Nvidia and startup Valar Atomics are exploring a novel approach to power and cool AI data centers, aiming for minimal local water consumption. This initiative combines Valar Atomics' helium-cooled microreactors with Nvidia's closed-loop cooling systems to meet the immense power demands of AI infrastructure while significantly reducing water usage. The goal is to address the substantial water consumption of traditional data centers, particularly in water-stressed regions, and mitigate their environmental impact.

An initial demonstration in Utah showcased Valar Atomics' Ward-250 reactor powering Nvidia's Blackwell architecture hardware. However, this demonstration was on a small scale, and scaling up to a commercial 30-megawatt data center presents significant challenges, including regulatory approvals and economic viability. Nevertheless, this collaboration highlights a clear trend among AI companies to secure dedicated, stable energy sources and reduce their environmental footprint.

👉 Read the full article on Gizmodo en Español

Anthropic is reportedly moving towards developing its own AI chips to reduce its reliance on major providers like Nvidia and Google. The company is said to be in discussions with Samsung Electronics for the manufacturing of a custom AI chip. This strategic shift reflects the understanding that in the competitive AI landscape, access to sufficient computing power for training and running models is as crucial as the models themselves.

The move is partly influenced by competitors like OpenAI, which recently unveiled its inference chip, "Jalapeño," developed with Broadcom. By developing its own silicon, Anthropic aims to gain advantages in cost per query, speed, energy efficiency, and deployment capabilities for its Claude models. While seeking to diversify its compute stack, Anthropic emphasizes that it will continue to leverage chips from existing partners such as Google, Amazon, and Nvidia, rather than severing those relationships.

👉 Read the full article on Gizmodo en Español

A new study proposes that wastewater could serve as an alternative source for critical minerals such as lithium, magnesium, and titanium. Researchers from Northeastern University, Yale, and other institutions analyzed ten non-traditional water sources, including oil and gas production water, coal power plant water, desalination plant brine, abandoned mine water, industrial effluents, and municipal wastewater. Recovering these minerals, essential for clean energy technologies, could help address issues like concentrated supply chains and environmental impact.

The study, published in Joule, found that magnesium and lithium were the most abundant critical materials, with uranium, titanium, fluorine, and silicon also present in medium to high concentrations. This approach aligns with the principles of a circular economy, complementing traditional mining rather than replacing it. A significant challenge remains in developing selective and energy-efficient recovery technologies, as these minerals are often present in low concentrations and mixed with other compounds.

👉 Read the full article on Gizmodo en Español