Google Tailors Eighth Generation TPUs to AI Agents

Key Takeaways

• AI-related goods accounted for as much as 80% of recent global export growth, according to one analysis; South Korean semiconductor exports jumped 259.4% year over year in October.
• AMRO projects ASEAN+3 growth of 4.1% in 2026 and 2027, while warning that power constraints, delayed productivity gains and fast market reactions could expose the region to an AI investment slowdown.
• Google’s TPU 8i Max doubles theoretical computing power and memory bandwidth within the existing rack footprint; Google says TPU Sync cuts memory-cache transfer delays by more than 90%.
• Google’s Agent Compute Engine pauses idle agent sessions and sets aside working memory, while TPU 8tx targets confidential inference and small-scale training.
• WashU students raised privacy concerns about possible access to university-affiliated AI chat logs, including chats through Claude for Education; Vice Provost Nathan Jacobs said access to Claude Edu chats in academic investigations was uncertain.
• At the University of Alaska, nearly 200 students and staff petitioned against AI-assisted guidelines, and regents delayed a vote until February after faculty objected in part to late disclosure of Claude’s role.
• UK and Norway plan deeper AI cooperation spanning compute, hardware, defence, safety testing and governance, coordinated by their respective government departments.
• Former FCC chair Tom Wheeler joined Known as an adviser; the startup is developing tools to identify autonomous AI agents and the organizations responsible for them.
• Rio is developing an AI-enabled WhatsApp health-care tool for more than six million residents after the proposal won the 2025–2026 Bloomberg Philanthropies Mayors Challenge.
• Articles also covered school-parent concerns about technology use, AI safety and investment discussions at the Lisbon Web Summit, and AI trip planning’s possible role in travel and hospitality.

AI Data Center Spending Drives Trade Growth and Asian Export Risks

U.S. investment in AI data centers has helped drive global trade growth, with Asian exporters supplying semiconductors, servers, and other electronics. Capital Economics expects global trade to grow at its strongest rate since the early 1990s, excluding rebounds after the COVID-19 pandemic and the 2008 to 2009 financial crisis. In October, South Korean exports rose 93% year over year, including a 259.4% jump in semiconductor exports, while China’s exports grew 35%. The article says up to 80% of recent global export growth came from AI-related goods, so a slowdown in data center spending could hurt exporters and growth across Asia. It also reports that the KOSPI has fallen 42% to 43% from its peak and warns that policymakers have few tools to prevent a global recession if the AI boom collapses.

Asia’s AI Boom Faces Three Timing Gaps

On November 12, 2026, an analysis of ASEAN+3 described how demand for semiconductors, advanced electronics, and digital infrastructure is supporting exports and investment. AMRO projects that ASEAN+3 economies will grow 4.1% in both 2026 and 2027. The article highlights gaps between AI investment and power infrastructure, between immediate investment gains and longer-term productivity improvements, and between fast-moving financial markets and slower economic changes. It says these gaps could leave data centers short of power, create unused infrastructure, or lead to sudden financial shifts if expectations change. A change in AI investment could affect exports, company earnings, share prices, and cross-border investment across the region.

WashU Students Debate Access to AI Chat Logs

WashU Vice Provost for Artificial Intelligence Nathan Jacobs said he believes the university may be able to view university-affiliated AI chat logs in student misconduct cases. He said AI chats fall under policies that allow access to university systems, and Claude for Education, launched on September 14, 2026, is operated by WashU. Jacobs was unsure whether the university could access Claude Edu chats during academic integrity investigations. Sophomores Johnny Jannotta, Sykes Carlsen, and Daniel Di Palma shared concerns about the breadth of access, the need for clear notice, and the effect on privacy in academic and personal life.

Lisbon Web Summit Highlights AI Safety and Investment

The Lisbon Web Summit in Lisbon, Portugal, brought together technology experts and industry leaders to discuss artificial intelligence. The event focused on AI safety and investment, with speakers and panel discussions about the technology’s future, risks, and possible consequences. It also offered networking opportunities and exhibitions of AI technologies. The article says the summit gave the industry a place to discuss responsible AI development and continued investment in research and development.

