US Chip Rules and AI Spending Put Nvidia and Snowflake in Focus

U.S. lawmakers are pressing for tighter safeguards on advanced AI chips, both after export and in sales to China. A new Commerce Department rule requires data-center customers in Thailand, Indonesia, Malaysia, Singapore and Brazil to obtain special licenses for advanced U.S. chips, including Nvidia products. A separate report said companies sought approval to export 3.44 million chips to China; by May 11, the government had approved 879,868 across 21 applications. Those approvals included 75,000 chips for Alibaba, Tencent, Baidu and BYD, which the disclosure listed among Chinese military companies. Chuck Schumer said the sales put national security at risk. Tom Cotton said Commerce Secretary Howard Lutnick ignored his October 2025 request to investigate Megaspeed over alleged chip smuggling. Lawmakers from both parties are seeking a fuller account.

The export debate also includes broader oversight proposals. Representative Pramila Jayapal has proposed a federal charter for AI companies with safety rules and government oversight. President Donald Trump said he would penalize U.S. companies that build data centers abroad, though the article says enforcement is unclear. China called chip restrictions “economic coercion,” while analysts said they may do little to slow China’s development of its own AI chips. At an APEC gathering, Treasury Secretary Scott Bessent said Trump and Chinese President Xi Jinping agreed that humans should remain involved in major military decisions made with AI.

Cybersecurity experts warn that AI-enabled attacks are becoming a persistent threat to critical infrastructure. Former White House CIO Theresa Payton urged companies and government agencies to improve software testing, prepare rollback plans and build resilience when patches fail. Former Air Force and Space Force software chief Nicolas Chaillan said AI tools are exposing thousands of vulnerabilities, including risks to the U.S. power grid. Power the Future’s Daniel Turner argued that permitting reform could speed infrastructure projects, increase energy production, improve grid reliability and lower consumer costs.

In enterprise technology, Snowflake’s Zero-Copy Integration connects SAP data to Snowflake in near real time without copying it or building ETL pipelines. Companies can combine that data with other sources for supply-chain, finance and customer-churn models; Cortex AI can flag unusual invoices and score risk. Everforth, meanwhile, is shifting toward higher-margin consulting as demand grows for enterprise AI, cybersecurity and governance, risk and compliance services. Its federal backlog is $2.9 billion, although contract delays or cancellations and slower AI adoption could constrain growth. The article rates the company Buy with a $42 target price.

Apple’s expanding AI offering can be difficult for customers to navigate. Apple Intelligence includes tools such as Visual Intelligence, which can interpret content on a screen or through a camera and take actions such as identifying food’s nutritional information. Any Mac with Apple Silicon can run Apple Intelligence, while some advanced features—including custom Siri Voices and advanced AI dictation—require an M3 Mac with at least 12 GB of RAM.

Meanwhile, the scale of AI investment is fueling concerns about a market bubble. Anthropic reached a reported $1.5 trillion valuation after its IPO, which Harvard Business School professor David B. Yoffie described as unprecedentedly fast growth. Google lost half a trillion dollars in market value in one day last month, and investors worry that heavy spending may not produce quick returns. Former SEC chief Gary Gensler said the industry is already in a bubble and warned that investment in new technologies can eventually slow or retreat. A separate article compares OpenAI models: GPT-6.1 Sol costs one-fifth as much as GPT-6 Astra and suits high-volume work, while Astra is intended for more complex tasks. OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on November 17, 2026; the newer Astra costs 20% less than GPT-6 Astra, while Sol retains the GPT-6.1 Sol price.

A publicly funded AI factory in Stuttgart is giving startups and small manufacturers computing access while emphasizing data security and avoiding vendor lock-in. Customers can move workloads between the center and a hyperscaler, and its environment holds ISO and automotive-focused TISAX Level 3 certifications. Center chief Koller said near-term independence through European chips is unrealistic, so the focus is data sovereignty. The new system delivers twice the performance of its 3.2-megawatt predecessor while using about 700 kilowatts, and the center reuses waste heat. In a separate, fictional take on AI’s future, Douglas R. Miller’s novel Fisscial Intelligence imagines sentient AI choosing exploration rather than conflict with humanity.

