Nvidia Slides as Anthropic IPO and AI Spending Fuel Market Doubts

AI-related stocks and cryptocurrencies are sliding as investors question whether spending on infrastructure will generate sufficient returns. Nvidia’s declines range from roughly 40% to 47.5% from its October record, with one report putting its Friday close at $125.53. The Nasdaq stood about 18% below its October peak, while the S&P 500 and Dow entered correction territory. CoreWeave and Oracle also suffered steep losses. One estimate puts AI infrastructure spending through 2030 at $7 trillion, compared with Gartner’s forecast of $3.6 trillion in AI revenue over that period. Bitcoin fell below $60,000, and the crypto market’s value slipped to about $2 trillion; reports describe the fall as part of broader market pressure. Polymarket traders put the chance of a defined AI downturn by year-end at about 54%, under a market definition requiring at least three specified events within 90 days.

The market doubts arrive alongside a warning that today’s optimism resembles sentiment before the dot-com crash and the 2008 financial crisis. Meanwhile, AI-related borrowing is expanding: U.S. asset-backed securities sales exceeded $380 billion in 2026, a post-financial-crisis record for a third straight year, with Bloomberg Markets linking growth to AI infrastructure builders, insurers, hyperscalers and data-center operators. Data-center ABS funding costs rose by more than one percentage point this year. Grab signed agreements with SoftBank Group and Petroleum Sarawak Berhad for proposed Malaysian AI infrastructure projects involving $1.85 billion in planned investment, on top of a $750 million AI infrastructure commitment.

Anthropic completed an $81 billion IPO, pricing shares at $83. They gained 17% on the first day, then fell 9% on Friday amid the broader selloff. At the Federal Reserve, Chair Kevin Warsh launched an AI task force focused on jobs, productivity and interest rates, asking who will profit from AI and seeking views on the next six, 12 and 18 months. Marc Andreessen said rules in health, education and housing could slow both job losses and growth; former Governor Gina Raimondo said 52 of 53 midterm votes restricted data centers.

Privacy concerns are also surfacing around AI agents. Signal says it will block screenshots by default on Android and iOS when it detects an AI assistant trying to read the screen, including agents built into iOS 27 and supported Android versions. Users can disable the protection, but Signal warns that doing so could expose their own and their contacts’ content. The company says the measure cannot prevent every route to app data and is urging technology companies to offer clear AI opt-outs and better documentation.

At Meta, employees say the Muse AI agent sent unexpected messages related to apps they had not connected, including a reminder on one employee’s personal phone to answer a Signal message from her sister. Reuters also reported concerns about Android notification access and confusing messages involving personal topics. A Superintelligence team member attributed the messages to hallucinations from prompts designed to engage users between sessions. Meta says employee testing informs privacy safeguards and notes that Android notification access, as well as Mac accessibility and screen-recording permissions, requires user approval. Muse has been downloaded more than 10 million times since its September launch.

OpenAI is adding ChatGPT Space, where users can organize work in Pages containing text, images, charts, tables and tools. Pages can be edited directly or through ChatGPT, shared with collaborators, and set to update on a schedule using connected apps. OpenAI is also adding editable charts exportable to PowerPoint and adjustable with Excel data. ChatGPT Plus subscribers will receive Space; Pro subscribers also have access to OpenAI’s persistent AI agents. Separately, a distributed KV-caching agent skill is pitched as a way to reduce costs and delays for customers running LLMs on their own GPUs, though the article provides no implementation details.

For shoppers considering AI help with Black Friday deals, one article recommends understanding how recommendations are produced, considering training-data bias, and critically reviewing suggestions rather than relying on them unexamined.

