Power BI local AI ran on an 8GB M1 Mac mini, but slowly

Power BI and Fabric work often raises two concerns when are involved: token cost and data privacy. Power BI MCP server was connected to a locally running Nvidia model on a 2020 M1 Mac mini. The machine had only 8 GB of memory and is the kind of older Mac mini that can be found used for about $300 to $400.

The setup worked, but performance was poor. The weak point appears to be the old Mac mini and its small memory, not necessarily the whole local-AI idea. Newer laptops with 128 GB of memory and AMD’s Unified , priced around $3,500, were presented as more realistic machines for running these AI tools fully locally inside a company.

That price could still make sense for a large business if employee laptops already cost a few thousand dollars and reduces privacy and usage-cost concerns.

Key points

  • A 2020 M1 Mac mini with 8 GB of memory ran Power BI MCP server with a local Nvidia model.
  • The used Mac mini price was estimated at about $300 to $400.
  • The setup worked, but it did not run well on that low-memory machine.
  • Local AI was framed as a way to reduce token cost and data privacy concerns.
  • Newer 128 GB memory laptops were suggested as more realistic hardware for local company AI tools.
Read original