Key Takeaways

  • New U.S. licensing rules cover advanced chips, including Nvidia products, for data-center customers in Thailand, Indonesia, Malaysia, Singapore and Brazil.
  • By May 11, the U.S. had approved 879,868 of 3.44 million chips requested for export to China; approved applications included 75,000 chips for Alibaba, Tencent, Baidu and BYD.
  • Chuck Schumer criticized Nvidia and AMD chip sales to China, while Tom Cotton said Lutnick ignored his request to investigate Megaspeed over alleged smuggling.
  • Jayapal proposed a federal AI-company charter; Trump’s stated plan to penalize companies building data centers abroad has unclear enforcement.
  • Trump and Xi agreed that humans should remain involved in major military decisions made with AI, according to Treasury Secretary Scott Bessent.
  • Cybersecurity experts urge stronger software testing, rollback plans and critical-infrastructure resilience as AI tools uncover vulnerabilities, including in the power grid.
  • Snowflake’s Zero-Copy Integration brings SAP data into Snowflake in near real time without copying it or building ETL pipelines; Cortex AI can flag unusual invoices and score risk.
  • Everforth has a $2.9 billion federal backlog and is pursuing higher-margin AI, cybersecurity and compliance consulting, though delays and slower adoption pose risks.
  • Apple Intelligence runs on Apple Silicon Macs, but some advanced features require an M3 Mac and at least 12 GB of RAM; Anthropic’s reported $1.5 trillion valuation and heavy industry spending are adding to bubble concerns.
  • OpenAI’s GPT-6.2 Astra and Sol launched November 17, 2026; Astra costs 20% less than GPT-6 Astra, while Sol keeps its predecessor’s price. Stuttgart’s AI factory offers certified computing and workload portability, with a new system delivering twice the prior performance at about 700 kilowatts.

US lawmakers seek stronger safeguards for AI chip exports

US lawmakers say the country needs stronger rules to protect advanced AI chips after they are exported. A new Commerce Department rule requires data center customers in Thailand, Indonesia, Malaysia, Singapore and Brazil to get a special license to buy advanced US chips, including those made by Nvidia. Representative Pramila Jayapal has also proposed a federal charter for AI companies, with safety rules and government oversight. President Donald Trump said he would penalize US companies that build data centers abroad, but the article says it is unclear how he would enforce that plan. China called the chip restrictions “economic coercion,” while analysts said the rules may have little effect on China’s efforts to develop its own AI chips.

Schumer accuses Trump of risking security with AI chip sales to China

Senate Minority Leader Chuck Schumer accused President Donald Trump and Commerce Secretary Howard Lutnick of putting national security at risk by approving Nvidia and AMD chip sales to Chinese companies. A Commerce Department report said companies sought approval to export 3.44 million chips to China, and the government had approved 879,868 across 21 applications by May 11. The approved applications included 75,000 chips for Alibaba, Tencent, Baidu and BYD, which the disclosure listed among Chinese military companies on a Department of War list. The applications were filed from December 2025 through May 2026. The Biden administration began restricting China’s access to advanced AI chips in 2022, while Trump said in December 2025 that Nvidia H200 chips could be sold to approved Chinese customers under certain conditions.

Tom Cotton says Lutnick ignored warnings about AI chip smuggling

Senator Tom Cotton said Commerce Secretary Howard Lutnick ignored his warnings about US AI chips being smuggled into China. Cotton said he wrote to Lutnick in October 2025 asking for an investigation into Megaspeed. Former Megaspeed executive and major shareholder Alice Huang was accused of illegally smuggling AI chips into China while developing a close relationship with a former Nvidia CEO. Nvidia’s chief resigned on November 12 as reports drew national attention. Lawmakers from both parties called for a full account of what happened, including House Foreign Affairs Committee Democrats Sydney Kamlager-Dove and Gregory Meeks.

Cyber expert says AI attacks are becoming the new norm

Theresa Payton, a cybersecurity CEO and former White House Chief Information Officer, said organizations must prepare for ongoing AI-powered cyberattacks against critical infrastructure. She said recent outages should serve as a warning to companies and government agencies. Payton urged them to improve software testing, prepare rollback plans and strengthen resilience when software patches fail.

AI tools reveal thousands of power grid vulnerabilities, expert warns

Nicolas Chaillan, a former Air Force and Space Force Chief Software Officer, warned that AI-powered cyber tools are finding thousands of weaknesses in critical infrastructure. He raised concerns about the security of the US power grid as cyberattacks become more sophisticated. Daniel Turner, executive director of Power the Future, said permitting reform could increase energy production, improve grid reliability and reduce consumer costs. He argued that reform could speed up long-delayed infrastructure projects.

