AI Stocks Slide as Anthropic and Nvidia Face Investor Scrutiny

AI-linked stocks face renewed scrutiny over whether expensive data-center and model investments can produce returns. The Nasdaq fell 5.28% on Friday in one report, while other accounts put its weekly decline at 10%; the S&P 500 and Dow entered correction territory after falling 10% from recent highs. The Magnificent Seven lost more than $4 trillion in combined market value over about six weeks. Investors cite high valuations, borrowing costs, regulation and uncertainty about data-center construction. Analysts disagree over whether the sell-off signals an AI bubble or a correction; a 10% market decline has occurred regularly over more than a century, while a bear market is a drop of at least 20%.

Anthropic’s record-breaking market debut did little to calm investors. Its shares rose after listing but by Friday traded less than $6 above the offering price. The company raised $91 billion in its debut, yet said it may need to borrow in the future. Bloomberg Economics’ Anna Wong said Anthropic’s results showed weakening pricing power for leading AI services. Investors also question whether AI companies can meet expectations and finance data centers, particularly if interest rates rise or construction faces new restrictions. AI firms are financially interconnected through investments, loans, chip sales and data-center financing, while construction and investment have also supported jobs and economic growth.

Market unease also centers on Nvidia, after reports of turmoil around CEO Jensen Huang and alleged ties to a Chinese spy. Nvidia shares fell for seven consecutive sessions despite reporting record quarterly revenue of $115.8 billion and forecasting $130 billion for the fourth quarter. Investor Michael Burry sold half his put options on Nvidia and the VanEck Semiconductor ETF to lock in gains, while keeping the rest and leaving his bearish April 2027 Nasdaq 100 puts unchanged. Some market observers point to Meta and Apple as examples of companies with underlying strength.

Anthropic says all 10 of the largest Fortune 500 banks are customers of Claude. Barclays uses Claude to process about 120,000 client emails daily, Citadel applies it to financial modeling, and BNY uses it in operational workflows. FIS and Amalgamated Bank project that a Claude-powered anti-money-laundering agent could reduce investigation time by about 98%; that is an expected result, not a measured outcome.

Several new AI tools focus on enterprise use and oversight. TrustHouse’s Secure AI Gateway is designed to govern agent identities and permissions, tool and Model Context Protocol access, data governance, business-unit boundaries and action records. It is meant to work with existing identity providers, security controls, infrastructure and model gateways. Fortress.ai launched Fortress Command, which combines AI agents, threat intelligence and human review to help utilities, federal agencies and defense organizations identify and resolve supply-chain cyber risks.

Thinking Machines released Inkling 2 Mini, an open-weight model with 26.4 billion total parameters and 3.4 billion active parameters. Its roughly 20GB checkpoint can run on Apple computers with at least 32GB of memory, an NVIDIA DGX Spark or GPUs with at least 32GB of VRAM. It handles text and images as input and generates text, and is available under Apache 2.0 on Hugging Face. The company reports 79.2% on SWE-bench Verified using bash only and a 20.1 Artificial Analysis Index score, tied with Qwen3.8-27B; it also reports weaker scores on HLE-Diamond and MultiChallenge.

Government and public-sector proposals raise questions about AI access and recordkeeping. US lawmakers Andrew Garbarino and John Moolenaar recommend barring Chinese-developed AI models from federal systems and high-risk critical infrastructure, including federal contract work, and preventing federal cybersecurity grants from paying for them. Their report says US organizations are adopting lower-cost Chinese open-weight models and argues that the United States lacks equally competitive alternatives. In Ulster County, a proposed policy would require people to create constituent records, with exceptions for approved speech-to-text and translation tools that do not train on county information. Committees are scheduled to consider the proposal on Dec. 10 and, if approved, the full Legislature on Dec. 15.

Other AI deployments span education, marketing and software development. Texas Sports Academy received more than $6.5 million from the state’s new voucher program—over twice as much as any other private school. Its 13 schools serve grades 3–12; students use an AI learning program for two hours each morning before sports. Brevo offers AI-assisted email drafting, personalization, automated campaigns, audience segmentation and performance reporting. Together AI, meanwhile, reported that a coding-agent incident exposed prompts involving requests to MiniMax M3, GLM-5.2 and Kimi K2.6. It found no evidence of further leakage, believes the data was deleted, and says downloaded prompts cannot be retrieved. The company tightened agent rules, adopted scoped or short-lived credentials where supported, and made zero data retention the default for new accounts.

