Uploaded March 2026 | Updated November 2026, 1 day ago
How do you bring the power of AI to your desk without compromising your data?
In this episode of the Plugged In Podcast, we sit down with Plugable Founder and CTO Bernie Thompson for a deep dive into the hardware requirements of the "Local AI" revolution. As the industry shifts from public cloud models to private, local execution, the demands on our hardware are changing—specifically regarding GPU power, VRAM, and high-speed connectivity
This conversation serves as the technical foundation for our upcoming TBT5-AI (Thunderbolt 5 AI Enclosure). Bernie explains why Thunderbolt 5 is the critical interface required to bridge the gap between a standard Windows laptop and the high-end GPU performance needed to run 120B parameter models locally.
What we discuss in this episode:
- The TBT5-AI Vision: Why a dedicated AI enclosure with 800W of power is necessary for production-grade local AI.
- Local vs. Cloud: The security, cost, and privacy benefits of keeping data on-premises for legal, financial, and government sectors.
- VRAM vs. Throughput: Why GPU memory capacity is the most important variable for running large-scale open-source models versus traditional gaming.
- Plugable Chat: How our new software "harness" enables smaller models to perform complex SQL queries and document retrieval effectively.
- The Engineering Mission: How Plugable is applying 15+ years of Thunderbolt expertise to solve modern AI compatibility challenges.
🕒 Chapters
00:00 Introduction: The Shift to Local AI
02:20 CTO Perspective: Why Thunderbolt 5 is the Key to AI
04:47 The Privacy Case for Local AI
07:26 Corporate Data Security: Avoiding the "Cloud Trap"
12:03 Inside the TBT5-AI: Engineering an 800W Enclosure
17:13 AI vs. Gaming: Why VRAM Matters More Than Throughput
22:28 Plugable’s Mission: Solving Hardware Compatibility
24:54 Introducing Plugable Chat: The Software Harness
30:54 Technical Innovations in SQL & Knowledge Retrieval
34:42 The Future: Scaling Local AI for the Enterprise
37:10 Final Thoughts: AI and the Future Desk
How do you bring the power of AI to your desk without compromising your data?
In this episode of the Plugged In Podcast, we sit down with Plugable Founder and CTO Bernie Thompson for a deep dive into the hardware requirements of the "Local AI" revolution. As the industry shifts from public cloud models to private, local execution, the demands on our hardware are changing—specifically regarding GPU power, VRAM, and high-speed connectivity
This conversation serves as the technical foundation for our upcoming TBT5-AI (Thunderbolt 5 AI Enclosure). Bernie explains why Thunderbolt 5 is the critical interface required to bridge the gap between a standard Windows laptop and the high-end GPU performance needed to run 120B parameter models locally.
What we discuss in this episode:
- The TBT5-AI Vision: Why a dedicated AI enclosure with 800W of power is necessary for production-grade local AI.
- Local vs. Cloud: The security, cost, and privacy benefits of keeping data on-premises for legal, financial, and government sectors.
- VRAM vs. Throughput: Why GPU memory capacity is the most important variable for running large-scale open-source models versus traditional gaming.
- Plugable Chat: How our new software "harness" enables smaller models to perform complex SQL queries and document retrieval effectively.
- The Engineering Mission: How Plugable is applying 15+ years of Thunderbolt expertise to solve modern AI compatibility challenges.
🕒 Chapters
00:00 Introduction: The Shift to Local AI
02:20 CTO Perspective: Why Thunderbolt 5 is the Key to AI
04:47 The Privacy Case for Local AI
07:26 Corporate Data Security: Avoiding the "Cloud Trap"
12:03 Inside the TBT5-AI: Engineering an 800W Enclosure
17:13 AI vs. Gaming: Why VRAM Matters More Than Throughput
22:28 Plugable’s Mission: Solving Hardware Compatibility
24:54 Introducing Plugable Chat: The Software Harness
30:54 Technical Innovations in SQL & Knowledge Retrieval
34:42 The Future: Scaling Local AI for the Enterprise
37:10 Final Thoughts: AI and the Future Desk










