Anthropic Plans IPO With $900.9 Billion in Commitments

Key Takeaways

• Holiday Extras says Codex rebuilt its iOS and Android travel app in two weeks; open-to-book conversion rose 32.5% and errors fell from 19% to 9%.
• Holiday Extras is extending Codex-assisted software building beyond engineering, while engineers review work before production.
• Amazon Bedrock inference costs can be affected by cache duration, service tier, reasoning effort and maxTokens; LiteLLM can provide an OpenAI-compatible gateway with AWS monitoring and controls.
• OpenAI’s Business Hub lets verified firms manage business information and feeds; eligible U.S. firms can offer clearly labeled Business Agent chats, including Shopify merchants.
• Anthropic plans an IPO of 825 million shares at $70–$78 each; its prospectus lists about $900.9 billion in purchase and lease commitments, mostly for infrastructure.
• Anthropic uses chips from Amazon Web Services, Google, Nvidia and AMD.
• Articles raise questions about Nvidia succession and talent risks following Jensen Huang’s reported resignation, while also noting continued investor focus on AI demand.
• Anne Neuberger’s proposed cyber plan targets five critical sectors, calls for zero exposed control systems and tested 72-hour manual operations, and proposes a joint drill within three to six months.
• Robinhood says more than 200,000 customers use agentic trading features; the company has over 28 million customers and reported a record equity-trading day.
• GMGN.AI supports meme-token trading across Solana, BNB Chain, Ethereum, Base, Monad and TRON, with copy trading, limit, take-profit and stop-loss orders.

Holiday Extras rebuilds its app with Codex in two weeks

Holiday Extras used OpenAI’s Codex to rebuild its travel app in two weeks, compared with more than six months for similar work before. The company says open-to-book conversion rose 32.5% and the app error rate dropped from 19% to 9%. The rebuilt app supports both iOS and Android, and the company moved from major annual releases to weekly updates. Holiday Extras first gave staff access to ChatGPT Enterprise, then used Codex to help customer experts and engineers build software.

Holiday Extras rebuilds its travel app in two weeks

Holiday Extras rebuilt its travel app in two weeks using Codex, after a similar rebuild was expected to take six months. Open-to-book conversion rose by 32.5%, while the app error rate fell from 19% to 9%. The company first provided staff with ChatGPT Enterprise, then used Codex to help customer experts build software. Liam Thompson, VP of AI Innovation and former design lead, said the question had shifted from whether he could build something to how he would build it.

Holiday Extras opens Codex software building to all staff

Holiday Extras is using Codex to let staff outside engineering build software, with engineers still involved before work reaches production. The company said a new app prototype took two days and reached production in two weeks, while the previous version took more than a year of design and testing with engineers. OpenAI also reported that a finance employee with decades of accounting experience rebuilt and improved costly finance systems. Company leaders said the time from an idea to working software is shrinking, and they are watching whether this approach works as more non-engineers deliver software.

Four Amazon Bedrock settings can change inference costs

Four request-level settings can affect the cost of Amazon Bedrock inference: cache duration, service tier, reasoning effort, and maxTokens. These choices change the number of tokens the service processes or the rate charged. The article explains how the settings were tested and how to check their effects. Costs may continue to rise even after teams tune prompts and application logic.

Build a self-service LLM gateway with Amazon Bedrock and LiteLLM

The article explains how to build a self-service large language model gateway using Amazon Bedrock Application Inference Profiles and LiteLLM. Developers can use an OpenAI-compatible API to connect tools such as Claude Code, Continue, and Cursor through one base URL and API key. LiteLLM runs on AWS Fargate, while Amazon RDS stores persistent key and spending data, ElastiCache for Valkey supports caching and rate limits, and AWS Secrets Manager stores secrets. The setup also uses Amazon CloudWatch for usage and quota monitoring and Langfuse for prompt-level tracking. The gateway gives platform, finance, and security teams ways to manage performance, costs, and access controls.

ChatGPT adds business profiles and direct chats for sellers

OpenAI has introduced Business Hub, where verified firms can update their details and add feeds with pricing and availability. Eligible ecommerce and online service firms in the U.S. can also use Business Agents, which let users open a separate, clearly labeled chat with a business from its ChatGPT answer. Eligible Shopify merchants can connect their stores through Business Hub or Shopify admin, using products already listed in Shopify Catalog. ChatGPT may also show a firm in relevant answers even if it has no Agent. The launch could bring extra dashboard work for small firms, and some insights will remain limited until businesses reach volume levels.