Key Takeaways

  • Nasdaq and other major indexes fell sharply as investors questioned AI valuations, data-center costs, borrowing needs and potential regulation; the S&P 500 and Dow entered corrections.
  • Anthropic’s $91 billion market debut did not stop the sell-off; its shares ended Friday less than $6 above the offering price, and the company said it may need to borrow in the future.
  • Nvidia shares fell for seven sessions despite $115.8 billion in quarterly revenue and a $130 billion fourth-quarter forecast; Michael Burry sold half his Nvidia and semiconductor ETF puts.
  • Anthropic says the 10 largest Fortune 500 banks use Claude. Barclays processes about 120,000 client emails a day with it, while FIS and Amalgamated Bank project a 98% reduction in anti-money-laundering investigation time.
  • TrustHouse’s Secure AI Gateway targets controls for agent identities, permissions, tools, data and action records; Fortress.ai’s platform uses AI agents and human review to address supply-chain cyber risks.
  • Thinking Machines’ 26.4-billion-parameter Inkling 2 Mini runs on specified 32GB Apple computers and GPUs, supports text and image input, and is available under Apache 2.0 on Hugging Face.
  • US lawmakers propose restricting Chinese-developed AI in federal systems and high-risk critical infrastructure, citing its lower cost and the lack of equally competitive US alternatives.
  • Ulster County proposes requiring human creation of constituent records, with limited exceptions for approved speech-to-text and translation tools; committee consideration is scheduled for Dec. 10.
  • Texas Sports Academy received more than $6.5 million in voucher funds; students at its 13 schools use an AI learning program for two hours each morning.
  • Together AI says a coding-agent incident exposed prompts involving MiniMax M3, GLM-5.2 and Kimi K2.6; it tightened controls and made zero data retention the default for new accounts.

AI Stocks Slide as Investors Question Spending and Profits

Investors are growing concerned about the huge cost of developing artificial intelligence and whether companies can earn enough to justify it. The Nasdaq fell 5.28 percent on Friday, wiping out its gains for the year, while the seven major technology companies known as the Magnificent Seven lost more than $4 trillion in combined market value in about six weeks. Concerns include restrictions on data centers after the US midterm elections and turmoil at Nvidia, where CEO Jensen Huang resigned amid reports about alleged ties to a Chinese spy. Anthropic, the maker of Claude AI, raised $91 billion in its market debut, but said it may need to borrow money in the future, adding to concerns about AI companies’ debt and complex financial ties.

A 10 Percent Market Drop Is Unsettling but Common

The S&P 500 fell 10 percent from its recent high, a decline Wall Street calls a correction. Such drops have happened regularly for more than a century, though investors worry a correction could turn into a bear market, which means a fall of at least 20 percent. Fears about AI valuations, costly data center investments, borrowing needs, possible interest rate increases, regulation, and Nvidia leadership turmoil have driven the recent sell-off. Some analysts say the AI boom may be a bubble, while others view the decline as a panic-driven correction; even Anthropic’s market debut did not stop the selling.

AI Stock Losses Raise Fears for Markets and the Economy

AI-related stocks and broader markets fell sharply, raising concerns that the AI boom could be a bubble and that its decline could affect the financial system and economy. The Nasdaq fell 10 percent for the week, while the S&P 500 and Dow entered corrections, meaning they had dropped 10 percent from recent highs. Investors cited concerns about tougher rules, data center construction, and turmoil at Nvidia after CEO Jensen Huang resigned. Some experts warn that AI investments may not deliver the returns investors expect, while others point to the strength of companies such as Meta and Apple. The article also notes that AI companies are closely connected through investments, loans, chip sales, and data center financing, and that data center construction and AI investment have supported jobs and economic growth.

Anthropic IPO Fails to Ease Wall Street’s AI Concerns

Anthropic’s record-breaking initial public offering did not stop technology stocks from falling, and the Nasdaq 100 dropped 10 percent for the week, its worst showing since March 2020. The stock rose after its debut but later fell to less than $6 above its public offering price by Friday. Investors remain concerned about AI companies’ ability to meet high expectations, the cost of funding data centers, possible interest rate increases, and risks to data center construction after Democratic midterm victories. Bloomberg Economics economist Anna Wong said Anthropic’s financial results showed weakening pricing power for leading AI services. Nvidia’s leadership turmoil has also unsettled investors, while analysts warned that higher borrowing costs could lower company valuations.

TrustHouse Introduces AI Gateway for Enterprise Risk Controls

TrustHouse announced a Secure AI Gateway designed to manage risks beyond access to AI models as companies use autonomous and semi-autonomous agents. Its approach covers agent identities and permissions, access to tools and Model Context Protocol servers, business unit boundaries, data governance, and records of AI actions. The company says policies can control which agents use approved skills, connectors, and enterprise services, and can assess whether requested actions are allowed in their business context. TrustHouse Chief AI Officer Chiru Bhavansikar said organizations need control over the identities, permissions, tools, and services AI systems use. The approach is intended to work with existing identity providers, security controls, AI infrastructure, and model gateways rather than replace them.

Fortress.ai launches AI platform to tackle critical supply chain risks

Fortress.ai introduced the Fortress Command Platform to help utilities, federal agencies and defense organizations find and address cyber risks in their supply chains. The platform combines AI agents and Fortress’s threat intelligence with human review. It analyzes customer data alongside information about vendors, products and vulnerabilities, then helps customers and suppliers resolve important risks. Fortress CEO Alex Santos said the goal is to give organizations clear priorities and show which issues they have fixed.

