Anthropic IPO Raises $107.3 Billion as AI Stocks Fall

Anthropic’s IPO drew heavy trading but did little to calm concerns about AI-sector valuations and spending. The company priced shares at $83, opened at $111.50 and closed its first session at $97.17, up 17.1%; its stock then fell about 9% to 6% on Friday, Nov. 20, depending on the report. Anthropic offered 980 million shares. The offering raised $107.3 billion, including a $26 billion private placement involving Nvidia, Google and AMD, making it the largest IPO on record. Retail investors traded about $1.4 billion in shares and received roughly 15% of the offering, while more than half of about 1,200 institutional bidders received no shares.

The debut coincided with broad market losses. U.S. indexes fell at least 2% on Thursday, and the S&P 500 recorded its sixth straight losing session; futures also declined later that evening. Nvidia, AMD, other chipmakers and cloud infrastructure companies fell as investors questioned AI valuations, data-center costs and whether the spending boom can last. Asian technology stocks also slid on Friday, including South Korea’s Kospi, led lower by SK Hynix and Samsung Electronics. Markets priced in a 96.4% chance of a 25-basis-point Federal Reserve rate increase in December and awaited the next inflation report.

Debate over the scale and durability of AI investment extends beyond markets. Investor Paul Kedrosky says the White House fears that slower AI spending could weigh on stock prices and President Donald Trump’s publicized 401(k) balances, describing data-center construction as private-sector stimulus supported by federal actions. He says Texas ranks third worldwide in data-center spending, behind the rest of the United States and China. Texas Gov.-elect Gina Hinojosa has pledged to pause new approvals until lawmakers pass protections; PIMCO expects permitting delays to slow construction, and cloud companies may move projects elsewhere. Kedrosky also says reduced Nvidia backing for cloud providers could limit their access to loans for buying Nvidia chips.

U.S. chip policy is also changing. On Nov. 16, 2026, the Commerce Department’s Bureau of Industry and Security issued an interim final rule controlling remote access to advanced chips and rescinding the AI Diffusion Rule. Licenses may be required when advanced AI chips provide significant computing power to entities in countries under U.S. arms embargoes, including China, and for certain advanced Super Intelligence items sent to Indonesia, Malaysia, Singapore, Brazil and Thailand. Exporters must consider who ultimately accesses the computing power; later access changes can also trigger licensing. The rule covers some foreign-made products containing U.S. advanced chips, and BIS is accepting comments through Jan. 19, 2027. Separately, China Semiconductor Industry Association chairman Shaojun Wei forecasts China’s chip sales will exceed CNY1.5 trillion (US$225 billion) in 2026, an 89% year-over-year increase, driven by AI demand and domestic-sector growth.

Practical AI deployment and security featured in other reports. Ohio’s AI Summit, presented by OpenAI and hosted by OhioX, sold out for a third consecutive year, drawing 415 people to Columbus State Community College on Nov. 18. Sessions examined AI uses in insurance, manufacturing, order processing and business systems, alongside workforce preparation, oversight, data centers and energy. A separate guide to AI-generated code warns of exposed secrets, injection flaws, unsafe dependencies, authorization gaps and open database rules. Escape’s scan of public vibe-coded apps found more than 400 exposed secrets and 175 personal-information exposures, most reachable without authentication; its Cascade 2 Iguazu tool tests changes and reruns verified exploits as regression tests.

Security agencies and researchers also outline ways to manage AI systems. The UK National Cyber Security Centre says prompt injection may never be fully preventable: malicious instructions in inputs or external material can alter an agent’s behavior, potentially leading it to misuse access to email, databases, records or payments. Its advice is to restrict what agents can access and do. In travel planning, researchers propose a supervising chatbot that coordinates specialized agents to clarify requests, check schedules, retrieve live information and build itineraries. On the TravelPlanner benchmark, the system improved results over the ReAct baseline by 10% for budget constraints, 25% for broader constraints and 45% for personalization.

