Microsoft Releases Windows Server 2028 Preview with Local AI Inference
The default-on NVMe change alone justifies evaluation, since Windows Server 2025 required administrators to opt in manually and the reported gains are material for database and virtualisation hosts. IT teams planning on-premises AI deployment should assess whether the limited early-access hybrid AI features fit their workloads and test vLLM or Llama.cpp integration on their own hardware. Compared with Linux-based inference stacks, Windows shops now get a native path instead of workarounds, which matters for enterprises across ASEAN building private data centres.
IT之家Source