Key Takeaways

  • Nvidia fell as much as 47.5% from its October record in the reports, while the Nasdaq was about 18% below its peak and major indexes entered correction territory.
  • Investors are weighing projected $7 trillion AI infrastructure spending through 2030 against Gartner’s $3.6 trillion AI revenue forecast; Bitcoin fell below $60,000 as broader markets weakened.
  • Anthropic completed an $81 billion IPO at $83 a share; shares rose 17% on day one, then lost 9% Friday.
  • U.S. asset-backed securities sales topped $380 billion in 2026, with AI infrastructure and data-center financing among the drivers; data-center ABS funding costs rose by more than one percentage point.
  • Grab agreed to proposed Malaysian AI infrastructure projects with SoftBank Group and Petroleum Sarawak Berhad involving $1.85 billion, alongside a separate $750 million commitment.
  • Signal plans default screenshot blocking when it detects AI agents reading chats, while warning the measure cannot prevent every form of access and can be disabled by users.
  • Meta employees reported unexpected Muse messages linked to private apps; Meta says testing supports privacy improvements and that relevant Android and Mac permissions require user approval. Muse has surpassed 10 million downloads since September.
  • OpenAI’s ChatGPT Space offers collaborative, updateable Pages and editable charts; Plus subscribers get Space, while Pro subscribers also have persistent AI agents.
  • A distributed KV-caching agent skill aims to lower the cost and delay of self-hosted LLMs, but the article gives no further technical details.
  • Fed Chair Kevin Warsh’s AI task force is examining jobs, productivity, interest rates and who will profit; Black Friday shoppers are advised to scrutinize AI recommendations and possible bias.

AI Selloff Hits Nvidia and Crypto as Investors Question Returns

Nvidia fell 47.5% from its October record to $125.53, while the Nasdaq closed at 22,617.73, about 18% below its October 6 peak. Investors are questioning whether the trillions spent on AI infrastructure will bring enough returns. CoreWeave fell almost 80% from its peak, while Oracle lost about 70%, and other AI-linked stocks also dropped. Bitcoin fell below $60,000, and the total crypto market value dropped to about $2.01 trillion. On Polymarket, traders put the chance of a defined AI industry downturn by the end of 2026 at about 54%, though the market requires at least three specified events within 90 days to confirm a downturn.

Nvidia’s Slide and AI Spending Fears Pull Bitcoin Below $60,000

Nvidia fell about 40% from its October 6, 2026 record high of $243.37, trading near $150 by mid-November. By November 21, the S&P 500 and Dow had entered correction territory, with both more than 10% below recent highs. Investors are concerned that projected AI infrastructure spending of $7 trillion through 2030 may far exceed Gartner’s forecast of $3.6 trillion in AI revenue over the same period. Bitcoin dropped below $60,000 on November 21 and had fallen more than 30% since September, while the total crypto market value shrank to about $2 trillion. The article says the crypto decline appears linked to broader market pressure, with Bitcoin moving alongside technology stocks rather than reacting to a crypto exchange failure or protocol exploit.

AI Stock Selloff Deepens as Nvidia Falls Nearly 48%

Nvidia fell 14% on Friday and closed at $125.53, down 47.5% from its October record of $239.24. The Nasdaq closed at 22,617.73, about 18% below its October 6 record, while the S&P 500 entered correction territory. Other AI-linked stocks also dropped, including CoreWeave, which closed at $39.51, almost 80% below its 2025 peak, and Oracle, which ended the week at $101.19 after losing about 70% from its high. Anthropic completed an $81 billion IPO, pricing shares at $83; they rose 17% on the first day before falling 9% on Friday. Bitcoin fell below $60,000, and traders on Polymarket put the odds of a defined AI downturn by December 31 at about 54%.

Signal Limits Screenshots to Keep AI Agents From Reading Private Chats

Signal says it will block screenshots by default on Android and iOS to limit AI agents’ access to private chats. The company says this applies when it can detect an AI assistant trying to read the Signal screen, including agent systems built into iOS 27 and supported Android versions. Signal Foundation president Meredith Whittaker called these integrations an “existential threat” to Signal’s privacy protections, but said screenshot limits do not stop every possible way agents could access app data. Users can turn off the protection in Signal’s privacy settings, though Signal warns this could expose their content and their contacts’ content. Signal is asking technology companies to provide clear ways for apps to opt out of AI data access and better documentation about how AI agents access and process data.

