Google is embarking on an ambitious initiative called "Project Suncatcher," planning to launch AI data centers into space. This project aims to power Tensor Processing Units (TPUs) on solar-powered satellites, orbiting 400 miles high to maintain constant sunlight. Google engineers believe constellations of about 80 such satellites could meet the escalating demand for AI computing power while significantly reducing the company's carbon footprint. Initial trials are set for early 2027 with the launch of two test satellites alongside Planet Labs, with the goal of making space-based data centers as affordable as Earth-based ones by the mid-2030s. These space systems could provide up to eight times more solar power than ground-based panels and lessen the need for land and water resources on Earth. Google's Trillium TPUs have already passed radiation tests, confirming their durability for a five-year mission. Meanwhile, the White House has blocked Nvidia from selling its most advanced AI chip, the Blackwell chip, to China. White House spokeswoman Karoline Leavitt confirmed this decision, citing national security concerns and tightening export rules on advanced technology. While President Trump had previously suggested discussing chip sales with Chinese President Xi Jinping, that conversation did not occur. Nvidia has been developing other chips for the Chinese market that comply with current export regulations, but the administration's latest stance indicates even these modified chips could face restrictions if deemed too powerful, potentially impacting Nvidia and the broader AI industry. In other developments, U.S. AI chip startup Tenstorrent plans to significantly expand its training program for Japanese engineers, aiming to accept 40 to 60 applicants in 2026, a more than fivefold increase from this year, reflecting the growing demand for AI chips. Separately, the Horizon Europe-funded RAIDO project is working on a platform for "Trustworthy and Green AI," focusing on efficient data use, energy-saving model designs, and ethical protections to reduce energy consumption during AI training and deployment, while also improving AI system explainability and bias detection.
Key Takeaways
- Google's "Project Suncatcher" aims to launch AI data centers with Tensor Processing Units (TPUs) into space using solar-powered satellites.
- Initial test satellites for Google's space AI project are scheduled for launch with Planet Labs in early 2027.
- Google anticipates that space-based AI data centers could become as affordable as Earth-based ones by the mid-2030s, offering up to eight times more solar power efficiency.
- The White House has blocked Nvidia from selling its advanced Blackwell AI chip to China due to national security concerns.
- White House spokeswoman Karoline Leavitt confirmed the ban on Nvidia's Blackwell chip, following ongoing discussions since August.
- President Trump had previously considered discussing chip sales with Chinese President Xi Jinping, but this discussion did not take place.
- The Biden administration is tightening export rules, potentially affecting even Nvidia's modified chips designed for China if they are considered too powerful.
- U.S. AI chip startup Tenstorrent plans to expand its training program for Japanese engineers to 40-60 applicants in 2026, a fivefold increase.
- The Horizon Europe-funded RAIDO project is developing a platform for "Trustworthy and Green AI," focusing on energy efficiency and ethical protections.
- RAIDO aims to reduce energy use during AI training and deployment, and enhance AI system explainability and bias detection.
Google plans solar powered AI data centers in space
Google engineers are exploring a bold plan to put AI data centers in space. This "Project Suncatcher" aims to meet the huge demand for AI computing power and lower Google's carbon footprint. They plan to launch Tensor Processing Units, or TPUs, on satellites into an orbit that keeps them constantly in sunlight. Google will launch two test satellites with Planet Labs in 2027, hoping this idea becomes affordable by 2035. This method could provide up to eight times more solar power than Earth-based panels.
Google aims for space based AI data centers
Google plans to launch AI data centers into space, with initial trials set for early 2027. Engineers believe constellations of about 80 solar-powered satellites, orbiting 400 miles high, can meet the growing demand for AI. This "Project Suncatcher" could become as affordable as Earth-based data centers by the mid-2030s due to lower launch costs. Space data centers would use solar panels that are up to eight times more efficient and reduce the need for land and water on Earth. Google sees this as a major step towards scaling AI computing in the future.
Google's Project Suncatcher aims for space AI
Google announced "Project Suncatcher," a bold plan to launch AI data centers into space using solar-powered satellites. This initiative aims to use constant solar energy to power AI chips, addressing the high energy needs of Earth-based data centers. Google's Trillium TPUs have already passed radiation tests, showing they can last for a five-year mission. The company plans to launch prototype satellites with Planet by 2027 to test the hardware in orbit. Google believes that by the mid-2030s, running data centers in space could cost about the same as on Earth.
White House blocks Nvidia's top AI chip sale to China
The White House announced that Nvidia cannot sell its most advanced AI chip, the Blackwell chip, to China. White House spokeswoman Karoline Leavitt confirmed this decision during a press conference. This follows ongoing discussions since August about whether a less powerful version of the chip could be sold. President Trump had previously suggested he might talk about the chips with Chinese President Xi Jinping, but that discussion did not happen.
US blocks Nvidia's advanced Blackwell AI chip for China
The White House announced Tuesday that Nvidia cannot sell its advanced Blackwell AI chip to China. The Biden administration is tightening export rules on advanced technology due to national security worries. Nvidia has been working on other chips for China that meet current export rules. However, the White House's latest statement suggests even these modified chips might be blocked if they are too powerful. This decision could significantly affect Nvidia and the wider AI industry.
Tenstorrent boosts Japan AI chip training for 2026
U.S. AI chip startup Tenstorrent plans to greatly increase its training program for Japanese engineers. In 2026, the company will accept 40 to 60 applicants, which is over five times more than this year. This expansion comes as the demand for AI chips continues to grow. Engineers in the program will travel to the U.S. for their training.
RAIDO Project creates green trustworthy AI
The RAIDO project, funded by Horizon Europe, is developing a platform for "Trustworthy and Green AI." This project combines efficient data use, energy-saving model designs, and ethical protections. It aims to reduce energy use during AI training and deployment, and improve how AI systems explain their decisions and detect bias. The Jo \
Sources
- Google wants to build solar-powered data centers
- Google plans to put datacentres in space to meet demand for AI
- Google has a ‘moonshot’ plan for AI data centers in space
- Nvidia cannot sell its most advanced AI chip to China, White House says
- Nvidia cannot sell its most advanced AI chip to China, White House says
- Tenstorrent to ramp up recruitment for Japan AI chip design training
- Energy-Efficient AI Models
Comments
Please log in to post a comment.