Nvidia’s leadership shift is prompting questions about the financing behind the AI buildout. After Jensen Huang resigned as CEO, president and director, CFO Colette Kress became interim CEO. Nvidia shares fell as much as 14% before closing down 4.3%; rival chipmakers AMD and Broadcom each rose more than 5%. Investors are watching whether Kress will maintain investments and purchase guarantees that help customers build data centers and support demand for Nvidia chips. The company reported $99 billion in equity investments at the end of July, $25 billion in future investment commitments and $360 billion in AI cloud-capacity purchase commitments. It also agreed to provide up to $105 billion in credit support for a data-center business intended for OpenAI and has a CoreWeave agreement to buy up to $6.3 billion in capacity it cannot sell, subject to service conditions. Kress reportedly opposed a proposed $250 billion OpenAI credit-support plan. Analysts say tighter deal rules could ease concerns about circular financing but also slow infrastructure investment. Nvidia paid $10 billion in federal and state income taxes in fiscal 2025; analysts differ on the outlook, with one warning that weaker long-term investment could slow growth and others pointing to strong demand and a two-to-three-year lead over its nearest GPU competitor. The board is considering internal and external CEO candidates, and Nvidia is due to report third-quarter earnings on Tuesday. A report also said President Donald Trump urged Huang not to resign.
Separate Reuters investigations found Chinese AI projects worth more than $3 billion—and, in one account, more than $3.2 billion—specify Nvidia chips subject to US export controls. Plans include a 6.2 billion yuan Fujian center naming H100s, a 3 billion yuan Shaanxi facility specifying H100 and H200 chips, a 2.26 billion yuan Jilin project and a proposed Inner Mongolia development listing 625 H200 servers. Wuxi authorities said an H800 cluster began operating in March 2025. The project totals do not establish that all processors were delivered, or that export laws were broken. Continued reliance on Nvidia systems may leave developers exposed to delays and higher costs, while switching to alternatives can require software changes and extra engineering.
AI is also changing university work and industrial operations. A survey of 1,054 UK undergraduates found 95% use AI and 94% use it for assessed work; the share who paste AI-generated text directly into assignments rose from 3% in 2024 to 12% in 2026. Trinity College Dublin checks essays through Turnitin in Blackboard, while its Schools set their own rules. Concerns persist about false flags: Vanderbilt disabled Turnitin’s AI detector after estimating its 1% false-positive rate could wrongly flag about 750 of 75,000 papers processed in 2022. In manufacturing, Infor and Anthropic introduced Infor Industry AI for Claude, allowing select Infor Velocity Suite customers to query operational data and request actions such as purchase orders or production-plan changes. Users must approve actions, and existing permissions apply; the connector uses the open Model Context Protocol and records actions in Infor. Customers need a Claude license and administrator activation. Aetina, meanwhile, offers compact, fanless Mini Series Edge AI Systems—including AIE-CO23/33-S1, AIE-CN33/43-A1, AIB-MO23/33-S1 and AIB-MN33/43-S1—for local visual inspection, anomaly detection, quality control, security monitoring and automated decisions.
Research on reasoning systems and reports on cybersecurity point to different opportunities and risks in AI. Researchers say their learned energy model, CRAM, solved 99.996% of Sudoku-Extreme test puzzles after 42 milliseconds of training on one GPU and all Maze-Unique puzzles after 132 milliseconds. They report at least 1,000-times-faster results than looped models on those tasks, and matching baseline Sudoku-Extreme accuracy with at least 282 times fewer inference FLOPs; they caution that simple puzzles may not show how looped reasoning scales. In security, researchers say attackers are using AntiFox, an AI-powered penetration-testing framework hosted on Alibaba Cloud, to target web applications and AWS environments with stolen credentials. Arctic Wolf described activity involving valid IAM credentials, an assumed EC2 role, access to a production secret, an ASPX web shell and S3 files. Check Point found more than 2,500 systems hosting ARTEX, whose autonomous agents can perform many operations with little human involvement. AWS said its services were not compromised and advised customers to limit permissions and enable Amazon GuardDuty.
