Toggle light / dark theme

Nvidia scandal may test fragile US-China truce, experts say

U.S. Commerce Dept looks to cut off China’s current ability to remotely control Nvidia chips outside China.

The geopolitical stakes around frontier AI compute just escalated significantly. 🏛️⚡️

While physical hardware embargoes (restricting shipping A100/H100/H200 accelerators to Chinese destinations) have been active under U.S. Export Administration Regulations (EAR), enforcing remote cloud access remained a major enforcement friction point.

Here is the breakdown of why this shift is happening now:

1. The Compute-as-a-Service Loophole.

Chinese AI labs and tech entities have routinely leased compute clusters housed in neutral or third-party jurisdictions (e.g., Southeast Asian cloud nodes). Because the silicon physically resides outside China, no raw hardware crosses restricted borders—yet foreign engineers execute training runs remotely.

2. Shift in Lobbying Dynamics.

Nvidia has historically pushed back against broad export bans to protect global commercial revenue and maintain market presence. However, following recent intelligence reports and congressional inquiries into data-center intermediaries (like Megaspeed), Capitol Hill’s appetite for tighter enforcement has expanded.

3. Next-Gen Technical & Regulatory Mechanics.

* Silicon Telemetry & Geofencing: Hardware-enforced remote attestation via on-chip Secure Enclaves.

* Strict KYC for IaaS Providers: Standardizing Know-Your-Customer mandates for global Infrastructure-as-a-Service vendors to prevent proxy compute leasing.

* Extraterritorial FDPR Enforcement: Expanding Foreign Direct Product Rules to cover third-party data centers hosting American AI architectures.

The Bigger Strategic Takeaway:

As enforcement shifts from physical customs checks to silicon-level tracking and remote access restrictions, domestic compute alternatives in China (e.g., Huawei’s Ascend platform + CANN stack) will face intense pressure to rapidly close the performance gap.

#ArtificialIntelligence #Semiconductors #NVIDIA #Geopolitics


Calls in Washington grow to tighten semiconductor export controls as next week’s Trump-Xi Apec meeting takes on added significance.

Caltech robot makes history by building a structure in space

For the first time, a robot has successfully assembled part of a structure in space without astronauts, demonstrating a technology that could dramatically change how large spacecraft and scientific instruments are built.

The robotic system was developed by researchers at the California Institute of Technology (Caltech) with funding from the Defense Advanced Research Projects Agency (DARPA). It is part of DARPA’s Novel Orbital and Moon Manufacturing, Materials, and Mass-efficient Design (NOM4D) program, an initiative exploring ways to manufacture and assemble enormous structures directly in space.

Rather than relying on astronauts to connect individual pieces, the robot uses an automated arm to retrieve structural tubes and connecting joints from nearby storage dispensers. It then joins the components into repeating sections, called bays, that form a larger supporting framework known as a truss.

Fabric molds could bring material-saving concrete beams within reach

Topology optimization is a computer design method that can help engineers design beams using reduced amounts of concrete and steel. But those intricate shapes produced from the tool are often too complex or expensive to pour into conventional rigid molds. The few that have been built have mostly relied on specialized equipment, such as concrete 3D printers or machine-carved molds, equipment that is out of reach in regions where new construction is expected to grow.

Now, MIT engineers have shown that highly efficient beams can be cast in something more accessible: sewn fabric. In a paper published in the Journal of Building Engineering, the team designed and built a 3-meter (about 10-foot) test beam using fabric molds created through an automated pattern-making process. Compared with a conventional beam designed to carry the same weight, the optimized beam used about one-third as much concrete and half as much steel reinforcement. When tested, its measured response was within about 2% of what the team’s computer model predicted.

Fabric formwork, in which concrete is poured into a mold made of textile, has been explored for years to create beams that vary in depth, rather than the rectangular shapes common in construction. But because fabric deforms under the weight of wet concrete, that flexibility has made the final shape hard to predict.

A single AI model simulates the behavior of many substances, from sliding sand to folding proteins

Today, a scientist who wants to simulate a physical process must pick the right tool for that one job, such as a solver built for flowing water or another designed to model the way a protein folds. Each one is hand-coded around the specific equations that dictate how that process behaves.

Researchers at MIT, the University of British Columbia (UBC) and elsewhere have now replaced that patchwork with a single deep-learning system, called WorldParticle, that could help engineers model different physical phenomena, from sliding sand to folding proteins, with one unified artificial intelligence architecture.

Scientists and engineers could use this system to model processes that typically require separate simulators, such as a fluid sloshing against a deformable solid. This could offer a faster, more accurate way to simulate how sloshing oil might impact an elastic valve in a storage tank, for instance. It could also help engineers study where and how granular materials, like sand or soil, could move when a soft robot touches them.

FLEX approach helps robots reason beyond literal commands

Humans are intuitive by nature. We can sense meaning beyond the literal words that we encounter. Now, a new artificial intelligence–assisted learning platform developed by MIT engineers is enabling robots to read larger meaning into a human’s natural language commands.

In a study to be presented at the annual Conference on Robot Learning (CoRL 2026), held in Austin, Texas, Nov. 9–12, researchers from MIT and New York University introduce FLEX, a flexible feature-editing pipeline for robot motion planning.

With FLEX, humans can interact with a robot by giving it instructions in natural, conversational language. The robot can not only follow the instructions literally but also read into the words to draw larger generalizations that help to improve its motion-planning task.

Where Will The World Store Its Fast Growing Volume Of Data?

Governments contribute too. NIST and other standards organizations will help develop security, interoperability, cryptography, and resilience frameworks. As data infrastructure expands into space and international waters, we will need international cooperation.

The emerging storage ecosystem will create new markets and investment opportunities in advanced materials, photonics, biotechnology, satellite infrastructure, energy, cooling, cybersecurity, and data management.

The AI revolution is transforming information consumption. Data is no longer a business and technology byproduct. It powers artificial intelligence, economic competitiveness, and national security. The colloquial phrase “data is the new gold ring” is true.

In rain and darkness, how should self-driving cars trust their sensors?

Many of today’s self-driving cars “see” the road through two senses at once. Cameras capture color, texture and the outline of a pedestrian, while light detection and ranging (LiDAR) fires laser pulses to measure exactly how far away that pedestrian stands. Together, they provide complementary information and outperform either sensor alone.

This pairing has a weakness, however. Trained mostly on clear daytime footage, the fusion model learns to lean on whichever sensor performs best there and may fail in challenging scenarios. Rain can scatter LiDAR beams, for instance, while a night drive leaves cameras battling darkness and glare.

To address this, assistant professor Zhao Na at Singapore University of Technology and Design (SUTD) led two studies that built different models. In “CCF: Complementary collaborative fusion for domain generalized multi-modal 3D object detection,” the research team rebalances how a detector weighs its sensors so it holds up in new surroundings. In “PanDA: Unsupervised domain adaptation for multimodal 3D panoptic segmentation in autonomous driving,” they harness unlabeled data from a new environment to help the model adapt once it arrives.