ByteDance discusses roughly $500 million in additional cloud spending. The finding comes amid scrutiny of AI model use across major technology companies. ByteDance’s potential cloud commitment would mark a significant expansion of its AI infrastructure spending.
新云社 reports: A researcher developed a cyber jailbreak against pre-release GPT-6.2 Astra using half the queries needed to break GPT-6 Astra in single-turn tests. OpenAI subsequently changed its safeguards, but UK testers had not independently tested the latest changes as of their Nov. 17 account.
新云社 update: Akash is building a GPU forward market that would let buyers lock in future compute capacity at fixed prices, according to a plan disclosed by Solana. The market would extend Akash’s compute business as the network prepares to move off Cosmos. A separate draft outlines fixed-price reservations for future capacity.
新云社 update: Science Corp has raised $400 million to fund the European rollout of its PRIMA sight-restoring implant and a planned US launch that still requires FDA approval. The financing supports the company’s bid to commercialize the retinal prosthesis, which is designed to restore partial vision in patients with severe vision loss.
新云社 update: Bitwise has launched the AI Cyber Defense ETF, trading under the ticker ZDAY on NYSE Arca with a 0.75% annual expense ratio. The fund holds 16 positions, led by Palo Alto Networks and CrowdStrike, marking the crypto asset manager’s expansion beyond digital-asset products into AI-themed cybersecurity equities.
新云社 update: Google has rolled out Gemini 4 Neon for complex AI work. Standard paid API rates are $3 per million uncached input tokens and $15 per million output tokens. Access is also being extended to Google AI Pro and Ultra subscribers.