Other coverage offers limited or broader market context. One article describes AI-powered decentralized exchanges as platforms using AI to improve transaction information, user experience and trading features, while saying Bitcoin’s longer-term direction may depend on institutional activity, ETF flows, adoption, regulation and economic conditions. A DW News video dated November 13, 2026, is titled “How Jensen Huang’s downfall became an AI security crisis,” but the article provides no details supporting that framing. On the same date, Tusk Venture Partners’ Bradley Tusk told Yahoo Finance that AI investors risk prioritizing prominent founders—including Huang, Elon Musk, Sam Altman and Dario Amodei—and future narratives over revenue, profits and other fundamentals. He questioned SpaceX’s valuation and urged investors to ask leaders to justify investments with facts rather than hype.
Key Takeaways
- Colette Kress is Nvidia’s interim CEO after Jensen Huang resigned; the board is considering internal and external candidates for a permanent successor.
- Nvidia shares dropped as much as 14% after the announcement, then closed 4.3% lower, while AMD and Broadcom each gained more than 5%.
- Nvidia’s reported commitments include $99 billion in equity investments, $25 billion in future investment commitments and $360 billion in AI cloud-capacity purchases; its OpenAI-related credit support can reach $105 billion.
- Reuters found more than $3 billion in Chinese AI projects specify restricted Nvidia chips, but project plans do not confirm delivery or establish export-law violations.
- A survey found 95% of 1,054 UK undergraduates use AI and 94% use it for assessed work; direct pasting of AI text rose from 3% in 2024 to 12% in 2026.
- Infor and Anthropic made Infor Industry AI for Claude available to select Infor Velocity Suite customers, with user approval, existing permissions and action logging built in.
- Aetina’s fanless Mini Series systems run AI locally for industrial inspection, anomaly detection, quality control and monitoring.
- Researchers report CRAM solved 99.996% of Sudoku-Extreme puzzles after 42 milliseconds of training and all Maze-Unique puzzles after 132 milliseconds.
- Researchers linked attacks using AntiFox and stolen credentials to AWS environments; AWS said its services were not compromised and recommended least-privilege access and Amazon GuardDuty.
- Bradley Tusk urged AI investors to assess fundamentals such as revenue and profits rather than relying on founder profiles or company narratives; the Bitcoin article identifies ETF flows and institutional activity as factors to watch.
Nvidia Faces Economic Questions After Jensen Huang Steps Down
Jensen Huang resigned as Nvidia’s CEO after founding the company 33 years ago, and Colette Kress became interim CEO. Analysts say Nvidia’s role supplying AI chips to companies such as Amazon, Anthropic, Meta, OpenAI and Oracle means the leadership change could affect the wider economy. Nvidia paid $10 billion in federal and state income taxes in fiscal 2025, and one analyst warned that weaker long-term investment could slow AI infrastructure growth. Others said demand remains strong and Nvidia is still about two to three years ahead of its nearest GPU competitor. The board will consider internal and external candidates for the permanent CEO role, and Nvidia is due to report third-quarter earnings on Tuesday.
Nvidia Shares Slide as Investors Question Customer Financing
After Jensen Huang resigned immediately as Nvidia’s CEO, president and director, the company’s shares fell as much as 14% before closing down 4.3%. Investors are weighing whether interim CEO Colette Kress will continue Nvidia’s investments in customers and purchase guarantees, which have helped AI companies expand data centers and also supported demand for Nvidia chips. A contract with CoreWeave, for example, allows Nvidia to buy up to $6.3 billion in computing capacity that CoreWeave cannot sell, subject to service conditions. AI cloud and data center power stocks fell, while Nvidia rivals AMD and Broadcom each gained more than 5%. South Korean venture capital firms said they would keep investing in AI but would watch Nvidia’s next moves closely.
