Nvidia secures H200 chip sales as Anthropic aids US military

Key Takeaways

• Nvidia has received Chinese approval and purchase orders for its H200 AI chips, restarting production after a year-long halt.
• Nvidia CEO Jensen Huang confirmed the H200 sales, which are capped at 50% of U.S. domestic volume for China.
• Nvidia is developing a custom Groq AI chip for the Chinese market, expected by May, for inference tasks.
• GMI Cloud is investing $12 billion to build a 1GW sovereign AI infrastructure in Kagoshima, Japan, for national security and economic competitiveness.
• The U.S. military is using AI, including tools from Anthropic, to reduce targeting analysis time from days to under four hours.
• Amazon employees report AI tools like Kiro sometimes generate flawed code, negatively impacting productivity.
• AI researcher Andrej Karpathy's 'autoresearch' system uses AI agents to run 700 experiments in two days, accelerating AI model development.
• Menlo Security launched a browser security platform to secure enterprise computing from AI agent risks, enforcing instruction-data separation.
• The FAA implemented new helicopter flight restrictions near airports based on AI analysis of thousands of aviation close calls.
• AI is creating challenges like students using it for assignments and the spread of deepfakes, as seen with Israeli PM Benjamin Netanyahu.

Nvidia prepares Groq chips for China market

Nvidia is developing a version of its Groq chips for the Chinese market, with availability expected in May. These chips are designed for AI inference tasks, where AI systems perform actions for users. Nvidia plans to combine these Groq chips with its forthcoming Vera Rubin chips, which cannot be sold in China. This move comes as Nvidia restarts production of its H200 chips after receiving export licenses from the U.S. government. The Groq chips are not downgraded versions but can be adapted to work with other systems.

China approves Nvidia H200 AI chip sales

Chinese authorities have approved multiple Chinese companies to purchase Nvidia's H200 AI chips. Nvidia CEO Jensen Huang confirmed that the company has received licenses and purchase orders from many Chinese customers. This approval follows months of waiting for licenses from both the U.S. and China. Nvidia has also been restarting production of the H200 chips, which had been halted due to regulatory hurdles. The U.S. had previously granted a license in February for small amounts of H200 products to specific China-based customers.

China approves Nvidia H200 AI chip sales

Chinese authorities have given the green light for Nvidia's H200 AI chip sales in China. Nvidia CEO Jensen Huang confirmed receiving licenses and purchase orders from numerous Chinese companies. The company is also restarting H200 chip manufacturing after a pause due to regulatory issues in both the U.S. and China. While the U.S. had previously approved limited exports, Beijing's approval is now secured. This development allows major tech firms like ByteDance, Tencent, and Alibaba to import the chips.

Nvidia gets China OK for H200 GPUs, restarts production

Nvidia CEO Jensen Huang confirmed that the company has received export licenses for multiple Chinese customers and has purchase orders for its H200 GPUs. Manufacturing of the H200 chips has also restarted, marking the first time its China supply chain has been active in over a year due to export restrictions. The U.S. government has granted these licenses, and Chinese authorities have also approved sales. This approval allows for H200 sales with a 50% volume cap relative to U.S. domestic sales.

Nvidia gets China approval for H200 chip sales

Nvidia has received approval from Beijing to sell its H200 artificial intelligence chips in China, and is reportedly developing a custom version of the Groq AI chip for the market. This approval allows Nvidia to resume shipments of its H200 series chips, which were a significant part of its previous revenue. While the U.S. authorized exports, China's internal hesitation was the main obstacle. Nvidia plans to use Groq chips for inference tasks and will pair them with its Vera Rubin chips, which are restricted in China.

China approves Nvidia H200 AI chip sales

Chinese authorities have approved Nvidia's H200 AI chip sales in the country, according to a source. Nvidia has been seeking this approval to maintain its market lead in China amidst U.S. export restrictions. The H200 chip is an upgrade to the A100, offering better performance and memory for AI tasks. The approval comes as Nvidia faces competition from domestic Chinese chipmakers and other international companies. The exact volume and value of H200 chip sales in China remain unclear.

AI changes work for professors and Amazon employees

Artificial intelligence is significantly changing the daily work of white-collar professionals. University professors are finding students using AI to complete assignments, which hinders the development of critical thinking skills. Meanwhile, Amazon employees report that the company is pushing AI integration in a way that sometimes hurts productivity, with tools like Kiro generating flawed code that requires extensive correction. These changes highlight the disruptive and often uncomfortable nature of rapid technological advancement in the workplace.

Military uses AI for faster targeting and analysis

The U.S. military is increasingly using artificial intelligence to speed up targeting and battlefield analysis. AI systems can process vast amounts of data from sensors and battlefields, allowing for quicker identification of potential targets and optimal weapon selection. This technology compresses the time for target processing from days to potentially under four hours, while still keeping a human in the loop. AI is also used for administrative tasks and to help war-game scenarios to minimize casualties. The Pentagon is evaluating AI tools from various companies, including Anthropic.