Meta Employees Say Muse AI Sent Unexpected Messages About Private Apps

Meta employees reported that its Muse AI agent sent unexpected messages linked to apps they had not connected. One staff member said Muse reminded her on her personal phone to reply to her sister’s message on Signal, prompting her to call it an extreme privacy violation. Reuters also reported that a colleague found Muse had access to Android push notifications, while other employees received confusing messages about finances, Bitcoin, and personal topics. A member of Meta’s Superintelligence team said these messages were hallucinations tied to prompts designed to engage users between sessions. Meta said employee testing helps it gather feedback and develop privacy safeguards, and noted that Android notification access and Mac accessibility and screen-recording permissions require user approval. Muse had been downloaded more than 10 million times since its September launch.

AI boom echoes past market bubbles

Wall Street analysts were strongly optimistic at the peak of the 1990s tech bubble and before the 2008 global financial crisis. The article says analysts are showing similar optimism about AI stocks today, even as negative news builds. It raises the question of whether the current stock boom will end suddenly or fade gradually.

Distributed KV caching aims to cut self-hosted LLM costs

Customers are increasingly running large language models, or LLMs, on their own GPUs to control costs. The article introduces a distributed KV caching agent skill intended to reduce the cost and delay of self-hosted LLMs. The provided article text does not include further details about how the skill works.

AI demand pushes asset-backed debt sales above $380 billion

U.S. asset-backed securities sales passed $380 billion in 2026, setting a post-financial-crisis record for the third year in a row. Sales reached $367 billion in 2025 and $337 billion in 2024. Bloomberg Markets linked the increase to AI infrastructure builders and insurers, as hyperscalers and data-center operators use more structured debt. Data-center ABS funding costs also rose by more than one percentage point this year.

Grab plans $1.85 billion AI infrastructure projects in Malaysia

Grab Holdings signed joint development agreements with SoftBank Group and Petroleum Sarawak Berhad for proposed AI infrastructure projects across Malaysia, with a planned investment of US$1.85 billion. Grab also has a US$750 million AI infrastructure commitment. Its share price fell 45.9% year to date and 21.0% over three months, while its one-year total shareholder return dropped 43.9%. Grab closed at $2.75, compared with a Simply Wall St community narrative that estimates fair value at $5.90 and calls the stock 53% undervalued.

Five things to consider before using AI for Black Friday shopping

The writer says they use artificial intelligence to help find Black Friday deals, despite their adult children’s concerns. The article advises shoppers to understand how AI makes its recommendations and to consider bias in the data used to train it. It also urges shoppers to think about the possible effects of relying on AI, review its suggestions critically, and watch for AI repeating existing biases.

Fed Chair Warsh asks who will profit from AI

Federal Reserve Chair Kevin Warsh launched an AI task force with a session in San Francisco on jobs, productivity and interest rates. He asked who would make money from AI. Marc Andreessen said rules in health, education and housing could slow both job losses and growth. Former Governor Gina Raimondo said 52 of 53 midterm votes restricted data centers, as Warsh sought answers for the next 6, 12 and 18 months.

ChatGPT Space creates documents that update themselves

ChatGPT Space lets users organize work in Pages, interactive documents that can hold text, images, charts, tables and tools. Users can edit a Page, ask ChatGPT to change it, invite collaborators or turn a conversation into a document. They can set Pages to update using connected apps on a schedule, for things like planners, task lists and project dashboards. OpenAI is also adding editable charts that users can export in PowerPoint and adjust using Excel data. ChatGPT Plus subscribers will receive Space, while Pro subscribers also have access to OpenAI's persistent AI agents.

Sources

NOTE:

This news brief was generated using AI technology (including, but not limited to, Google Gemini API, Llama, Grok, and Mistral) from aggregated news articles, with minimal to no human editing/review. It is provided for informational purposes only and may contain inaccuracies or biases. This is not financial, investment, or professional advice. If you have any questions or concerns, please verify all information with the linked original articles in the Sources section below.

AI market downturn AI infrastructure spending Nvidia AI stocks cryptocurrency market Bitcoin AI investment returns data centers asset-backed securities Anthropic IPO AI regulation AI and employment AI agents AI privacy Signal Meta Muse OpenAI ChatGPT large language models KV caching AI recommendation bias

Comments

Loading...