Nvidia Interim CEO Colette Kress May Reconsider AI Funding
After Jensen Huang resigned, Nvidia CFO Colette Kress took over as interim CEO, raising questions about the company’s investments in AI firms that buy its GPUs. Nvidia had equity investments valued at $99 billion at the end of July, with another $25 billion in future investment commitments and $360 billion in AI cloud capacity purchase commitments. It also agreed to provide up to $105 billion in credit support for a data center business intended for OpenAI and signed a $6.3 billion cloud capacity deal with CoreWeave. Kress is reported to have opposed a proposed $250 billion OpenAI credit support plan, and stricter rules on future deals could ease concerns about circular financing but slow AI infrastructure investment. Possible CEO candidates mentioned include Ian Buck, René Haas, Tim Cook and Lisa Su, while reports said President Donald Trump urged Huang not to resign.
Chinese AI Projects Worth Billions Still Plan to Use Restricted Nvidia Chips
A Reuters report found that Chinese AI computing projects worth more than $3 billion rely on Nvidia processors restricted by the United States. Plans include a 6.2 billion yuan center in Fujian province that names H100 chips, a 3 billion yuan project in Shaanxi specifying H100 and H200 chips, and projects in Jilin and Inner Mongolia that also plan to use Nvidia processors. Some facilities have moved forward: a computing cluster with Nvidia H800 chips began operating in Wuxi in March 2025. The projects show continued reliance on Nvidia technology despite US export controls, which could expose developers to supply delays and higher costs. The report did not establish that processors for all planned projects had been delivered.
China Plans $3.2 Billion in AI Projects Around Restricted Nvidia Chips
A Reuters investigation found that more than $3.2 billion in Chinese AI projects specify Nvidia chips covered by US export controls. The figure represents the total value of projects, not confirmed chip purchases or shipments. Plans include a Fujian center valued at 6.2 billion yuan, a 3 billion yuan facility in Shaanxi, a 2.26 billion yuan project in Jilin and a proposed Inner Mongolia development listing 625 Nvidia H200 servers. Wuxi authorities reported that an H800 cluster began operating in March 2025, but planned projects do not prove that all chips were obtained or that export laws were broken. Nvidia’s chips and software remain widely used, while US restrictions that began in 2022 have led Chinese developers to weigh familiar Nvidia systems against alternative processors that may require software changes and extra engineering.
AI tools are reshaping student work and university rules
Trinity College Dublin uses Turnitin through Blackboard to check essays for similarity and AI use before lecturers read them. A survey of 1,054 UK undergraduates found that 95 per cent use AI, and 94 per cent use it for assessed work. The share who paste AI generated text directly into their work rose from 3 per cent in 2024 to 12 per cent in 2026. Trinity allows each School to set its own AI rules, while concerns remain about AI detectors wrongly flagging students. Vanderbilt turned off Turnitin’s detector after estimating that its 1 per cent false positive rate could wrongly flag about 750 of the 75,000 papers it processed in 2022.
Aetina brings real time AI to industrial edge systems
Aetina’s Mini Series Edge AI Systems let engineers and manufacturers run AI directly on devices at the edge instead of sending data to the cloud. The systems include the AIE-CO23/33-S1, AIE-CN33/43-A1, AIB-MO23/33-S1 and AIB-MN33/43-S1 models. They use connected cameras and sensors for real time visual inspection, anomaly detection, quality control, security monitoring and automated decisions. Their compact, fanless design suits production lines, factory equipment and smart inspection stations. Running AI locally can reduce delays, improve response times and increase efficiency.
CRAM solves Sudoku and maze puzzles up to 1,000 times faster
Researchers developed CRAM, a learned version of the energy model created by Hopfield and Tank in the 1980s. After 42 milliseconds of training on one GPU, CRAM solved 99.996 per cent of Sudoku-Extreme test puzzles; after 132 milliseconds, it solved 100 per cent of Maze-Unique puzzles. The researchers say those results came at least 1,000 times faster than looped models on the same tasks. CRAM also matched baseline looped models’ highest Sudoku-Extreme accuracy while using at least 282 times fewer inference FLOPs. The work presents energy based models as a promising way to train reasoning systems, while warning that simple puzzles may not test how well looped reasoning methods can scale.
