Amazon Offloads $8B in Nvidia Chips, California Bans AI Firing Decisions

Amazon is moving forward with a plan to transfer roughly $8 billion in Nvidia Grace Blackwell chips to a special-purpose vehicle, offering investors up to a 10 percent equity stake. The company would then lease the chips back while continuing to operate them across data centers in five US states, including Nevada and Virginia. Each accelerator costs between $60,000 and $70,000, meaning Amazon would offload somewhere between 114,000 and 133,000 chips. The move would help Amazon reduce costs on its balance sheet while maintaining full use of the hardware.

The chip deal fits into a much larger spending picture. Amazon expects to shell out $220 billion in capital expenditure this year, a steep jump from $131 billion in 2025, with most of that going toward its cloud unit AWS. Along the way, Amazon has also been raising AI chip rental prices by 15 percent. It is one of several big tech companies pouring enormous sums into AI infrastructure, a trend that investor Michael Burry has compared to the dot-com bubble. Burry warns that five major tech companies have hidden commitments totaling over $1.5 trillion that do not appear on their balance sheets, and that AI spending now represents 2.07 percent of US GDP.

Meanwhile, California is stepping in with new rules. Governor Gavin Newsom signed the "No Robo Bosses" law, taking effect July 1, 2027, which requires a human to verify and approve any AI-based decision to discipline or fire a worker. Employers must disclose when AI was used and provide data records on request. Violations carry a $500 penalty per infraction. The law aims to keep companies accountable when algorithms make workplace decisions that affect people's jobs.

On the security front, a new report identifies seven platforms to watch for securing AI agent identities and permissions. By April 2026, about 38 percent of organizations have more than 100 AI agents running, yet only 9.5 percent secure more than 81 percent of them. OpenAI adopted Ory for identity management, a system that serves 1.2 billion weekly active users. Global enrollment in AI security training courses has surged by 665 percent, with U.S. enrollment up 358 percent. Still, nearly 40 percent of businesses lack adequate AI security expertise.

Concerns about an AI slowdown are also circulating. Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman have both called for the industry to ease off. The St. Louis Fed found that without data center construction, the US economy would already be in a recession. Investors are being advised to check their portfolio exposure to AI and consider Treasury yields above 5 percent as a safer alternative. Separately, ServiceNow executive Paul Fipps said at Fortune's AIQ Summit that companies must redesign business processes rather than simply automate existing work, and that AI governance is becoming a business bottleneck.

In other developments, XtalPi and Fangda Carbon deployed a predictive AI model for graphite electrode production that is now fully operational. The model draws on Fangda Carbon's 60 years of manufacturing data and XtalPi's algorithmic capabilities to evaluate formulation candidates before physical trials, helping identify cost-saving blend proportions. Additionally, AI groups aligned with OpenAI and Anthropic spent $55.7 million on campaign advertising and outreach during midterm elections, mostly in primaries, supporting candidates across party lines while keeping their involvement largely hidden.

Key Takeaways

  • Amazon plans to transfer $8 billion in Nvidia Grace Blackwell chips to a special-purpose vehicle, offering investors up to a 10 percent equity stake and leasing the chips back.
  • Amazon expects $220 billion in capital expenditure this year, up from $131 billion in 2025, mostly funding its AWS cloud unit.
  • Investor Michael Burry warns AI spending by five major tech companies mirrors dot-com bubble risks, with hidden commitments exceeding $1.5 trillion off balance sheets.
  • California's "No Robo Bosses" law, effective July 1, 2027, requires human approval for AI-driven disciplinary decisions and imposes a $500 penalty per violation.
  • About 38 percent of organizations now run more than 100 AI agents, but only 9.5 percent secure more than 81 percent of them, according to a new security report.
  • OpenAI adopted Ory for identity management, a platform handling 1.2 billion weekly active users across its systems.
  • Global enrollment in AI security training courses surged 665 percent, though nearly 40 percent of businesses still lack adequate AI security expertise.
  • Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman have both called for the AI industry to slow down amid economic concerns.
  • The St. Louis Fed found that without data center construction, the US economy would already be in a recession.
  • AI groups aligned with OpenAI and Anthropic spent $55.7 million on midterm election advertising and outreach, mostly in safe primary seats.

Amazon plans to sell $8 billion in Nvidia chips to investors

Amazon is trying to sell about $8 billion of advanced Nvidia Grace Blackwell chips to outside investors. The company would create a special vehicle to hold the chips and then lease them back from that vehicle. This move would help Amazon reduce costs on its balance sheet. The chips are already being used in data centers across several US states.

Amazon in talks to move $8bn Nvidia chips off its balance sheet

Amazon has held talks with investors to transfer thousands of Nvidia Grace Blackwell chips into a special-purpose vehicle. The company would lease the chips back and offer investors up to a 10 percent equity stake. The chips are deployed in more than a dozen data centers across five US states including Nevada and Virginia. Amazon expects to spend $220 billion in capital expenditure this year, mostly on its cloud unit AWS.