Anthropic IPO puts Amazon, Alphabet and Nvidia ties in focus

Anthropic, the maker of Claude AI models, aims to sell 825 million shares for $70 to $78 each. CNBC reported that pricing may happen on Nov. 18, with trading starting the next day. Founded in 2021, Anthropic sells Claude through its platform and apps, including the Claude Code coding agent. Its prospectus lists about $900.9 billion in purchase and lease commitments, mostly for computing and infrastructure, with about $70.7 billion due in the 12 months after Sept. 30. The company uses chips from Amazon Web Services, Google, Nvidia and Advanced Micro Devices, and the article ranks six AI stocks by how directly they benefit from Anthropic’s spending.

Jensen Huang’s Nvidia exit raises concern about talent loss

Jensen Huang resigned as Nvidia’s CEO after admitting to a relationship with a woman described as an alleged Chinese intelligence asset. Huang co-founded Nvidia in 1993 and played a central role in its chip design, strategy and public image. The company’s GPUs, first developed for computer games, became important for AI because they can process many equations at once. Analysts cited in the article see no obvious successor and warn that Huang’s departure could distract the company or lead to a broader loss of talent. Nvidia employs tens of thousands of workers, including leading chip designers, but questions remain about who will replace Huang and whether the company can keep its technical work on track.

Starbucks CEO Brian Niccol discusses the company’s next steps

Yahoo Finance featured Starbucks CEO Brian Niccol in a video about the company’s comeback. Niccol discussed the next step in expanding Starbucks’ offerings. The article provides no further details about the plans.

Neuberger urges CEOs to lead national grid cyber defense

Anne Neuberger proposed a national cyber plan covering five sectors and a few hundred operators, warning that AI is shortening the time between finding a weakness and exploiting it. The plan calls for defenders to find exposed systems, set priorities and deliver tested workarounds quickly. It covers power, banking and payments, water, communications and hospitals, with sector councils committing to zero exposed control systems and proven 72-hour manual operations. The plan would have ISACs send anonymized monthly scorecards to a White House force, while AI labs scan code and firmware and report findings through Gold Eagle. It also calls for long-term renewal of the Cybersecurity Information Sharing Act of 2015, liability protection for shared fixes, and a joint drill modeled on GridEx within 3 to 6 months.

Robinhood sets trading record as prediction markets and AI tools grow

Robinhood Markets had its highest ever equity trading day on the Thursday before a Wolfe Research event, Chief Brokerage Officer Steve Quirk said. He highlighted prediction markets, agentic trading and wealth management as areas for growth, and said more than 200,000 customers use agentic trading features. Robinhood has more than 28 million customers, and the average account size has grown to about $15,000 from about $3,000. Quirk also pointed to international expansion and adviser services as ways to support deposit growth, while noting uncertainty around sports contracts in prediction markets.

GMGN.AI offers fast meme token trading across six blockchains

GMGN.AI is a multi-chain trading platform for meme tokens. Users can trade on Solana, BNB Chain, Ethereum, Base, Monad and TRON while finding new launches and trending tokens. The platform is available on the web, mobile app, Telegram bot, gmgn-cli and AI agents. It supports 24/7 copy trading, limit orders, take-profit orders and stop-loss orders.

Investors discuss Nvidia leadership risks and strong AI demand

Storm Uru, co-head of Liontrust’s Global Innovation Team, and Sarah Kunst, managing director of Cleo Capital, discussed Nvidia and investment opportunities in a video. They covered the company’s leadership crisis, succession risks and demand for AI. The video was published on November 12, 2026, and runs for 28 minutes.

Sources

• Holiday Extras is rethinking what it can build with Codex
• Holiday Extras rebuilt its app in two weeks
• Holiday Extras lets everyone build with Codex
• Four Amazon Bedrock inference choices that affect cost
• Building a self-service LLM gateway with Amazon Bedrock and LiteLLM
• ChatGPT Is Now Shelf Space for Business
• Forget Buying All 7 "Magnificent Seven" Stocks. Here's How to Rank 6 Top AI Stocks Right Now (Hint: It Has to Do With Anthropic.)
• Jensen Huang leaving Nvidia risks a 'brain drain' from the AI chipmaker
• Yahoo Finance Invest: AI in Action
• Neuberger wants CEOs to own grid defense
• Robinhood Markets Hits Record Trading Day, Sees Growth in Prediction Markets and AI Trading
• GMGN.AI | The Fastest Multi-Chain Meme Trading Terminal
• Panel discusses Nvidia turmoil, AI demand and investment opportunities