UK and Norway Plan Wider Cooperation on Artificial Intelligence

UK Artificial Intelligence Minister Kanishka Narayan and Norwegian Minister for Digitalisation and Public Governance Torgeir Micaelsen met on November 12, 2026, at the Atlantic Futures Forum. They announced plans to deepen AI cooperation, building on the countries’ 2024 strategic partnership and their shared goals for security, resilience, growth, and innovation. Potential work includes compute infrastructure, semiconductors and AI hardware, defence applications, AI safety testing, and international AI governance. The ministers also plan to exchange experience on research and innovation policy and on AI’s links with quantum technology, biotechnology, and robotics. The UK Department for Business, Innovation, Science and Trade and Norway’s Ministry of Digitalisation and Public Governance will coordinate the work.

Rio uses AI and WhatsApp to connect residents with health care

Rio de Janeiro is developing an AI enabled WhatsApp tool to help its more than six million residents get health care. The tool will identify people at risk of missing services, contact them, and take action to help. The idea won the 2025 to 2026 Bloomberg Philanthropies Mayors Challenge, which received proposals from 50 cities. Rio is now working with local health care providers to put the idea into practice.

Former FCC chair Tom Wheeler joins AI startup Known as adviser

Tom Wheeler, who led the Federal Communications Commission from 2013 to 2017, will advise AI startup Known. The company is developing technology to identify autonomous AI agents and determine which organizations are responsible for them. Wheeler said clear accountability will be important as these systems take on more responsibility, and that regulators, standards groups, and businesses can help build it. Known Interim CEO Allie Kline said Wheeler’s experience connecting government and industry will support the company’s growth. Other advisers include Vint Cerf, Vinh Ngueyn, Teresa Shea, Paul Sagan, and Edna Conway.

University of Alaska faculty challenge AI guidelines after delayed disclosure

The University of Alaska system’s chief information technology officer used Anthropic’s Claude to help draft its AI Engaged guidelines. Nearly 200 students and staff, mostly in Anchorage, signed a petition asking regents to adopt a short statement written by people instead. Hundreds of faculty members oppose the draft in part because the use of AI was not disclosed until last month. The chief information technology officer said humans made several rounds of edits and only about 20% of the original two page draft remained. Regents delayed a vote until February to allow more public input.

Google builds specialized TPUs for AI agents and secure computing

Google announced new eighth generation TPU systems, including the TPU 8i Max for long context agent work and the TPU 8tx for confidential inference and small scale training. The 8i Max combines two TPU 8i chips in one package, doubling theoretical computing power and memory bandwidth within the existing rack footprint. Google says its TPU Sync software can move large AI memory caches to other storage and cut transfer delays by more than 90%. The company also introduced Agent Compute Engine, a Google Cloud service that pauses idle agent sessions and sets aside their working memory. The TPU 8tx is aimed at customers such as regulated businesses and governments that need to handle sensitive data securely.

Boulder parents protest the growing use of AI and tech in schools

Parents in Boulder joined a national protest against what they see as too much technology in schools, including AI, Chromebooks, and tablets. They gathered outside Google’s campus and raised concerns about having little choice over devices sent home with students. Anna Segur said school records showed her son made 460 YouTube visits in one month and watched 39 hours of YouTube at school over six months. Parents also questioned whether strong peer reviewed research shows that education technology improves learning. They called for more time and evidence before schools expand its use.

Ajay Aluri examines AI’s role in travel and hospitality

Ajay Aluri contributed to Marketplace reporting on the growing use of AI options for trip planning. The report asks what role AI will play in travel and hospitality. The article is dated November 12, 2026.

Sources

• Global Trade's Reliance on AI Infrastructure Investment: If the 'Single Engine' Stalls, Asia Takes the Direct Hit
• Asia's AI Boom Has a Timing Problem
• WashU can see students’ AI chat logs in misconduct cases. What do students think?
• Lisbon Web Summit puts AI safety and investment in focus
• United Kingdom-Norway Joint Statement of intent on Artificial Intelligence
• Your city probably isn’t using AI like this one
• Former FCC Chair Tom Wheeler to Advise AI Startup
• UA faculty clash over postponed artificial intelligence guidelines
• Google reworks TPU silicon into specialized systems for agents and security
• Boulder parents join national protest against AI, tech in education
• What role will AI play in travel and hospitality?