AI servers

A new open-source project called Hob is being presented as an on-device alternative to Claude Cowork and ChatGPT. It runs locally rather than in the cloud, appealing to users who want AI assistance without sending data to external servers. Discussion is centred on its privacy advantages and how it compares with established commercial AI assistants.

A new guide ranks the top inference servers for deploying open-source AI models in production environments. It compares options such as vLLM, Triton Inference Server, and Text Generation Inference on performance, scalability, and ease of use, offering engineering teams practical advice for choosing infrastructure as open-source model adoption grows in real-world applications.

Attention is turning to the water footprint of data centers, the facilities that power cloud computing and artificial intelligence. Servers generate heat that must be cooled, and many cooling systems rely on evaporating large volumes of water. As new centers are built near residential areas, questions are growing about the strain on local supplies and the lack of public disclosure about usage.

A security challenge dubbed PHP2Shell offered a Bitcoin reward to anyone who could break into an unpatched server running PHP. A hacker successfully exploited the server and claimed the bounty. The challenge was designed to test claims that AI tools now make it easy to exploit known vulnerabilities, and the successful breach has reignited debate over how AI is changing offensive security.

Reports say thieves are stealing high-value AI data center equipment, such as GPUs and servers, and reselling it for large profits. The trend highlights how the AI hardware boom has turned components into prized targets, raising security concerns for companies building out data centers to meet surging demand.