Nvidia’s leadership change is prompting questions about how much capital will continue flowing into AI companies. The articles report that Jensen Huang left after 33 years and finance chief Colette Kress became interim CEO while the board searches for a permanent successor. AI researcher Gary Marcus says Nvidia has acted like a bank, investing in firms that then buy its chips; OpenAI, Anthropic and CoreWeave are among those that have relied on this funding model. Marcus warns that Kress could be more cautious, potentially tightening funding for AI developers that are losing money. He also points to possible federal and congressional investigations into Nvidia’s dealings involving China. After the leadership news, European and Asian shares edged lower, though losses were limited.
Separate investigations in Malaysia and Singapore are examining Nvidia hardware shipments. Malaysia is checking whether equipment imported by a Megaspeed subsidiary was later sent abroad. Prime Minister Anwar Ibrahim said there was no evidence it had left the country, and Megaspeed said it hired a U.S. law firm to audit its GPU inventory and export-control compliance. In Singapore, Chinese citizen Shi Bingzhe faces 41 charges tied to transactions involving Superphi and NJ Global Financial and 917 servers containing restricted Nvidia hardware. Authorities have not established that the equipment reached China. If convicted, Shi could face up to 10 years in prison and a fine of S$500,000 per charge.
Google has released Gemini 4 Hydrogen, a Google DeepMind model for coding and everyday agent tasks, available through the Gemini API but not for self-hosting. Google says it works 2.5 times faster than Gemini 3.8 Flash and scores 57% on Terminal-Bench 4.1. It has a 1,048,576-token context window, supports text, images, audio, video, PDFs, tables and charts, and offers four thinking levels, including the new xhigh option. API pricing is $1.50 per million input tokens and $7.50 per million output tokens. Elsewhere, Elev8 lets clients connect ChatGPT or Claude through the Model Context Protocol to read and explain account information. The connection is read-only: it cannot trade, transfer funds or change account settings, and clients can disconnect at any time.
AI’s expanding use brings security risks alongside new capabilities. It can help government and security teams analyze satellite imagery, detect unusual activity and respond to emergencies, but can also enable phishing, voice fakes, targeted scams and faster cyberattacks. The article recommends restricting AI access to sensitive systems, monitoring activity, keeping detailed records and requiring human approval for high-impact actions. It describes a November 4, 2026 incident in which hackers changed refrigeration settings at three French blood-service storage sites; alarm systems enabled technicians to intervene before blood products were harmed. These concerns will feature at GITEX GLOBAL 2026 in Dubai, December 7–11, where GITEX Cyber Valley will address autonomous-agent risks, connected systems, data sovereignty, AI security, cyber resilience and post-quantum readiness. Exhibitors include Acronis, Kaspersky, SentinelOne and Trellix, and the UAE Cybersecurity Council will participate.
Demand for AI infrastructure is also lifting ITCEN PNS. The company reported third-quarter consolidated revenue of 239.5 billion won, up 14.9% year over year, and net profit of 1.1 billion won, although it remained at an operating loss. GPU server and memory sales at subsidiary C Platform drove the increase. Standalone revenue rose 17.2% to 20.3 billion won, with net profit of 2.9 billion won, partly reflecting convertible-bond valuation gains. ITCEN PNS plans to expand AI security and post-quantum cryptography services, including vulnerability checks, consulting and migration to PQC systems. CEO Park Won-gyu says the company will focus on profitability and operational efficiency while growing revenue.
Other organizations are putting AI to work in professional services and research. Law firm Ropes & Gray will let lawyers count up to 100 hours of AI training toward billable targets, alongside investments in AI-powered legal tools. Bracco and Avicenna.AI received U.S. FDA clearance for UPTUNITY-Aneurysm, software that helps radiologists identify suspected saccular brain aneurysms larger than 3 mm on CT angiography and track known aneurysms over time. It supports, but does not replace, physician review; the companies are preparing a U.S. launch and Bracco plans to show it at RSNA in Chicago, November 29–December 3. Researchers’ ESGPred framework combines language models and graph learning to predict ESG scores from sustainability reports and company relationships. Tested on 5,266 reports from 2,134 public companies published between 2020 and 2023, it predicted overall and individual Environmental, Social and Governance ratings, with graph learning outperforming text-only methods.
Key Takeaways
- Jensen Huang left Nvidia after 33 years; interim CEO Colette Kress now leads while the board searches for a permanent successor.
- Gary Marcus warns a more cautious Nvidia investment strategy could affect funding for OpenAI, Anthropic and other AI developers that rely on capital while struggling to profit.
- European and Asian markets edged lower after the Nvidia leadership news, with reported losses limited.
