PLUS: IBM's $11B bet, a Tesla robot's suspicious collapse, and Michael Burry calls OpenAI the next Netscape
Good morning
In a significant reversal of its tech policy, the United States will now allow Nvidia to sell its powerful H200 AI chips to approved customers in China. The move marks a strategic compromise in the ongoing tech trade war between the two nations.
This policy signals a shift from outright restriction to managed competition with Beijing. The key question is whether this new approach can secure U.S. economic interests without ceding a critical technological advantage in the global AI race.
In today’s Next in AI:
US reverses ban on Nvidia chips for China
A Tesla robot's suspicious on-stage collapse
IBM's $11B bet on real-time AI data
Trump's plan for a federal AI rulebook
The Great Chip Thaw
The Recap: In a major policy reversal, the White House will now allow Nvidia to sell its powerful H200 AI chips to approved customers in China. The move aims to find a compromise in the ongoing tech trade war.
Unpacked:
This decision follows a failed attempt to export a less powerful chip, the H20, which Beijing blocked, giving domestic companies like Huawei more room to grow.
As part of the deal, the U.S. government will receive a 25% cut of the revenue from Nvidia’s H200 sales in China, a condition that will also apply to other American firms like AMD and Intel.
National security experts remain divided, with some arguing the original ban was critical for slowing China’s military AI development while others believe it has failed to stop progress from Chinese companies like DeepSeek.
Bottom line: This policy re-opens a massive market for Nvidia, positioning its technology to remain the global standard even as it keeps the most advanced chips off the table. It signals a strategic shift from total restriction to managed competition in the global race for AI dominance.
Tesla's Robot Reality Check
The Recap: A viral video from a Tesla event shows an Optimus robot suspiciously collapsing after making a hand motion consistent with an operator removing a VR headset, exposing the gap between AI hype and reality.
Unpacked:
The robot's "phantom headset" motion strongly suggests it was remotely teleoperated by a human, not acting autonomously as the event's "autonomy" theme implied.
This follows a pattern of misleading demos, including a 2024 video of Optimus folding laundry and another event where robots pouring beer were later confirmed to be human-controlled.
The failure underscores the immense difficulty of true robotic autonomy, a challenge companies like Figure are also tackling, often with more transparency about their development process.
Bottom line: The incident highlights the critical difference between polished demos and genuine autonomous AI, a crucial distinction in a field filled with hype. It serves as a reminder to look past the performance and question the underlying technology driving these humanoid robots.
IBM's $11B Bet on Real-Time AI Data
The Recap: IBM is acquiring data-streaming platform Confluent for an enterprise value of $11 billion, a major move to bolster its AI services with real-time data processing capabilities.
Unpacked:
The acquisition addresses the growing need for AI tools to process information in real time, a critical function for modern enterprise applications.
With a total enterprise value of $11 billion, this marks one of IBM's largest takeovers and a significant investment in its AI strategy.
Both companies anticipate the major acquisition will be finalized and closed by mid-2026.
Bottom line: This move shows IBM is aggressively pursuing the foundational data infrastructure necessary for advanced AI. By owning Confluent, the company can offer its enterprise clients a more integrated solution for building powerful, data-hungry models.
Trump's 'One Rulebook' for AI
The Recap: President Trump confirmed he will sign an executive order to create a single federal framework for AI regulation, effectively preempting states from creating their own rules. He announced his plan on Truth Social, arguing a unified approach is essential to prevent AI from being “destroyed in its infancy.”
Unpacked:
The administration’s goal is to avoid a complex patchwork of 50 different state-level AI laws, which tech leaders argue could slow down innovation and hinder America’s global competitiveness.
The move faces significant bipartisan pushback, with critics and safety groups warning it could shield AI companies from accountability. Florida Governor Ron DeSantis called the effort “federal government overreach” that would subsidize Big Tech.
This executive action follows after Congress killed an earlier attempt by Republicans in July to enact a 10-year moratorium on state AI laws, signaling a shift in strategy from legislative to executive power.
Bottom line: This executive order sets the stage for a major clash between the federal government's push for rapid innovation and the states' efforts to protect consumers. The resulting legal and political battles will likely define the landscape of AI governance in the U.S. for years to come.
AI Pulse
Michael Burry predicted OpenAI is the "next Netscape," calling the AI leader "doomed and hemorrhaging cash" and suggesting its continued funding is a ploy to set up a massive IPO.
Martian launched a $1 million prize to advance AI interpretability research, aiming to move the field from "alchemy" to a scientific understanding of how models work.
OpenAI renamed its “Cameo” deepfake feature to “Characters” in the Sora app following a trademark lawsuit from the celebrity video platform of the same name.