Infor and Anthropic connect Claude to manufacturing operations
Infor and Anthropic introduced Infor Industry AI for Claude, which brings Infor’s manufacturing knowledge and controls into Claude conversations and Infor systems. Production planners, buyers and customer service representatives can ask questions in everyday language, review Infor data and request actions such as creating purchase orders or changing production plans. Users must review and approve recommended actions, and existing Infor permissions determine what they can see and do. The connector uses the open Model Context Protocol and records actions in Infor. The service is now available to select Infor Velocity Suite customers, who need their own Claude license and an administrator to activate the connector.
AI agents automate cloud attacks using stolen credentials
Researchers say attackers are using AntiFox, an AI powered penetration testing framework hosted on Alibaba Cloud, to attack web applications and AWS environments, steal credentials and seek cloud secrets. Arctic Wolf reported that attackers used valid IAM credentials and, in some cases, an assumed EC2 role to explore AWS services and access a production secret. They also described an attack that used an ASPX web shell to read server configuration and database connection details, and another that accessed S3 files. Check Point found more than 2,500 systems hosting ARTEX, whose autonomous agents can carry out large numbers of operations with little human involvement. AWS said its services were not compromised and urged customers to follow security practices, including limiting permissions and enabling Amazon GuardDuty.
AI crypto platforms draw attention as Bitcoin ETFs see outflows
Bitcoin ETF outflows can offer clues about institutional demand and wider market participation. The article describes AI-powered decentralised exchanges as blockchain trading platforms that use AI to improve transaction information, user experience and trading features. It says Bitcoin’s longer-term direction may depend on institutional activity, ETF flows, crypto adoption, regulation and broader economic conditions.
Jensen Huang’s downfall raises an AI security crisis
A DW News video published on November 13, 2026, is titled “How Jensen Huang's downfall became an AI security crisis.” The provided article page includes the video title and date but does not give details about the reported downfall or security crisis.
Bradley Tusk warns AI investors to focus on company fundamentals
Tusk Venture Partners co-founder and Managing Partner Bradley Tusk discussed AI bubble risks with Yahoo Finance Executive Editor Brian Sozzi on November 13, 2026. Tusk said investors may focus on celebrity founders such as Jensen Huang, Elon Musk, Sam Altman and Dario Amodei, and on what companies might become, rather than profits, revenue or other fundamentals. He questioned SpaceX’s valuation, saying he saw nothing to justify it. Tusk advised investors to ask company leaders why their businesses are sound investments based on facts, not narratives, and said leaders who rely on hype may avoid giving a clear answer.
Sources
- How Jensen Huang’s resignation from Nvidia is threatening AI-powered economy
- Nvidia Falls 4% After Jensen Huang's Exit; Fate of AI Investment 'Financing Support for Customers' Is the Wild Card
- Nvidia CEO transition — Jensen Huang steps down, sparking AI investment concerns
- Nvidia processors power Chinese AI data centre projects worth $3 billion despite US export controls: Report
- China’s $3.2 Billion AI Pipeline Still Calls for Nvidia Chips Washington Restricts
- AI Is Training To Be A Better Student Than You Are
- AI Tools for Engineers: Aetina Mini Series Edge AI Systems
- Classical Energies Learn to Reason 1,000x Faster than Looped Models
- Infor, Anthropic Bring Claude Into Manufacturing ERP
- Attackers are Using Agentic AI Offensive Tools to Automate Intrusions
- Could AI-Powered Crypto Trading Platforms Gain Attention as Bitcoin (BTC) Faces ETF Outflows?
- How Jensen Huang's downfall became an AI security crisis
- The biggest AI bubble risk: investing in celebrity founders over fundamentals
Comments
Please log in to post a comment.