Filmmaker Jia Zhangke explores AI in movies

Chinese filmmaker Jia Zhangke is experimenting with AI tools in filmmaking, having produced two short films using the technology. He believes in approaching new technologies with curiosity to understand their potential benefits for film expression and production. Jia sees AI as a new medium that can offer new perspectives and possibilities. He also discussed how social media inspired the structure of his film 'A Touch of Sin,' highlighting how new media influences artistic creation.

Andrej Karpathy's AI agents run 700 experiments in 2 days

AI researcher Andrej Karpathy has developed a system called 'autoresearch' that uses autonomous AI agents to speed up AI research. In a recent experiment, these agents ran 700 experiments in two days, adjusting training code and initial settings for AI models. This technique led to significant performance gains in training AI models. Karpathy believes this approach will become standard in AI labs, potentially accelerating progress. The system involves an agent modifying code, optimizing a metric, and operating within a time limit.

Menlo Security adds AI agent security to browser platform

Menlo Security has launched a new browser security platform to address the risks posed by AI agents in enterprise computing. The platform provides a unified control plane for both human and non-human actors, applying governance and threat prevention at machine speed. It enforces instruction-data separation to prevent agents from misinterpreting malicious data as commands. The system also offers visibility into agent activity and enforces least-privileged access to prevent data breaches. This aims to provide consistent protection against sophisticated threats in the agentic economy.

Braznex embeds AI in trading for better execution

Braznex is advancing AI integration in global multi-asset trading by embedding intelligence directly into its trading systems. Unlike platforms that use AI for analytics, Braznex's AI-native framework works alongside its Order Management System and Smart Order Router. It uses machine learning to optimize routing decisions, model execution slippage, and recalculate margin requirements in real-time. This approach aims to close the intelligence gap between institutional and retail investors by providing advanced execution quality and risk intelligence.

GMI Cloud launches $12 billion AI infrastructure in Japan

GMI Cloud is launching a $12 billion initiative to build a 1GW sovereign AI infrastructure in Kagoshima, Japan. This project aims to ensure Japan has control over its AI computing power, which is considered crucial for national security and economic competitiveness. The facility will be an AI Factory designed for Physical AI, supporting robotics, autonomous vehicles, and industrial infrastructure. GMI Cloud is partnering with Wistron and VAST Data to build a green and sustainable AI infrastructure.

Netanyahu uses video to counter AI fakes

Israeli Prime Minister Benjamin Netanyahu posted a video to prove he is alive, countering claims that a previous address was an AI-generated fake. Some online accounts, linked to Iran, falsely identified a sixth finger on his hand in the video as evidence of AI manipulation. This incident highlights the 'liar's dividend,' where real videos can be dismissed as fake, sowing distrust. The phenomenon is becoming a significant challenge during conflicts, making it difficult to discern truth from AI-generated deception.

Sealevel Systems designs edge AI hardware for aerospace

Sealevel Systems is engineering AI-ready edge hardware for aerospace and defense systems. The company focuses on modular architectures, stable computing cores, and thoughtful design to meet demands for performance, power, and thermal management in harsh environments. Their approach uses modular designs with computer-on-modules to allow for system upgrades without full redesigns, supporting long program lifecycles. Sealevel's hardware is optimized for ingesting sensor data and can include AI acceleration through CPUs, GPUs, NPUs, or TPUs.

AI analysis leads FAA to restrict helicopter flights near airports

The Federal Aviation Administration (FAA) will now use radar to track helicopters flying near busy airports, a change prompted by AI analysis of aviation close calls. Previously, helicopter pilots were responsible for visual separation with other aircraft using a 'see and avoid' method. However, a fatal helicopter crash in January 2025 highlighted the risks, revealing that the FAA had data on thousands of close calls that went unanalyzed. AI tools were then used to identify risks nationwide, leading to new protocols to enhance safety.

Sources

• Exclusive: Nvidia preparing Groq chips that can be sold in Chinese market, sources say
• Chinese authorities approve Nvidia's H200 AI chip sales, source says
• Chinese authorities approve Nvidia's H200 AI chip sales, source says
• Jensen says Nvidia has received orders from Chinese customers for H200 GPUs, licenses from US gov't
• Nvidia receives China’s approval for H200 chip sales
• Chinese authorities approve Nvidia’s H200 AI chip sales, source says
• How AI is actually changing day-to-day work
• How the military is using AI in war
• Jia Zhangke Talks AI Tools In Filmmaking at 2026 Filmart
• ‘The Karpathy Loop’: Former OpenAI researcher’s autonomous agents ran 700 experiments in 2 days—and gave a glimpse of where AI is heading
• Menlo Security takes on AI agent risk with new browser security platform
• As Institutional AI Moves From Analytics to Execution, Braznex Advances Natively Embedded Intelligence for Global Multi-Asset Trading
• GMI Cloud Unveils $12 Billion, 1GW Sovereign AI Infrastructure Initiative in Japan
• Netanyahu posts ‘proof of life’ video as AI sows doubts about what’s real
• Engineering Edge AI Hardware for Aerospace & Defense
• AI probe of aviation close calls leads FAA to limit how helicopters can fly around busy airports