- Malaysia is investigating possible onward shipments of Megaspeed-imported AI hardware; Singapore has filed 41 charges against Shi Bingzhe involving transactions linked to 917 servers with restricted Nvidia hardware. Neither investigation has established that the equipment reached China.
- Google Gemini 4 Hydrogen has a 1,048,576-token context window, costs $1.50 per million input tokens and $7.50 per million output tokens, and Google says it runs 2.5 times faster than Gemini 3.8 Flash.
- Elev8’s MCP connection lets clients use ChatGPT or Claude to inspect brokerage information, but it cannot place trades, move money or change accounts.
- GITEX GLOBAL 2026 runs December 7–11 in Dubai, with AI security, cyber resilience and post-quantum readiness on the agenda.
- ITCEN PNS’s third-quarter revenue rose 14.9% to 239.5 billion won, driven by GPU server and memory sales, though the company still reported an operating loss.
- Ropes & Gray will count up to 100 hours of AI training toward lawyers’ billable targets; Bracco and Avicenna.AI received FDA clearance for aneurysm-detection software that supports rather than replaces physician decisions.
- ESGPred’s testing across 5,266 sustainability reports found graph-based learning improved ESG-score prediction compared with text-only approaches.
Nvidia Leadership Change Raises Questions About AI Funding
Nvidia named finance chief Colette Kress interim CEO as its board searches for a permanent successor to Jensen Huang. AI researcher Gary Marcus said Huang had taken large risks by investing in companies across the AI industry, while Kress may be more cautious. Marcus warned that a funding pullback could leave AI companies short of money, since many large language model developers are losing money and need major investments. He said such a change could shake investor confidence and threaten the wider AI boom.
Jensen Huang’s Exit Could Change Nvidia’s Role in the AI Boom
Jensen Huang left Nvidia after 33 years, and finance chief Colette Kress became interim CEO. AI researcher Gary Marcus said Nvidia has acted like a bank by investing in AI companies that then use the money to buy Nvidia chips. OpenAI, Anthropic and CoreWeave have relied on this kind of funding, even as many AI businesses struggle to make a profit. Marcus said Kress may take a more cautious approach to large investments and will also face possible federal and congressional investigations into Nvidia’s dealings involving China.
Governments Face Growing Security Risks as AI Use Expands
Artificial intelligence can help governments and security teams analyze satellite images, spot unusual activity, protect networks and respond to emergencies. But AI can also help criminals create convincing phishing messages, fake voices and targeted scams, and speed up cyberattacks. The article says organizations in sectors such as defense, banking, energy, healthcare and transportation should limit AI access to sensitive systems and use monitoring, detailed records and human approval for high-impact actions. It also describes a November 4, 2026 incident in which hackers changed refrigeration settings at three French blood service storage sites, but alarm systems let technicians act before blood products were harmed.
GITEX GLOBAL 2026 to Focus on AI Security and Cyber Resilience
GITEX GLOBAL 2026 will take place December 7 to 11 in Dubai, UAE, with a cybersecurity showcase focused on AI security, cyber resilience and readiness for post-quantum technology. Its GITEX Cyber Valley program will examine risks from autonomous AI agents, connected systems and data sovereignty through meetings and briefings for cybersecurity leaders. Confirmed exhibitors include Acronis, Kaspersky, SentinelOne and Trellix, among others, and the UAE Cybersecurity Council will also take part. The event will discuss how organizations can protect AI systems, data and infrastructure as they adopt new technologies.
Malaysia and Singapore Investigate Nvidia Hardware Shipments
Malaysia is investigating whether AI hardware imported by a Megaspeed subsidiary was later sent abroad, while Singapore has filed charges linked to 917 servers containing restricted Nvidia hardware. Malaysian Prime Minister Anwar Ibrahim said there was no evidence that Megaspeed-related equipment had left the country, and Megaspeed said it hired a U.S. law firm to audit its GPU inventory and export-control compliance. In Singapore, Chinese citizen Shi Bingzhe faces 41 charges related to transactions by Superphi and NJ Global Financial between May 2024 and February 2025. The investigations have not established that the hardware reached China; Shi could face up to 10 years in prison and a fine of S$500,000 per charge if convicted.
Google launches Gemini 4 Hydrogen for faster coding and agent work
Google released Gemini 4 Hydrogen, a model from Google DeepMind designed for coding and everyday agent tasks. It has a 1,048,576-token window and runs through Google’s Gemini API, with no option to host it yourself. Google says Hydrogen completes tasks 2.5 times faster than Gemini 3.8 Flash and scores 57% on Terminal-Bench 4.1. API prices are $1.50 per million input tokens and $7.50 per million output tokens. Developers can set four thinking levels, including the new xhigh option, and the model supports input such as text, images, audio, video, PDFs, tables and charts.