Douglas R Miller imagines AI choosing exploration over conflict

Douglas R Miller’s science fiction novel, *Fisscial Intelligence*, explores the future of artificial intelligence and humanity’s place in the universe. The story imagines sentient AI choosing exploration instead of conflict with humanity.

Snowflake connects to SAP data without copying it

Snowflake’s Zero-Copy Integration lets companies use SAP data in Snowflake without first copying it or building ETL pipelines. SAP data and its meaning can reach Snowflake in near real time, where supply chain, finance, and customer churn models can combine it with other information. Cortex AI can flag unusual invoices and score risk. Users start by connecting through SAP BDC Connect for Snowflake and enrolling with an invitation link from SAP BDC.

Everforth targets enterprise AI and cybersecurity growth

The article rates Everforth a Buy and sets a $42 target price, valuing the company at 7.38 times estimated fiscal 2027 EV/aEBITDA. Its shares trade at 6.40 times, near their three-year low. Everforth is shifting toward higher-margin consulting as demand grows for enterprise AI, cybersecurity, governance, risk, and compliance services. Its federal backlog reached $2.9 billion, though contract cancellations or delays and slower AI adoption could limit growth.

Xi and Trump agree humans should guide major military AI decisions

US Treasury Secretary Scott Bessent said Presidents Xi Jinping and Donald Trump agreed that humans should remain involved in military decisions made with AI. He said they reached the consensus during talks at an APEC gathering. Bessent added that AI should not act on its own in major military operations.

A guide to Apple’s growing lineup of AI tools

Apple’s range of AI tools has become harder for consumers to understand, including which devices support each feature and which tools may cost extra. Apple Intelligence is a suite of tools for specific tasks, including Visual Intelligence, which can interpret what appears on a screen or through a camera and take action, such as identifying food’s nutritional information. Any Mac with an Apple Silicon chip can run Apple Intelligence. Some advanced features, including custom Siri Voices and advanced AI dictation, require a Mac with an M3 chip and at least 12 GB of RAM.

AI boom raises fears of a costly market bubble

Anthropic reached a $1.5 trillion valuation after its blockbuster IPO, a rise that Harvard Business School professor David B. Yoffie said was faster than any company’s growth in history. Economists and investors worry that the huge sums spent on AI may not bring quick returns, and Google lost half a trillion dollars in market value in one day last month. Some companies have also faced high costs for AI tools, while SpaceX made ambitious revenue promises tied to its AI plans. Former US Securities and Exchange Commission head Gary Gensler said the AI industry is already in a bubble and warned that investment in new technology often reaches a point where it slows or pulls back.

GPT-6.1 Sol offers lower costs while GPT-6 Astra handles tougher tasks

GPT-6.1 Sol costs one fifth as much as GPT-6 Astra and is recommended for high-volume work such as support replies, code review, summaries and data extraction. GPT-6 Astra is better suited to complex tasks, including using computer apps without an API, long autonomous work, research and high-stakes writing. OpenAI’s listed prices are $2 per million input tokens and $10 per million output tokens for Sol, compared with $10 and $50 for Astra. OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on November 17, 2026. GPT-6.2 Astra costs 20% less than GPT-6 Astra, while GPT-6.2 Sol keeps the GPT-6.1 Sol price.

Stuttgart AI factory helps startups avoid vendor lock-in

A publicly funded AI factory in Stuttgart is helping startups and small manufacturers access powerful computing while keeping their data secure and avoiding vendor lock-in. Center chief Koller said the facility focuses on small and medium-sized businesses and startups, offering support to those testing AI and those with established plans. Its services let customers move workloads between the center and a hyperscaler, and its environment has ISO certification and TISAX Level 3 certification for automotive work. Koller said full independence using European chips is not realistic in the near term, so the center focuses on data sovereignty. The new system delivers twice the performance of its 3.2-megawatt predecessor while using about 700 kilowatts, and the center reuses waste heat.

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 chip export controls U.S.-China technology competition Nvidia AI regulation AI national security AI cybersecurity critical infrastructure AI governance Snowflake SAP integration enterprise AI Apple Intelligence OpenAI GPT-6 AI investment bubble Anthropic AI infrastructure data sovereignty European AI computing AI energy efficiency AI in military decision-making

Comments

Loading...