U.S. limits on high-end computing exports are tightening China’s access to Nvidia chips, a key resource for Chinese companies trying to keep pace in AI. In other chip news, Broadcom shares fell after a report said Google was cutting shipments of its AI chips, which feature in many of Broadcom’s products.
Anthropic’s public debut drew strikingly different figures across the supplied reports: one says the Claude maker raised $81 billion in the largest IPO ever, with shares opening more than 30% above the offer price and closing up 17%; another reports $117 billion raised through its IPO and private placements. A separate account values its market debut at nearly $1.4 trillion. Despite the attention, the listings did not revive enthusiasm for AI stocks, as investors continued reassessing risks and Anthropic expected both revenue and expenses to rise sharply.
AI’s demand for power is also shaping Redwood Materials’ strategy. CTO Colin Campbell, a Tesla veteran, is helping the company turn used EV batteries into storage for AI data centers; the report says this could become more important to Redwood than recycling. Meanwhile, Amazon SageMaker AI now flags inference endpoints with rising latency, idle GPUs or missed performance improvements. Users can benchmark alternatives for response time, throughput and cost per million tokens, then decide whether to apply changes.
Competition in consumer AI continues: JPMorgan analyst Mark DeVries says Google Gemini could challenge Meta’s Muse, describing Gemini as more advanced and easier to use. Gemini can answer questions, generate text and create art. A PYMNTS survey found 22.5% of consumers had used generative AI to manage money in the year through September 2026. Among surveyed AI users, 65% used ChatGPT and 52.9% Gemini; people used these tools for budgeting, savings and credit tips, comparing products, checking fees and drafting messages about disputes or support.
Research points to limitations and risks beyond market competition. CompAct-Bench tests whether language-model agents can compress work history and still complete long tasks in information seeking, software engineering and workspace settings. Compression often reduced accuracy, including for leading Claude and GPT models, sometimes by distorting task status or losing earlier results. A separate social-network study found that sharing beliefs alongside opinions increased the number of belief echo chambers, though those chambers involved fewer people.
Markets and regulators are responding to uncertainty around AI. The Nasdaq Composite fell about 16.3% from its October 6 peak near 27,600 to close near 23,100 on November 20, entering its third correction of 2026 and erasing its year-to-date gains. AI stock losses and concern about tighter Federal Reserve policy weighed on shares; the Nasdaq fell more than 3% that day and the S&P 500 about 2%. The European Central Bank plans targeted on-site checks at a small number of banks, including reviews of loan origination standards, as it examines concerns about circular financing in AI. It also plans to assess banks’ resilience to geopolitical risks and market shocks.
Key Takeaways
- U.S. export limits are restricting China’s access to high-end computing and Nvidia chips, challenging Chinese AI companies.
- Reports give differing figures for Anthropic’s financing and debut: $81 billion raised in one account, $117 billion in another, and a reported valuation near $1.4 trillion; shares closed 17% above the offering price in one report.
- Anthropic’s high-profile debut did not revive the AI trade as investors continued to reassess risk.
- Google reportedly cut AI chip shipments, sending Broadcom shares lower.
- Redwood Materials is exploring used EV batteries for AI data-center storage, a business that could eclipse recycling in importance.
- Google Gemini is seen by JPMorgan analyst Mark DeVries as a potential challenger to Meta’s Muse; Gemini offers question answering, text generation and art creation.
- In a survey, 22.5% of consumers used generative AI for money management; 65% of surveyed AI users used ChatGPT and 52.9% used Gemini.
- CompAct-Bench found that compressing agents’ work histories often hurt task accuracy, including for leading Claude and GPT models.
- The Nasdaq entered its third correction of 2026, down about 16.3% from its October 6 peak and erasing its gains for the year.
- Amazon SageMaker AI now identifies inference endpoint latency, idle GPU and performance issues, while the ECB plans bank inspections that include scrutiny of AI-related circular financing concerns.
US chip limits threaten China’s AI ambitions
New U.S. rules limit China’s access to high-end computing power, dealing a blow to Beijing’s artificial intelligence ambitions. Chinese tech companies were already struggling to keep pace with U.S. competitors in advanced AI. The article focuses on their reliance on Nvidia chips and the challenges they face as access tightens.
Broadcom shares fall after report Google cuts AI chip shipments
Broadcom shares fell after a report said Google is cutting shipments of its AI chips. The chips are a key component in many of Broadcom’s products. The report added to an already bad day for the company’s stock.
Anthropic’s IPO fails to revive the AI trade
Anthropic, an artificial intelligence model maker, raised $117 billion through its initial public offering and private placements on Thursday. The company is holding onto cash as it expects both revenue and expenses to rise sharply. But the IPO did not revive the AI trade, which had been fading since early October as investors reassessed the risks.
Anthropic’s record IPO succeeds as AI doubts persist
Anthropic, the company behind Claude, raised $81 billion in the largest IPO ever. Its shares opened more than 30% above the offering price on Thursday, then closed up 17% as the broader market fell. The debut was a success for Anthropic and its backers, but doubts about AI remained. The article also previews a proposed Ukraine peace deal from Jared Kushner and a popular snack aboard the USS George Washington.
