Summary4 min read

Podcast Summary

Reshaping Workflows with Dell Pro Precision and NVIDIA RTX Pro GPUs

Episode: Live from GTC: What Makes GB300 So Special with Matt Allard
Host: Logan Lawler
Guest: Matt Allard, NVIDIA
Date: March 17, 2026

Episode Overview

This episode explores the groundbreaking features of the newly released Grace Blackwell GB300, live from NVIDIA’s GTC 2026 event. Host Logan Lawler is joined by Matt Allard from NVIDIA for an energetic deep dive into what sets GB300 apart, how it leverages data center-level architectures within workstations, and its transformative impact on AI workflows and enterprise computing. The conversation includes firsthand insights on live demos, agentic AI, and the significance of running advanced models locally with full control and security.

Key Discussion Points & Insights

1. The Excitement Around GB300 at GTC (00:19–01:36)

Live Demos at GTC:
- Matt Allard highlights the buzz at the event, noting that Dell and NVIDIA are demonstrating GB300 on the Dell Pro Max in real time.
- Quote:
  - “We have a demo on the Dell Pro Max with GV300 really showing what it’s capable of and the fact that you’ve got this huge amount of power, really data center level performance and capability ... a game changer for folks that want to do AI on a local system, buying their firewall.” (Matt Allard, 00:46)
Local AI and Data Sovereignty:
- Logan points out the rising demand for on-premises AI solutions due to privacy, data sovereignty, and latency concerns.
- The GB300 addresses these needs by offering “20x the tops performance” and the ability to “run a billion parameter model at, you know, FB4.” (Logan Lawler, 01:11)

2. What Makes GB300 "Special"? (01:37–02:37)

Data Center Architecture in a Workstation:
- Allard explains that the GB300 isn’t just inspired by, but directly derived from, NVIDIA’s data center architecture—specifically, the Grace Blackwell platform.
- Quote:
  - “This is the exact same processors, CPU and GPU that you’d find in our NVL servers. So it really is data center level performance that comes with a huge amount of memory.” (Matt Allard, 01:44)
Memory and Model Capacity:
- GB300 features a massive “748 gigabytes of coherent memory,” enabling the loading of up to “a trillion parameter model.”
- This allows for experimentation with cutting-edge foundation models locally before deploying to full-scale NVL data centers.
- Quote:
  - “Really amazing to be able to experiment with big foundational models before you deploy.” (Matt Allard, 02:08)
Seamless Scalability:
- Thanks to architectural consistency, workloads can easily shift from the GB300 workstation up to Dell’s NVL data center products.
- Quote:
  - “Because it has the same architecture, it’s really easy to then shift up to the Dell data center products.” (Matt Allard, 02:27)

3. Keynote Highlights & Agentic AI (02:37–03:21)

Nemo Claw Launch and Agentic AI:
- Matt Allard singles out the debut of Nemo Claw, emphasizing its importance for the rise of agentic AI—AI agents capable of continuous, autonomous operation.
- Quote:
  - “Agentic AI is having an inflection point. It’s a moment that’s going to really drive people to use continuously operating agents...” (Matt Allard, 02:49)
  - Devices like “Dell Promax GV10 are perfect for that.”
Running AI Agents Locally:
- The new GB300 can seamlessly run lightweight agentic frameworks—Open Claw and Nemo Claw—enabling advanced experimentation and deployment directly on the device.
- Quote:
  - “You can run a lightweight device running Open Claw or Nemo Claw directly on that device and really get to see what’s possible with agents.” (Matt Allard, 03:08)

4. Open Claw, Nemo Claw, & Security (03:21–03:52)

Ease of Deployment & Security Controls:
- Logan mentions the complexity of permissions with Open Claw, to which Allard responds by emphasizing the ease and security enhancements with Nemo Claw and Open Shell.
- Quote:
  - “Minimally use Open Shell, which is now part of Open Claw. Open Shell gives you ... a sandbox, so it gives you permissions and policies that will keep it from running a little bit wild. ... Nemo Claw makes it ... a blueprint and an example of just set it and go.” (Matt Allard, 03:28)

Notable Quotes & Memorable Moments

On Data Center Power in a Workstation:
- “This is the exact same processors, CPU and GPU that you’d find in our NVL servers.” (Matt Allard, 01:44)
On Local AI Control:
- “People are not worried about the cloud, but they want to have, you know, sovereignty over their data.” (Logan Lawler, 01:11)
On Agentic AI’s Rise:
- “Agentic AI is having an inflection point. It’s a moment that’s going to really drive people to use continuously operating agents...” (Matt Allard, 02:49)
On Security in AI Agents:
- “Openclaw has some security issues and so running it with the Open Shell is great. Nemo Claw makes it even easier. It’s really a blueprint and an example of just set it and go.” (Matt Allard, 03:28)

