AI ·
What hardware you need to run an LLM yourself
How model size maps to VRAM, why bandwidth sets token speed, and which GPU or server to rent for 8B to 284B models, with the arithmetic shown.
Notes on hosting, security and infrastructure from the Server Pals team.
AI ·
How model size maps to VRAM, why bandwidth sets token speed, and which GPU or server to rent for 8B to 284B models, with the arithmetic shown.
AI ·
Grid connections, liquid cooling and 100 kW racks now decide who can host AI workloads. What the constraints are in 2026 and what to ask a provider before you sign.
Where the Server Pals catalog stands, how a provider gets onto the site, how a listing is built, and what we are working on adding next.
AI ·
The best AI models in October 2026: Claude, GPT-6, Gemini 4 and Grok API prices, plus the open-weight models you can self-host, with sizes and licences.
Industry ·
Control panel price rises, a registry change, gigawatt campuses and grid politics: what moved in hosting during October 2026 and what it means for anyone buying a server this quarter.
Security ·
The flaws CISA confirmed under active attack in October 2026, from Citrix and Ivanti gateways to backup and media servers, and the patch windows that make sense for a hosting operation.
Security ·
CVE-2026-109618 and CVE-2026-109619 were patched on October 21, 2026 after the details leaked. What they are, why the config workaround is not enough, and how to check your servers.