Redwood Materials weighs AI battery storage against recycling
Redwood Materials CTO Colin Campbell, a Tesla veteran, is helping turn used electric vehicle batteries into a business that can power AI. The company’s work connects battery reuse with energy needs at AI data centers. The article’s headline says AI battery storage could become more important to Redwood than recycling.
Anthropic’s IPO fails to reignite the AI trade
The article examines why Anthropic’s large initial public offering did not revive investor interest in AI stocks. Business Insider investing reporter William Edwards covers the US stock market and the wider economy. The article is available only to Business Insider subscribers.
Anthropic’s huge debut puts AI in the market spotlight
Anthropic, the company behind the Claude chatbot, began public trading in a stock market debut valued at nearly $1.4 trillion. The article describes it as one of the largest debuts in history, behind SpaceX’s record-setting debut in June. Financial expert Kelly Brothers joined KCRA to discuss what the listing could mean for the market. He also discussed the release of Grand Theft Auto VI and its possible impact on Wall Street.
CompAct-Bench tests how AI agents handle compressed context
Researchers introduced CompAct-Bench to test how well language model agents can compress their work history and still finish long tasks. The benchmark covers information seeking, software engineering and workspace tasks, using BrowseComp, SWE-bench and an internal CompanyBench evaluation. Tests found that context compression often reduced task accuracy, even for leading Claude and GPT models. Mistakes came not only from missing information but also from distorted task states, such as treating unresolved issues as settled or forgetting earlier results.
JPMorgan sees Google Gemini challenging Meta’s Muse
JPMorgan analyst Mark DeVries said Google’s Gemini could challenge Meta’s personal AI agent, Muse, and support gains in Google stock. He described Gemini as more advanced and easier to use. Gemini can answer questions, generate text and create art. The article reported DeVries’s comments on November 19, 2026.
Consumers use AI to compare cards and understand fees
A PYMNTS Intelligence survey found that 22.5% of consumers had used generative AI to manage money in the 12 months before September 2026. Among AI users surveyed from April through September, 65% used ChatGPT and 52.9% used Gemini, while older users favored Gemini by about five percentage points. Power users often asked AI for savings or credit tips, budgets and product comparisons. They also used it to review bank or card fees and draft messages about disputes or support, while product comparisons were the most common measured task among mainstream users.
Nasdaq enters its third correction of 2026 as AI shares fall
The Nasdaq Composite fell about 16.3% from its October 6 peak of approximately 27,600, closing near 23,100 on November 20. That decline put the index in its third correction of 2026 and erased its gains for the year. AI stock losses and concern that the Federal Reserve may tighten policy weighed on the market. The Nasdaq fell more than 3% on November 20, while the S&P 500 dropped about 2% during the day. The current decline is deeper than the Nasdaq’s corrections in March and April and in July, both of which were followed by rebounds to record highs.
SageMaker AI flags rising latency and idle GPUs on inference endpoints
Amazon SageMaker AI now monitors inference endpoints for rising latency, idle GPUs, and missed performance improvements as workloads change. It sends findings through AWS Health, with one notification for each endpoint. Users can open the endpoint in Amazon SageMaker Studio and benchmark its current setup against alternatives for response time, throughput, and cost per million tokens. These benchmarks use separate computing resources, and the endpoint does not change unless users choose to apply a result. The article explains how to review, apply, or dismiss recommendations and how to opt an endpoint out of monitoring.
Study finds belief sharing can increase the number of echo chambers
Researchers studied how sharing beliefs along with opinions affects echo chambers in social networks. They expanded an existing opinion dynamics model with belief merging and kept agents’ beliefs and opinions coherent according to shared values. Their measures examined echo chambers through homogeneity, segregation, and reinforcement. Simulations found echo chambers with belief sharing, and showed that linking beliefs to opinions increased the number of belief echo chambers. Those chambers involved fewer agents.
ECB plans 2027 checks on banks’ AI investment risks
The European Central Bank plans targeted on-site inspections of a small number of banks to review their loan origination standards. Sharon Donnery, an ECB supervisory board member, announced the plan at a JPMorgan event on November 19. The checks come as concern grows about circular financing in the AI industry. The ECB also plans to focus on banks’ ability to withstand geopolitical risks and market shocks.
Sources
- China’s AI Firms Are Addicted to Nvidia Chips. Now They Are Facing Withdrawal.
- Broadcom Stock Falls on Report Google is Cutting AI Chip Shipments
- Anthropic IPO Didn’t Save the AI Trade
- Anthropic Lives Up to Wall Street Hype, but AI Doubts Linger | The 10-Point for Nov. 20
- Redwood Materials CTO says AI battery storage could eclipse recycling
- Why Anthropic's big IPO couldn't revive the AI trade
- Making Cents: What does an AI company entering public trading mean for the market?
- CompAct-Bench: A Benchmark for Working-Context Compaction in Long-Horizon Agent Tasks
- Google's Gemini Tasks Seen As Challenger To Meta's Muse
- Consumers Ask AI Twice: Which Card and Why the Fee?
- Nasdaq slides into its third correction of 2026 as AI stocks stumble
- Amazon SageMaker AI now detects latency drift, idle GPUs, and missed optimizations on your inference endpoints
- Coherent belief and opinion propagation produces more echo chambers: corrected version
- Reg Wrap: ECB tells JPMorgan it will inspect banks’ AI investment risks in 2027
Comments
Please log in to post a comment.