World markets edge lower as investors weigh Nvidia leadership news
Shares fell modestly in Europe and Asia as investors considered how Jensen Huang’s resignation as Nvidia CEO might affect the global artificial intelligence industry. The losses remained limited, according to the article. Investors were assessing the possible impact of the Nvidia news on the AI sector. The article does not provide further details about the resignation or market moves.
Ropes & Gray lets lawyers count AI training toward billable goals
Ropes & Gray will allow lawyers to count up to 100 hours of AI training toward their billable targets. The firm said the policy supports its use of technology and innovation in legal work. It has also invested in AI-powered tools intended to make legal work more efficient and effective. The article says the policy is part of the firm’s effort to improve client service and support business growth.
AI infrastructure demand lifts ITCEN PNS revenue and restores net profit
ITCEN PNS reported third-quarter consolidated revenue of 239.5 billion won, up 14.9% from a year earlier, and net profit of 1.1 billion won, while it continued to post an operating loss. GPU server and memory sales at subsidiary C Platform drove the revenue increase as investment in AI infrastructure boosted demand. On a standalone basis, revenue rose 17.2% to 20.3 billion won, and the company reported a 2.9 billion won net profit, partly due to valuation gains on convertible bonds. ITCEN PNS plans to grow AI security and post-quantum cryptography businesses, including vulnerability checks, security consulting and migration to PQC systems. CEO Park Won-gyu said the company would focus on improving profits and operational efficiency while growing revenue.
Elev8 lets traders connect AI assistants to read account data
Global CFD broker Elev8 has introduced a feature that lets clients connect a preferred AI assistant to their brokerage account using the Model Context Protocol, or MCP. The read-only connection can show and explain account balances, trading conditions, instruments, trade history, profits and losses, open positions and pending orders. The assistant cannot place trades, move money or change an account. Elev8 provides setup guides for ChatGPT Web and Desktop and Claude Web and Desktop, and says setup takes a few minutes. Account passwords are not shared, clients can disconnect at any time, and the AI assistants remain independent third-party software.
Bracco and Avicenna.AI gain US clearance for brain aneurysm software
Bracco and Avicenna.AI announced US FDA clearance for UPTUNITY-Aneurysm, software that helps radiologists detect and monitor brain aneurysms on head CT angiography (CTA) images. It identifies suspected saccular aneurysms larger than 3mm and provides information about their location and size. For known aneurysms, it measures and compares scans over time. The software supports, but does not replace, a physician’s review and clinical decisions. The companies are preparing to introduce it commercially in the USA, and Bracco plans to showcase it at the RSNA annual meeting in Chicago from 29 November to 3 December.
ESGPred combines language models and graph learning to predict ESG scores
Researchers introduced ESGPred, a modular framework that predicts Environmental, Social and Governance (ESG) scores by combining domain-specific language modeling with graph learning. The system uses text from sustainability reports and relationships among the companies being rated. Researchers tested it on 5,266 reports from 2,134 publicly listed companies published between 2020 and 2023, using LSEG Data & Analytics ratings as the ground truth. ESGPred predicted overall ESG ratings and separate Environmental, Social and Governance scores. The results showed that graph-based learning improved performance compared with text-only methods.
Sources
- Nvidia Funded AI Boom Under Jensen Huang, but Gary Marcus Questions if New CEO Will Keep Writing the Checks: ‘Whole Thing Might Collapse’
- Jensen Huang is out. An AI expert explains what his exit could mean for the AI boom
- AI and National Security: Why Strategic Restraint Matters
- GITEX GLOBAL 2026 Puts AI Security, Cyber Resilience and Post-Quantum Readiness in Focus
- Malaysia and Singapore step up Nvidia AI chip probes as charges brought against Chinese national over 917 smuggled servers — firm at center of Jensen Huang controversy announces inventory audit to account for all restricted GPUs
- Google Releases Gemini 4 Hydrogen: 2.5x Faster Than Gemini 3.8 Flash
- World shares log modest losses as investors assess impact on AI sector from the Nvidia scandal
- BigLaw firm makes statement with new billable hours/AI training policy
- ITCEN PNS posts higher revenue as AI demand lifts sales; net profit returns
- Global broker Elev8 enables AI access to help clients see more of their trading data
- Bracco, Avicenna.AI announce US FDA clearance for new brain aneurysm detection and assessment software
- A Modular Framework for ESG Prediction through Domain-Specific Language Modeling and Graph Learning
Comments
Please log in to post a comment.