Michael Burry takes profits on Nvidia and chip ETF bets as tech stocks fall

Investor Michael Burry sold half of his put options on Nvidia and the VanEck Semiconductor ETF, locking in gains after a steep technology sell-off. He kept the remaining options and left his bearish April 2027 Nasdaq 100 puts unchanged. Burry also sold about one-third of his Zoetis call options and increased his stake in MercadoLibre by 20%. Nvidia shares had fallen for seven straight sessions, even after the company reported record quarterly revenue of $115.8 billion and forecast fourth-quarter revenue of $130 billion.

Barclays, Citadel and BNY use Anthropic’s Claude in financial work

Anthropic said the 10 largest Fortune 500 banks are its customers, as financial firms use its Claude AI for research, coding, compliance and operations. Barclays uses Claude to process about 120,000 client emails each day, and Citadel applies it to financial modeling. BNY uses Claude in operational workflows. FIS and Amalgamated Bank expect a Claude-powered anti-money laundering agent to cut investigation time by about 98%, though that figure is a projection, not a measured result.

Texas sports and AI school network receives over $6.5 million in voucher funds

Texas Sports Academy has received more than $6.5 million from Texas’s new voucher program, more than twice the amount received by any other private school. Its 13 schools serve students in grades 3 through 12 and focus on sports such as baseball, basketball, football and volleyball. Students spend two hours each morning using an AI-based learning program before focusing on sports. The network’s founder, Mackenzie Price, also created Alpha School and has donated more than $1.5 million to Gov. Greg Abbott and other supporters of the $1 billion voucher program.

US lawmakers call for limits on Chinese AI in government and critical infrastructure

Republican lawmakers Andrew Garbarino and John Moolenaar urged a ban on Chinese-developed AI models in US federal systems and high-risk critical infrastructure. A congressional report made 12 recommendations, including blocking the models from federal contract work and preventing federal cybersecurity grants from paying for them. The lawmakers said US companies are adopting Chinese open-weight models because they offer capable technology at lower costs, while the US lacks equally competitive alternatives. Their seven-month investigation also examined AI use at Airbnb, Anysphere and DoorDash.

Ulster County proposes limits on generative AI in public records

Ulster County Legislature Majority Leader Abe Uchitelle introduced a proposal to limit generative AI use in creating and maintaining constituent records. The policy would require a person to create county records, with exceptions for approved speech-to-text and translation tools that do not use county information to train their models. Uchitelle said AI could create biased language that may affect residents for years because county records can be permanent. He asked elected officials and department heads to review the proposal and share how they use generative AI. The Ways and Means Committee is scheduled to consider it on Dec. 10, followed by the full Legislature on Dec. 15 if approved.

Thinking Machines releases 26.4B Inkling 2 Mini for 32GB computers

Thinking Machines introduced Inkling 2 Mini, an open-weight model with 26.4 billion total parameters and 3.4 billion active parameters. It reuses its middle layers three times per token, allowing its roughly 20GB checkpoint to run on Apple computers with at least 32GB of memory, one NVIDIA DGX Spark, or GPUs with at least 32GB of VRAM. The model accepts text and images and produces text, and it is available under the Apache 2.0 license on Hugging Face. Thinking Machines reported a score of 79.2% on SWE-bench Verified using bash only and 20.1 on the Artificial Analysis Index, tied with Qwen3.8-27B. Its reported weaker results include 15.2% on HLE-Diamond and 50.0% on MultiChallenge.

Brevo uses AI and automation to support email marketing

Brevo is a digital marketing platform that uses artificial intelligence and automation to help businesses manage email campaigns. Its features can help marketers draft email copy and subject lines, personalize messages, and send automated emails after customer actions, such as signing up or leaving items in a shopping cart. Brevo also offers audience segmentation and reports on open rates, clicks, and conversions. The article says these tools can save time, support more relevant campaigns, and help businesses improve engagement and sales. It describes possible uses for online stores, educational institutions, and service-based businesses.

Together AI reports coding agent exposed user prompts

Together AI said a coding agent incident exposed user prompts involving requests to MiniMax M3, GLM-5.2, and Kimi K2.6. The company said five credentials used in testing were Dragos canaries and that it found no evidence of further leakage beyond the incident. Together AI worked with an outside forensic firm and law enforcement, and said it believes the data was deleted, though downloaded prompts cannot be retrieved. Its review of the prompt contents is still underway to decide who should receive direct notice. The company says it now uses scoped or short-lived credentials where supported, checks credentials on pull requests, tightened rules for AI coding agents, and made zero data retention the default for new accounts; existing users must enable it in account settings.

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 stocks AI market volatility AI investment data centers Anthropic Claude Nvidia AI valuations enterprise AI AI governance AI security cybersecurity open-weight models Inkling 2 Mini AI regulation Chinese AI models public-sector AI AI in banking AI education data privacy

Comments

Loading...