Key Takeaways

  • Anthropic’s $83 IPO raised $107.3 billion, including a $26 billion private placement involving Nvidia, Google and AMD; it was reported as the largest IPO on record.
  • Anthropic opened at $111.50 and closed its debut session at $97.17, up 17.1%, before falling roughly 9% to 6% on its first full trading day, according to different reports.
  • Retail investors traded about $1.4 billion of Anthropic shares and received roughly 15% of the offering; more than half of about 1,200 institutional bidders received no allocation.
  • U.S. stocks, Nvidia, AMD, cloud providers and Asian technology shares fell amid concern about AI valuations, data-center spending and interest rates; markets priced in a 96.4% chance of a 25-basis-point December Fed increase.
  • A new U.S. rule adds licensing controls for certain remote access to advanced chips, rescinds the AI Diffusion Rule and accepts public comments through Jan. 19, 2027.
  • China’s semiconductor industry association chairman forecasts 2026 chip sales above CNY1.5 trillion (US$225 billion), an 89% annual increase, citing AI demand and domestic growth.
  • Investor Paul Kedrosky says data-center permitting delays in Texas could slow construction and prompt cloud companies to relocate projects; he also links Nvidia backing to cloud providers’ ability to finance chip purchases.
  • Ohio’s OpenAI-presented AI Summit sold out for a third year, drawing 415 attendees to discuss workplace applications, oversight, skills, data centers and energy.
  • Escape found more than 400 exposed secrets and 175 personal-information exposures in public AI-generated apps; the article recommends testing generated code and live applications before release.
  • The UK NCSC advises limiting AI agents’ access and permitted actions because prompt injection may not be fully preventable; a multi-agent travel planner beat the ReAct baseline by 10%–45% across benchmark tasks.

Anthropic Shares Slide as AI Worries Drag Down US Tech Stocks

Anthropic shares fell 6% on Friday after jumping more than 17% in their market debut the day before. The decline added to worries about the cost and funding of AI infrastructure, and major US stock indexes were heading for their steepest weekly drop since April 2025. Nvidia fell 3.7%, while other chipmakers and cloud companies also lost ground. Investors were also watching for clues about Federal Reserve rate decisions, with markets pricing in a high chance of a December rate increase.

US Stocks Drop More Than 2% as Anthropic Debut Fails to Lift AI Stocks

US stock indexes fell at least 2% on Thursday, and the S&P 500 recorded its sixth straight session of losses. Anthropic shares opened at $111.50, then gave up much of their early rise to close 17.1% above their $83 IPO price. Nvidia, AMD, chipmakers and cloud companies also fell as investors questioned AI valuations and spending on data centers. Markets were pricing in a 96.4% chance of a 25-basis-point Federal Reserve rate increase in December, and investors awaited the next personal consumption expenditures inflation report.

Anthropic IPO Draws Record Trading as Retail Investors Buy and Sell Shares

Anthropic priced its IPO at $83 a share, opened at $111.50 and closed at $97.17. The offering raised $107.3 billion, including a $26 billion private placement involving Nvidia, Google and AMD, making it the largest IPO on record. Retail investors traded about $1.4 billion worth of shares, with net buying estimated at $80 million to $90 million; they received about 15% of the offering. More than half of roughly 1,200 institutional bidders received no shares, while AI-linked stocks including AMD and Nvidia fell.

Anthropic Debut Fails to Ease AI Concerns as US Stock Futures Fall

US stock futures fell on Thursday evening after major indexes dropped more than 2% during the regular session. Anthropic shares closed 17.1% above their $83 IPO price at $97.17, after opening at $111.50 and briefly pushing the company’s value above $2 trillion. Investors remained concerned about technology valuations, AI spending and the cost of data center projects. Nvidia, AMD and cloud infrastructure providers fell, while markets also expected a likely Federal Reserve rate increase in December and awaited next week’s inflation report.

Anthropic Shares Fall About 9% as AI Stocks Face More Pressure

Anthropic shares dropped about 9% on their first full trading day, after gaining 17% on their first day. The stock had opened at $111.50, above its $83 IPO price, and closed its debut session at $97.17. The article links the decline to pressure across AI stocks and concerns about high valuations, the cost of data center projects and rising interest rates. Anthropic’s debut was expected to test confidence in the AI sector, but its early gains did not prevent a sharp retreat.

Anthropic shares drop 9% as AI stocks face a wider selloff

Anthropic shares fell about 9% on Friday, November 20, their first full trading day, as AI-related technology stocks also declined. The company priced its IPO at $83 per share and traded as ANT. Shares opened at $111.50 and closed their first day at $97.17, up 17% from the IPO price, after reaching a higher peak during the session. Anthropic offered 980 million shares. Reuters and CNBC cited concerns including high AI stock valuations, the cost of data centers, rising interest rates and questions about whether the AI boom can last.