Key Segment Timestamps

00:19–01:36 — Introduction to GB300 & Event Demos
01:37–02:37 — Data Center Architecture in Workstations; Memory & Scalability
02:37–03:21 — Keynote Recap; Agentic AI & Nemo Claw
03:21–03:52 — Security, Permissions, and Ease of Use in AI Agent Deployment
03:52–04:20 — Invitation to GTC Booth; Wrap-Up

Final Thoughts

This fast-moving episode offers a front-row view into how Dell Pro Precision and NVIDIA’s GB300 are redefining enterprise AI workflows by putting data center might into the hands of workstation users. With the flexibility to experiment at scale, heightened security, and exciting advances in AI automation via tools like Nemo Claw, listeners gain not just technical knowledge but a sense of the possibilities for the next era of high-performance AI.

Loading summary

Transcript16 lines

[00:00]
A
Foreign. Welcome to Reshaping Workflows with Dell Pro Precision and Nvidia, where innovation meets real world impact in high performance computing.
[00:20]
B
Welcome back. We're live from GTC 2026 with reshaping workflows with Delpro Precision and Nvidia RTX GPUs. But guess what, we're not talking about RTX GPUs right now with my good friend Matt Allard, former boss now at Nvidia, we're talking about Grace Blackwell. So Matt, first off, let's start with station GB300. What's the hottest news coming out of GTC this year around this product?
[00:46]
C
Well, the excitement about the product is that we're actually showing live demos on it. So we have a demo on the Dell Pro Max with GV300 really showing what it's capable of and the fact that you've got this huge amount of power, really data center level performance and capability is a game changer for folks that want to do AI on a local system, buying their firewall.
[01:11]
B
Correct. I mean I think we've all seen that, you know, GPUs have gotten more powerful, people are not worried about the cloud, but they want to have, you know, sovereignty over their data. They want it to reside locally, they don't want latency and all that. So I mean the GB300 is a perfect, you know, answer. It's 20x, you know, the tops performance or run a billion parameter model at you know, FB4. But let me ask you, you're kind of on the inside, right?
[01:37]
C
Yeah.
[01:37]
B
What else makes, you know, as you call station, we call GB300 kind of the same thing. What makes that product special?
[01:45]
C
Really? What makes that product special is that it derived from our data center architecture. Race Blackwell. If you look at the Spark, your version of the promax with GB10, that's kind of a shrunk down version of our data center products really reduced to a system on a chip. This is the exact same processors, CPU and GPU that you'd find in our NVL servers. So it really is data center level performance that comes with a huge amount of memory. So 748 gigabytes of coherent memory, that means you can fit up to a trillion parameter model. Really amazing to be able to experiment with big foundational models before you deploy. And of course because it has the same architecture, it's really easy to then shift up to the Dell data center products. NVL Data Center I couldn't have said it better.
[02:37]
B
This is why I was my Former manager. The last question I want to end it with. Did you watch the keynote today?
[02:43]
C
I did.
[02:44]
B
What was your biggest aha moment or the favorite thing that for you coming out of the keynote?
[02:50]
C
So I'm so focused on the PC client products. I was really excited about the new Nemo Claw capability. That is really amazing because Agentic AI is having an inflection point. It's a moment that's going to really drive people to use continuously operating agents and things like The Dell Promax GV10 are perfect for that because you can run a lightweight device running Open Claw or Nemo Claw directly on that device and really get to see what's possible with agents.
[03:21]
B
I'll tell you what's possible because I've set up openclaw. You got to set your permissions. That's why you need Nemo Claw.
[03:28]
C
That's right. Yeah, that's right. So you could minimally use Open Shell, which is now part of openclaw. Open Shell gives you essentially a sandbox, so it gives you permissions and policies that will keep it from running a little bit wild. Openclaw has some security issues and so running it with the Open Shell is great. Nemo Claw makes it easy, even easier. It's really a blueprint and an example of just set it and go.
[03:53]
B
I love it. So, Matt, always a pleasure talking to you. If you happen to be at GTC, stop by the Dell booth, booth 714 to see one of the only working Open Claw slash Nemo Claw SL Open Shell SL Jensen special demos that are running out there right now. And with that, it's a wrap.
[04:21]
A
Do what you want. Do what you want. Do what you this podcast was produced
[04:35]
C
in partnership with Amaze Media Labs.