Amazon to offload $8B of AI chips and lease them back

Amazon is planning to transfer about $8 billion worth of Nvidia Grace Blackwell accelerators to a special purpose vehicle. That vehicle would issue debt and a 10 percent equity stake to outside investors. Each Grace Blackwell accelerator costs between $60,000 and $70,000, suggesting Amazon would offload between 114,000 and 133,000 chips. Amazon would continue using the chips through a lease agreement after the transfer.

Amazon raises chip rental prices and explores $8 billion Nvidia leaseback

Amazon is raising AI chip rental prices by 15 percent and is also exploring a deal to move roughly $8 billion in Nvidia Grace Blackwell chips into a special-purpose vehicle. The chips already operate at facilities in five US states including Nevada and Virginia. Amazon's capital expenditure budget for 2026 is set at $200 billion, up from $131 billion in 2025. The company has been using various financing tools to fund its AI infrastructure expansion.

Investor warns tech AI spending may form a dangerous bubble

Investor Michael Burry warned that capital spending by five major tech companies on AI infrastructure resembles the risky trends seen before the dot-com bubble. He noted that these companies have hidden commitments off their balance sheets under controversial accounting rules. Burry said the investment level is the highest share of US GDP since the late 1990s tech peak. Several AI infrastructure companies have also paused or postponed their planned public offerings due to market concerns.

Investor warns AI spending mirrors risky dot-com era

Investor Michael Burry warns that AI spending by five major tech companies has reached levels similar to the dot-com bubble of the late 1990s. He says these companies have hidden commitments totaling over $1.5 trillion that do not appear on their balance sheets. Burry notes that investment now equals 2.07% of US GDP. Several AI infrastructure companies have paused or delayed their IPO plans. Public support for data center projects has also declined.

California's No Robo Bosses law requires human oversight

California Governor Gavin Newsom signed the No Robo Bosses law, set to take effect on July 1, 2027. The law requires that a human must verify and approve any AI-based decision to discipline or fire a worker. Employers must disclose if AI was used and provide data records upon request. Companies that violate the law face a $500 penalty per violation. The law aims to keep employers accountable for AI-driven workplace decisions.

China's AI investment reaches mid-sized companies

China's AI investment is expanding beyond large tech companies into mid-sized firms. The trend involves increased server purchases and data center construction. This shift was reported on October 2, 2026. The growth shows that AI infrastructure spending is becoming more widespread across different company sizes in China.

Seven AI agent security platforms to watch in 2026

A new report highlights seven platforms designed to secure AI agent identities and permissions. By April 2026, about 38% of organizations have more than 100 AI agents running. Only 9.5% of organizations secure more than 81% of their deployed agents. The report reviews platforms including Ory, which OpenAI adopted for identity management serving 1.2 billion weekly active users. The analysis covers identity management, access control, and deployment flexibility.

Global demand for AI security training surges

Global enrollment in AI security training courses rose by 665%, with U.S. enrollment up 358%. This growth far outpaces the 67% rise in all AI-related courses. Training for agentic AI workflows increased more than sixfold. However, nearly 40% of businesses still lack adequate AI security expertise. Experts stress the need for strong human oversight and zero trust principles in AI deployments.

AI slowdown could impact your money and retirement savings

Major AI leaders like Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman have called for the industry to slow down. An AI slowdown could hurt the US economy because spending on data centers and software drives growth. The St. Louis Fed found that without data center construction, the economy would already be in a recession. Investors are advised to check their portfolio exposure to AI and consider Treasury yields above 5% as a safer alternative.

ServiceNow exec says AI agents deliver life-changing results at Standard Chartered

Paul Fipps, president of global customer operations at ServiceNow, spoke at Fortune's AIQ Summit about companies using AI agents. He said some customers cannot rely on humans to keep AI agents in check. Fipps emphasized that companies must redesign business processes rather than just automate existing work. He also stressed that AI governance is becoming a business bottleneck and that trust drives speed in AI strategy.

XtalPi and Fangda Carbon deploy predictive AI for graphite electrode production

XtalPi and Fangda Carbon developed a raw material selection AI model that passed acceptance testing and is now operational. The model evaluates formulation candidates before physical trials, meeting targets for cost optimization and predictive accuracy. It combines Fangda Carbon's 60 years of manufacturing data with XtalPi's data engineering and algorithmic capabilities. The system helps identify cost-saving blend proportions and recommends formulation adjustments when supply or pricing changes.

AI groups spend millions influencing midterm elections

AI political groups aligned with OpenAI and Anthropic spent $55.7 million on campaign advertising and outreach. Most spending occurred in primaries, with $52.6 million spent on safe seats. Only seven of 95 reviewed ads mentioned AI directly, and none mentioned data centers. The groups supported candidates favorable to their agendas while hiding their involvement, and both OpenAI and Anthropic backed candidates across party lines.

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 Transfer Amazon AWS AI Spending Dot-com Bubble Comparison California AI Law AI Security AI Agent Management AI Industry Concerns AI Election Influence AI Predictive Model Graphite Electrode Production AI Governance AI Workplace Decisions AI Data Center Construction AI Economic Impact AI Investment Advice AI Executive Views AI Campaign Advertising AI Hidden Commitments

Comments

Loading...