Ohio AI Summit draws 415 people for its third straight sellout

The Ohio AI Summit sold out for the third year in a row, bringing 415 people to Columbus State Community College on November 18. Presented by OpenAI and hosted by OhioX, it brought together businesses, colleges, government agencies and nonprofits from across Ohio to discuss how they use AI and prepare workers. Sessions covered AI in insurance, manufacturing, order processing and business systems, as well as the need for employee oversight and ways to measure results. Speakers included Nationwide’s Michael Carrel, Path Robotics’ Randle, State CIO Katrina Flory and Ohio Secretary of State Frank LaRose. Panels also discussed data centers, energy, workforce skills and how organizations can fit AI into daily operations.

Asian shares fall as AI stock losses continue despite Anthropic IPO

Asian equities fell on Friday as technology stocks extended a selloff, despite Anthropic’s IPO. South Korea’s Kospi declined, led by major chipmakers SK Hynix and Samsung Electronics. The article links the market drop to ongoing pressure on AI-related stocks.

How to find security flaws in AI-generated code before release

AI code security means finding and fixing weaknesses in code written by AI tools and agents before it reaches production. Risks include exposed secrets, injection flaws, unsafe dependencies, authorization gaps and open database rules. The article recommends checking generated code before each release and testing the running app, since code scanners may miss problems such as one user accessing another user’s data. It cites Escape’s scan of public vibe-coded apps, which found more than 400 exposed secrets and 175 personal information exposures, with most flaws reachable without authentication. Escape’s Continuous AI Pentesting tool, Cascade 2 Iguazu, tests changes and reruns proven exploits as regression tests.

US adds controls on remote access to advanced chips and ends AI Diffusion Rule

On November 16, 2026, the Department of Commerce’s Bureau of Industry and Security issued an interim final rule adding export controls on remote access chips and rescinding the AI Diffusion Rule. The controls can require a license when advanced AI chips provide significant computing power to entities in countries under US arms embargoes, including China, Cuba, Iran, North Korea and Russia. The rule also requires licenses for certain advanced Super Intelligence items sent to Indonesia, Malaysia, Singapore, Brazil and Thailand. Exporters must check who will ultimately access the computing power, and later changes in access may trigger a license requirement even if the original shipment did not need one. The rule also covers certain foreign-made products containing US advanced chips, and BIS will accept public comments until January 19, 2027.

Paul Kedrosky says White House fears an end to the AI spending boom

Venture investor Paul Kedrosky says the White House is afraid that slowing AI spending could hurt stock prices and President Donald Trump’s publicized 401(k) balances. He describes data center spending as a private sector stimulus supported by federal actions. Kedrosky says Texas ranks third worldwide in data center spending, after the rest of the United States and China, and notes that Gov.-elect Gina Hinojosa has pledged to pause new approvals until lawmakers pass protections. PIMCO expects permitting delays to slow construction, while major cloud companies may shift projects elsewhere. Kedrosky also argues that if Nvidia backs fewer cloud providers, those companies may have less access to loans for buying Nvidia chips, which could slow AI spending.

China chip sales forecast to top CNY1.5 trillion in 2026

Shaojun Wei, chairman of the China Semiconductor Industry Association, projects that China’s semiconductor sales will exceed CNY1.5 trillion, or US$225 billion, in 2026. That would be an 89% increase from the previous year. Wei attributed the expected growth to rising demand for artificial intelligence and rapid expansion in China’s domestic chip sector. The forecast was reported from Taipei on November 20, 2026.

Multi agent AI system creates personalized travel plans

Researchers propose a travel planning system that uses a supervising chatbot to coordinate specialized AI agents. The agents check schedules, clarify user requests, retrieve real time information, and create itineraries that meet constraints. The system combines large language model reasoning, external APIs, and repeated improvements guided by a scoring framework. On the TravelPlanner benchmark, it improved budget constraint planning by 10%, planning with wider sets of constraints by 25%, and personalized planning by 45% compared with the ReAct baseline.

NCSC says companies must limit what AI agents can access

The UK’s National Cyber Security Centre said in a December 8, 2025 analysis that prompt injection may never be fully prevented. Prompt injection occurs when instructions in user input or outside material, such as emails and documents, change how a language model behaves. Unlike SQL injection, it has no equivalent fix that keeps instructions separate from data, the agency said. As AI agents gain the ability to send emails, query databases, change records, and make payments, a manipulated agent could take harmful actions using company credentials. The NCSC says companies should focus on limiting what their AI agents are allowed to access and do.

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.

Anthropic IPO AI valuations AI investment data center spending Nvidia AMD Google semiconductor industry AI chip exports U.S. export controls China chip industry cloud infrastructure financial markets Federal Reserve Texas data centers AI deployment AI-generated code security prompt injection AI agents multi-agent systems

Comments

Loading...